Professional Documents
Culture Documents
ICTD 2009 Proceedings
ICTD 2009 Proceedings
PROCEEDINGS
April 17-19, 2009
Carnegie Mellon University in Qatar
Education City, Doha, Qatar
Host Organization:
Organizing Partner:
Platinum Sponsors:
Gold Sponsors:
Bronze Sponsors:
Media Partner:
Technical Sponsors:
Table of Contents
Introduction 1
Political Incentives and Policy Outcomes: Who Benefits from Technology- 173
Enabled Service Centers?
Jennifer Bussell
ICT Governance in Higher Education: Case Study of the Rise and Fall 348
of Open Source in a gulf University
Sofiane M. Sahyraoui
ICTD State of the Union: Where Have We Reached and Where Are We 357
Headed?
Rabin Patra, Joyojeet Pal, Sergiu Nedevschi
Global Youth Connectivity (GYC) – Using ICT for Peaceful Recovery 485
and Long-term Change
Anne Bertrand
Conference Chair
Advisory Board
Publications Chairs
Poster Chairs
Demo Chairs
Regional Chairs
Africa
Dr. Gary Marsden Dr. Tim Waema
University of Capetown, South Africa University of Nairobi, Kenya
Middle East
Dr. Adnan Abu Dayya Dr. Fouad Mrad
Qatar University American University of Beirut
East Asia
Dr. Jack Linchuan Qiu Rinalia Abdul Rahim
The Chinese University of Hong Kong Global Knowledge Partnership, Malaysia
South Asia
Anita Gurumurthy Dr. Umar Saif
IT for Change, India Lahore University of Management
Sciences, Pakistan
North America
Dr. John Bennett Claudia Morrell
University of Colorado at Boulder Multinational Development of Women
in Technology
Europe
John Traxler
University of Wolverhampton, UK
Oceania
Christina Higa Dr. Esther Batiri Williams
University of Hawaii The University of the South Pacific
Publicity Coordinators
Logistics Coordinators
Renee Barcelona Sarah Belousov
Carnegie Mellon University in Qatar Carnegie Mellon University
Kara Nesimiuk Ermine Teves
Carnegie Mellon University in Qatar Carnegie Mellon University
Website Coordinators
ICTD is the premier series of scholarly conferences on the use of ICTs for
development, spanning technical and social science domains. The call for papers
attracted a record 250 submissions. All papers were put through a double-blind peer-
review process. The Program Co-Chairs assigned papers to our Senior Program
Committee members who oversaw a review process involving three Program
Committee reviewers per paper: one with deep expertise about the subject matter;
another with broad background in the area; and one drawn from an altogether
different discipline. Our continuing hope is that this encourages a convergence of
vocabulary and ideas within the ICTD field, while maintaining the integrity of
different disciplines. The Senior Program Committee members then meta-reviewed
the papers. Authors were allowed a brief rebuttal to reviewer comments before final
acceptance decisions were made and revisions were finalized.
Ultimately, 19 papers were selected for oral presentation, and another 27 papers were
chosen as full papers for poster presentation; an acceptance rate of just over 18%.
These papers represent some of the best work being done in ICTD today. They focus
on a wide variety of development goals, and involve a broad and innovative range of
digital technologies. They draw from all continents of the global South, and focus on
all stages of the ICTD lifecycle: from readiness through design and adoption to use
and impact. They also tell us about all levels, from the individual through
communities and projects to ICTD programmes and policies. We hope that you will
find them an insightful, provocative, and informative contribution to our fast-growing
field of research and practice.
We also need to thank a number of people without whom the program could not have
been put together. First, we wish to thank our Honorary Chairs, Dr. Hessa Sultan Al-
Jaber, Secretary General of The Supreme Council of Information and Communication
Technology (ictQatar), and Dr. Kentaro Toyama, Assistant Managing Director of
Microsoft Research India (MSRI) for their tremendous support in making this
conference a success. We are of course deeply indebted to our Program Committee:
those who did the hard work of reviewing and our senior PC members who managed
the review process so effectively; all together it is they who ultimately steer the course
of ICTD. We thank the Regional Chairs for their assistance in promoting the
conference, and our Advisory Board which provided guidance and moral support. We
also thank the Publication Chairs, Yonina Cooper and Thrishantha Nanayakkara, who
2
made this proceedings possible; and Faheem Hussain and Tapan Parikh, who carried
the load in organisation of the poster presentations.
The conference program for ICTD2009 went well beyond papers, and was
significantly enhanced by demonstrations, organised by Bill Thies and Khaled Harras,
and by a series of panels and workshops, organised by Joe Mertz and Joyojeet Pal.
We are very grateful for their input and hard work.
This conference would not have been possible without the tireless efforts of a number
of organizers and volunteers, notably our local organizing committee chairs Elaine
Farah, Murry Evans, and Erin Stewart, and the many others on the local organizing
committee, especially Dean Charles Thorpe, Sarah Belousov, Ermine Teves, Renee
Barcelona, Kara Nesimiuk, Andy Zrimsek, Noha Al Afifi, Shams Hassan, Aaron
Lyvers, Shamila Khader, Ray Corcoran, Bob Gaus, Freddie Dias, Daniel Freeman,
and Jim Gartner. If there are others we did not name explicitly, it is our lapse.
The success of the conference is in part due to our many sponsors and partners. We
are extremely grateful to our organizing partner, The Supreme Council of Information
and Communication Technology (ictQatar), to our media partner Al Jazeera
Children’s Channel (JCC), to our technical sponsors IEEE and ACM, and to our
financial sponsors, the Qatar National Research Fund (Platinum Sponsor); Canada’s
International Development Research Centre (Platinum Sponsor); Qatar Telecom
(Platinum Sponsor); ExxonMobil (Gold Sponsor); Microsoft Corporation (Gold
Sponsor); IBM (Bronze Sponsor); the Computer Science program at Carnegie Mellon
University in Qatar (Bronze Sponsor); and other sponsors who asked not to be
publicly acknowledged.
Finally, we are indebted to Carnegie Mellon University in Qatar for hosting the
conference at their campus in Education City, and to the numerous employees of
Carnegie Mellon University (in both the Pittsburgh and Doha campuses), especially
the TechBridgeWorld team, who truly made this conference a success.
Thanks to everyone who contributed to ICTD2009 in so many ways, and thank you
for participating! We are confident you will find the following papers, abstracts, and
information not only interesting and useful, but the seeds for further research,
innovation, and developmental impact.
Abstract—Is the dramatic increase in Internet use associated democracy and regime transitions literature. To be sure, “the
with a commensurate rise in democracy? Few previous studies trouble with the zealots of technology as an instrument of
have drawn on multiple perception-based measures of governance democratic liberation is not that they misconceive technology
to assess the Internet’s effects on the process of democratization.
This paper uses perception-based time series data on “Voice & but that they fail to understand democracy” [2]. ‘In other
Accountability,” “Political Stability,” and “Rule of Law” to pro- words, “it turns out there is no simple general answer to
vide insights into democratic tendency. The results of regression the question: Is the technology democratizing?’ until we have
analysis suggest that the level of “Voice & Accountability” in a made clear what sort of democracy we intend.” We address this
country increases with Internet use, while the level of “Political question first before proceeding with a more detailed literature
Stability” decreases with increasing Internet use. Additionally,
Internet use was found to increase significantly for countries review.
with increasing levels of “Voice & Accountability.” In contrast, Barber’s notion of “strong democracy” comprises the careful
“Rule of Law” was not significantly affected by a country’s level and prudent judgment of citizens who participate in deliber-
of Internet use. Increasing cell phone use did not seem to affect ative, self-governing communities. Schmitter and Karl write
either “Voice & Accountability,” “Political Stability” or “Rule that, “modern political democracy is a system of governance
of Law.” In turn, cell phone use was not affected by any of
these three measures of democratic tendency. When limiting our in which rulers are held accountable for their actions in the
analysis to autocratic regimes, we noted a significant negative public realm by citizens, acting indirectly through the compe-
effect of Internet and cell phone use on “Political Stability” and tition and cooperation of their elected representatives”[3]. The
found that the “Rule of Law” and “Political Stability” metrics two authors emphasize that citizens are the most distinctive
drove ICT adoption. element in democracies. “All regimes have rulers and a public
Index terms—cell phone, democracy, fixed effects model, ICT,
realm, but only to the extent that they are democratic do
internet
they have citizens” [3]. In contemporary studies of democracy
and particularly in pluralist theory, “a vibrant civil society
I. I NTRODUCTION is usually regarded as an essential for good governance
Does the globalization of the Internet have a democratizing and effective democratic consolidation” [4]. In other words,
effect? The question has already been posed by numerous regular elections are not sufficient. As Zakaria noted, illiberal
studies but these have largely taken the form of qualitative case democracies have free elections but citizens remain cut off
studies and/or large theoretical analyses. In terms of a rigorous, from real power due to the lack of civil liberties [5].
quantitative establishment of the democratization effects of the Huber et al. write that the most basic feature of democracy
Internet, however, the jury is still out [1]. At the heart of is power sharing [6]. They identify three clusters of power
this debate, moreover, lies a more fundamental question about as primarily relevant for the chances of democracy: (1) the
the essence of democracy. In fact, “unless we are clear about balance of power in civil society; (2) the balance of power
what democracy means to us, and what kind of democracy we between state and society; and (3) the transnational balance of
envision, technology is as likely to stunt as to enhance the civic power that shape the first two and constrain political decision-
polity” [2]. The purpose of this paper is to contribute more making. By remaining diverse and independent of the state,
rigorous data-driven analysis to the literature on Internet and political participation by civil society acts as a channel of
democracy since “there is no doubt that rigorous and data- public voice and accountability, and a way of challenging
driven analysis of this relationship will benefit scholars and and checking the unbridled power of authoritarian regimes
policymakers alike” [1]. [4]. The structure of state-society relations is equally relevant
Previous research on the topic of Internet and democracy for democracy. As Huber et al. note, “the power of the state
can be characterized as lacking a serious perusal of the needs to be counterbalanced by the organizational strength of
4
the civil society to make democracy possible; the state must when Internet access and cell phone use is significantly more
not be so strong and autonomous from all social forces as prevalent and globalized than in the 1990s. If a statistically
to overpower civil society and rule without accountability.” significant relationship between Internet and democracy does
Clearly then a governing body that fails to follow the “rule exist, then it is more likely to manifest itself now and not in the
of law,” should not be considered democratic [3]. These 1990s. Third, we draw on both Internet and cell phone data per
elements of democracy are not sufficient conditions for a stable 100 inhabitants per 181 countries to assess the impact of the
democracy, but they are necessary and indispensable to the information revolution on democratization. We use regression
persistence of democratic governance. analysis to determine whether Internet or cell phone use has
The italicized terms above represent the fundamentals be- had a correlative effect on measures of democratic tendency.
hind the sort of democracy we intend: active citizen par- We also model whether the collection of democratic measures
ticipation, good governance, accountability, power sharing, has had a correlative effect on Internet or cell phone use.
balance of power and rule of law. The few quantitative studies The paper is structured as follows: the first section reviews
that do exist on Internet and democracy tend to aggregate the current debate and literature on Internet and democracy.
these fundamentals of democracy into a single index. Doing The second section explicates the datasets used in this study
so means these lose important information on how these while the third section formalizes the statistical models em-
individual components of democracy may be affected by the ployed in the regression analysis. Section four reviews the
growing prevalence of global Internet access. Furthermore, results and provides an interpretation of the findings. The fifth
past quantitative and qualitative studies tend to focus primarily and final section concludes the study.
on the impact of the Internet on established democracies. They
also focus on the 1990s almost exclusively, a serious limitation II. L ITERATURE R EVIEW
that remains surprisingly understated in the literature. Equally The Internet and democracy literature comprises two distinct
problematic in the current literature is the interchangeable schools of thought each comprising a host of qualitative
use of the terms “Internet” and “information revolution.” The research and some quantitative inquiry. In this section we
terms are purposefully not differentiated on the basis that the review in some detail the qualitative and quantitative literatures
predominant feature of the information society is the spread that have contributed to both schools of thought over the past
of the Internet. While this is true of Western democracies, it ten to fifteen years. In so doing, we compare and contrast the
is certainly not true for the majority of developing, nondemo- main arguments along with the respective findings. As noted
cratic countries, where cell phones are the most widely spread in the introduction, one common shortcoming of the Internet
communication technology after radios [7]. Indeed, the irony is and democracy literature is the tendency to oversimplify our
that “those who might most benefit from the net’s democratic understanding of democracy. The purpose of this literature
and informational potential are least likely to have access to review is thus to redress this gap evident in previous studies.
it, the tools to gain access, or the educational background to The first school of thought is often referred to as the
take advantage of the tools” [2]. more populist school of thought. This strand of the literature
This paper seeks to redress each of these shortcomings. subscribes to the viewpoint that the Internet has democratic
First, since the boundaries of the term “democracy,” and how relevance and impact [8], [9], [10], [11]. According to these
it is measured, is subject to lively debate, we use multiple authors, the Internet will decentralize access to communication
perception-based measures of governance for our dependent and information while increasing citizen access [12]. Best
variables. Governance indicators provide a better set of proxies and Wade write that “the Internet’s collective characteristics
for the sort of democracy we intend as identified above. We (e.g., low cost, multidirectional capability, etc.), helps make
therefore draw on the following three World Bank indicators: this possible.” We first review the qualitative literature that
(1) Voice and Accountability (VA) measuring perceptions comprises this school of thought followed by quantitative
of the extent to which a country’s citizens are able to par- studies.
ticipate in selecting their government, as well as freedom Dahl previously observed that telecommunications tech-
of expression, freedom of association, and a free media; (2) nologies have a key role in making possible the advanced
Rule of Law (RL) measuring perceptions of the extent to democratic country, where policy is firmly anchored in the
which agents have confidence in and abide by the rules of judgment of the “demos” [13]. In his list of the procedural
society, and in particular the quality of contract enforcement, minimal conditions that must be present for modern political
property rights, the police, and the courts, as well as the democracy to exist, Dahl thus argues that citizens should
likelihood of crime and violence; and (3). Political Stability have the right to seek out alternative sources of information.
and Absence of Violence (PS) measuring perceptions of Rheingold has dubbed the Internet as “the great equalizer”
the likelihood that the government will be destabilized or because it can “equalize the balance of power between citizens
overthrown by unconstitutional or violent means, including and power barons” [14]. The idea here “is that the Internet
politically motivated violence and terrorism. These metrics are will serve as a mass audience, and will politicize them in the
drawn from the World Bank Governance Indicators Research process” [12]. Anderson et al. draw on sociological research
Database (see http://info.worldbank.org/governance/wgi/). to show that electronic networks lead to the “break-down of
Second, we draw on data from 2000 through 2006, a time status-based social structures” and “increased participation in
5
discussion, decision-making, and task processes by those who for example). At the international level, the authors make the
typically are political or economically disadvantaged” [15]. bold claim that the Internet has “contributed to the rise of a
Etzioni sees in the Internet the possibility of an advancement more multicentric world structure in which nation-states have
of the state of public affairs through “teledemocracy” [16]. seen their preeminence lessen and non-governmental actors
Other scholars claim that the Internet will “enable a Jeffer- take the stage” [1]. The salient point here is that groups and
sonian revolution” [17]. Grossman argues that a “third epoch individuals can far more efficiently form coalitions of conse-
of democracy is arriving by the hand of technology,” and quence with a range of powerful collectives. As Best and Wade
claims that a “new, hybrid electronic republic’ is now forming rightly note, there are obvious democratic elements to this,
to displace the creation of Montesquieu, Locke, Madison and including the need for “nation-states to provide democratic
their contemporaries” [18], [12]. The practice of “electronic rights to their citizens so as to build legitimacy on the global
democracy,” according to Browning, will differ substantially stage” [1].
from previous renderings of democracy [19]. Perhaps the most We now turn to the quantitative studies that comprise the
provocative claim associated with this more populist school of first school of thought. One of the earlier statistical studies
thought is the one made by Negroponte, who argues that the on this side of the literature was carried out by Kedzie,
nation-state will evaporate as a consequence of the information who provides an account of how information communication
revolution [20]. Snider suggests that citizens need only be technologies contributed to the “third wave” of democracy
potentially informed in order to hold government accountable [26]. Prior to the fall of the Soviet Union and the prolifer-
[21]. ation of new democracies in Eurasia, the mainstream theory
Barber opines that by favoring decentralization, the mul- of democratization held that democracy followed economic
tiplication of choice, and consumer sovereignty, new tech- growth and development [27]. To be sure, one of the few robust
nologies such as the internet or cell phone have already, findings in the literature is that democracy is more likely in
albeit inadvertently, benefited democratic political culture [2]. more developed countries [28], [29], [30]. Longdregan and
According to Barber, “democracy is a form of government Poole have also shown that the most significant predictor of
that depends on information and communication. It is obvious transitions to authoritarianism is poverty [31], [32]. “In short,
then that new technologies of information and communications after 20 years of observation and analysis during the third wave
can be nurturing to democracy. They can challenge passivity, of academic interest in democratization, we can be reasonably
they can enhance information equality, they can overcome certain that a positive relationship between development and
sectarianism and prejudice, and they can facilitate participation democracy exists, though we do not know why” [33].
in deliberative political processes.” Hill and Hughes argue that Kedzie, however, was more interested in testing another
those who subscribe to the populist school have reasons to be potential causal mechanism, the “dictator’s dilemma” hypoth-
optimistic: “If the mere fact that political discourse against esis, which suggests that the globalization of markets places
repressive governments is taking place is a good in itself, then pressure on authoritarian regimes to keep their countries’
the utopians have reason to celebrate [22]. Perhaps the Internet communication borders open. He reasoned that the ensuing
will bring about a wider democratic revolution in the world” massive flow of information would not only allow for “the
[22]. efficient passage of commercial information, but also for more
Bimber is more cautious, arguing that while the Internet is ’democratic’ information” [1]. As Bimber observes, the most
accelerating the process of issue group formation and action important predictions about the Internet’s impact on politics
(in America), the structure of political power has not been amount to “causal claims regarding the effect of information
revolutionized or qualitatively transformed into a new epoch of flow on political participation and the organization of interests”
democracy [12]. According to Bimber, other scholars believe [12]. Other scholars have made related arguments. Webster,
that the Internet has a “transformative potential” because it for example, writes that the Internet has helped to facilitate
facilitates a kind of “one-to-one interaction among citizens a new form of capitalism called “information capitalism” in
and between citizens and government.” Along these lines, which global labor markets require highly flexible workers
some scholars such as Corrado and Firestone write that the who continuously adapt and learn [34]. Regimes that impose
Internet has the potential to promote “unmediated” commu- restrictions on information capitalism forgo the financial re-
nication and thereby decrease citizens’ reliance on officials turns possible by tapping into the information economy [35],
and organizations [23]. In sum, what distinguishes the populist [36], [37].
enthusiasm for the Internet is the “idea that elites and political In his study, Kedzie employs regression analysis to compare
intermediaries will grow less important” [12]. how much of the variation in democracy is explained by both
In the more contemporary, qualitative literature, Steele and traditional predictors of democracy and the strength of Internet
Stein, argue that the Internet amplifies trends in international diffusion by drawing on data from 144 countries [26]. For
relations [24]. Rosenau and Johnson address the impact of the his set of control variables, Kedzie included more traditional
Internet at both the individual and international level [25]. At predictors of democracy including economic development,
the individual level, the authors argue that the Internet can be education, human development and health. He also included
used as a tool to organize collectively to effect social and polit- indicators of pre-Internet information communication tech-
ical change around the world (see http://www.DigiActive.org, nologies (ICTs). His results suggest that the Internet is indeed
6
a strong predictor of democracy, more so than traditional variation in the Freedom House data, which makes meaningful
determinants of democracy. In a follow up study, Richards statistical analysis more difficult.
assessed the relationship between the Internet and physical In contrast to the populist literature, the second school
integrity [38]. His findings support Kedzie’s. However, the of thought disputes the majority of claims that exist vis-à-vis
latter study faces an important limitation since Kedzie’s (rather the relationship between Internet and democracy. The counter-
simple) longitudinal analysis draws on data from 1993. At arguments are based on both qualitative and quantitative re-
this point during the early 1990s, the Internet was hardly search. In terms of qualitative research, several scholars argue
globalized. that the Internet is merely an extension of the ruling class and
Best and Wade recognize this important short coming in centralized control [39], [40], [41], [34], [42]. According to
Kedzie’s study and therefore explore the global effect of the Neuman, even if the increase of ICTs had led to an increase
Internet on democracy over a ten year period, 1992−2002 [1]. in the motivation to communicate - which he argues has not
They aggregate political and civil rights data from Freedom happened - then ICTs would have become centralized by
House to formulate a democracy index, which serves as their government turning them into social control mechanisms [43].
dependent variable. The number of Internet users per 1, 000 Scholars who subscribe to this school of thought maintain that
represents their independent variable while the following mea- mass media information technologies discourage collective
sures are used as control variables: GDP per capita, education behavior, ”unless the rise in couch potatoes can be considered
and literacy rates, life expectancy, urbanization, prevalence of a social movement” [44], [45], [46].
non-Internet ICTs. Their analysis shows that a statistically In contrast to Snider’s argument about the mere potential of
powerful correlation exists between Internet diffusion and citizens being informed acting as a source of accountability, if
level of democratization. “The more salient observation to power is measured by the potential for “monopoly and control
make, however, is that while economic prevalence and liter- over information and communication, it is evident that the new
acy maintain relatively constant correlations with democracy, technology can become a dangerous facilitator of tyranny”
the correlation for Internet prevalence gradually strengthens, [21], [2]. Indeed, while the Internet may enable citizens to
almost to the same level as economic prevalence” [1]. subvert political hierarchy, Barber notes that with increased
The authors suggest that this dynamic reflects the grow- participation comes the peril of political and economic surveil-
ing significant relationship between Internet prevalence and lance. The populist school of thought is often blind to “how
democracy: “perhaps this is an indication that the Internet easily liberating technologies become tools of repression” [2].
has come of age as a correlate of democracy” [1]. Indeed, Bimber rejects the supposition that the Internet will
they posit that this growth in correlation strength might “be have significant effects on public life, point out that “both
expected given the positive network externalities, the network theory and empirical evidence cast grave doubt on the
effect’ that is a salient property of the Internet” [1]. However, communication-action connection at the core of the populist
the coefficients from the regression analysis reveal that Internet theory” [12]. Lippmann argued that the capacity of ICTs to
usage is only able to predict a minimal amount of the variation recreate politics is constrained by human nature, ie. cognitive
in democracy: “to generate one point of democracy, an extra processing, and not by the technical properties of the media
500 Internet users per 1,000 citizens is needed, or an extra themselves [47]. The Internet, then, is no different than other
$5, 882 of GDP per capita is needed” [1]. The scale of democ- ICTs even if the new medium differs from previous technolo-
racy runs from 2 to 14. In terms of democracy’s traditional gies in a fundamental way, namely allowing social bonding
determinants, GDP was a weak predictor while literacy turned to occur asynchronously. In sum, the Internet is “hardly
out to have no significance whatsoever. The other control producing the first dramatic expansion in communication:
variables used were either insignificant or internally correlated. telephone, radio, and television also expanded communication
While Best and Wade’s important contribution to the litera- profoundly.” There seems no compelling reason to believe
ture on Internet and democracy is one of the few contemporary that the communication capacity of the Net will have such a
quantitative studies carried out thus far, their approach does dramatically different effect than have other advances in point-
face a number of important limitations [1]. For one, their to-point and broadcast communication” [12]. Moreover, Page
democracy index needs to be unpacked and “its constituent argues, new ICTs may very well overcome spatial distance
components, such as freedom of the press, openness of the but his far from sufficient for establishing vibrant forms of
electoral process,” for example, tested against traditional de- political communication and deliberation [48].
terminants of democracy to determine whether one compo- Furthermore, “if democracy is to be understood as delib-
nent provides more explanatory power than others. Another erative and participatory activity on the part of responsible
limitation of their data is the fact that it extends only to citizens, it will have to resist the innovative forms of dem-
2002. This should be updated today due to the rapid pace agoguery that accompany innovative technology and that are
of ICT diffusion over the past several years. In addition, too often overlooked by enthusiasts [2]. Aristotle wrote that
several scholars have criticized the Freedom House data with the basis of a democratic state is liberty. Barber adds that a
regards to conceptualization, measurement and aggregation “free society is free only to the degree that its citizens are
issues (Munck and Verkuilen, 2002; Rydland et al., 2008). informed and that communication among them is open and
Furthermore, as discussed subsequently, there is little to no informed [2]. However, recent research and empirical work
7
confirms that governments increasingly have the upper hand is global in scope and uses time series data from 1992 to
in controlling and regulating the impact of the information 2002. We seek to build on their work by using data from
revolution [49], [50], [51], [52], [53], [54], [55], [56], [57], 2000 to 2006 and using the World Bank Governance metrics
[58]. As Goldsmith observes, “if governments can raise the as our measures of democratic tendency. Best and Wade
cost of Net transactions, they can regulate the transactions” combined the Freedom House metrics of political rights and
[59]. Beilock and Dimitrova found that countries with lower civil liberties as their measure of democratic tendency. As
Freedom House scores for civil liberties had significantly described subsequently we feel the Freedom House data are
lower Internet usage (even when controlling for economic not well suited to a regression study such as this one.
development) [60].
De Mesquita and George Downs also argue that government A. Our Approach to Measuring Democratic Tendency Using
elites (e.g., in Singapore) have learned to “stifle the bottom-up the World Bank Governance Indicators
democratic potential of the Internet and still promote economic Dahl characterizes a government with power vested in a
growth, contrary to Kedzie’s dictator’s dilemma argument” plurality as follows [13]:
[61]. As Bimber notes, the “central theoretical problem for
1) Control over governmental decisions about policy is
the populist claim is the absence of a clear link between
constitutionally vested in elected officials.
increases in information and increases in popular political
2) Elected officials are chosen and peacefully removed
action” [12] To this end, McLuhan’s old dictum may be wrong:
in relatively frequent, fair and free elections in which
“the medium is not the whole message. Content matters, and
coercion is quite limited.
there is simply no overwhelming reason to believe that a
3) Practically all adults have the right to vote in these
new medium will necessarily enhance the political quality
elections.
of communicative content” [12]. In short, “technology need
4) Most adults also have the right to run for the public
not inevitably corrupt democracy, but its potential for benign
offices for which candidates run in these elections.
dominion cannot be ignored” [2].
5) Citizens have an effectively enforced right to freedom
In terms of quantitative analysis, Scheufele and Nisbet’s
of expression, particularly political expression, including
2002 statistical study suggests that the Internet does not
criticism of the officials, the conduct of the government,
increase democracy. “Through linear regression, they find
the prevailing political, economic, and social system, and
that mass media broadcasting (e.g., television, newspapers)
the dominant ideology.
plays a far more effective role than the Internet in promoting
6) They also have access to alternative sources of informa-
democratic citizenship” [1], [62]. Given that an established
tion that are not monopolized by the government or any
body of quantitative research on this topic has yet to material-
other single group.
ize, Scheufele and Nisbet do caution against generalizing the
7) Finally, they have an effectively enforced right to form
results of their study, which focused exclusively on the United
and join autonomous associations, including political
States. Other scholars interested in this line of research have
associations, such as political parties and interest groups,
questioned the supposed direction of causation drummed up
that attempt to influence the government by competing
by the populist school of thought. Using multiple measures
in elections and by other peaceful means.
of regime type, Milner’s statistical analysis demonstrates that,
ceteris paribus, democracies permit much greater online ac- The first four points largely describe procedural aspects of
cess, both in terms of Internet users per capita and Internet a democracy, whereas the last three points delineate the com-
hosts per capita [63]. To this end, the information revolution munication aspect necessary for a well-functioning democratic
may merely be reinforcing pre-existing dynamics. regime. In fact, Diamond goes further and notes that “[s]ome
Milner’s study uses data from 1991 − 2001 to measure the insist on a fairly robust (though still procedural) definition
influence of regime type of adoption of the internet. This study of democracy, like Dahl’s ‘polyarchy.’ By this conception,
attempts to address a slightly different question—whether democracy requires not only free, fair, and competitive elec-
there is a relationship between measures of democracy and tions, but also the freedoms that make them truly meaningful
ICT penetration—but we build on her work by extending the (such as freedom of organization and freedom of expression),
range of years to 2006 (although we begin measurement in alternative sources of information, and institutions to ensure
2000). We adopt a fixed effect model and control for the time that government policies depend on the votes and preferences
component directly in the model. of citizens” [64]. Expanding on Dahl, Diamond has developed
Beilock and Dimitrova develop a model to explain global his own list of characteristics of a democracy [65]:
country differences in Internet use using income, measures of 1) Control of the state and its key decisions and allocations
freedom, region dummies, and development indices [60]. Their lies, in fact as well as in constitutional theory, with
data is a cross section from 2001 and does not take democratic elected officials (and not democratically unaccountable
variables directly into account. actors or foreign powers); in particular, the military is
Best and Wade ask the question closest to that addressed subordinate to the authority of elected civilian officials.
in this paper [1]. They ask whether Internet penetration has 2) Executive power is constrained, constitutionally and in
an effect on the level of democracy in a country. Their study fact, by the autonomous power of other government in-
8
stitutions (such as an independent judiciary, parliament, respondents, as well as thousands of experts working for the
and other mechanisms of horizontal accountability). private sector, NGOs, and public sector agencies” [66].
3) Not only are electoral outcomes uncertain, with a sig- In points eight through ten, Diamond gives a description
nificant opposition vote and the presumption of party of the role of law in a democracy. The World Bank has a
alternation in government, but no group that adheres to governance metric that expresses some of this: “measuring
constitutional principles is denied the right to form a perceptions of the extent to which agents have confidence in
party and contest elections (even if electoral thresholds and abide by the rules of society, and in particular the quality
and other rules exclude small parties from winning of contract enforcement, property rights, the police, and the
representation in parliament). courts, as well as the likelihood of crime and violence” [66].
4) Cultural, ethnic, religious, and other minority groups This approximates the World Banks “Rule of Law” metric and
(as well as historically disadvantaged majorities) are not we propose it as an empirical measure of Diamond’s points
prohibited (legally or in practice) from expressing their eight through ten.
interests in the political process or from speaking their The World Bank also has a metric measuring political sta-
language or practicing their culture. bility: “perceptions of the likelihood that the government will
5) Beyond parties and elections, citizens have multiple, be destabilized or overthrown by unconstitutional or violent
ongoing channels for expression and representation of means, including politically-motivated violence and terrorism”
their interests and values, including diverse, independent [66]. While not enumerated in either Diamond’s or Dahl’s
associations and movements, which they have the free- lists, Amartya Sen postulated an empirical correlation between
dom to form and join. democratic regimes and political stability [67]. He notes both
6) There are alternative sources of information (including the “political incentives provided by democratic governance”
independent media) to which citizens have (politically) to prevent crises and specifically that the “positive role of
unfettered access. political and civil right applies to the prevention of economic
7) Individuals also have substantial freedom of belief, opin- and social disasters in general” [67]. Thus we investigate the
ion, discussion, speech, publication, assembly, demon- World Banks “Political Stability” metric as another measure
stration, and petition. of democratic tendency.
8) Citizens are politically equal under the law (even though In measuring the relationship between ICT penetration and
they are invariably unequal in their political resources). these democratic variables, it is clear that country wealth is a
9) Individual and group liberties are effectively protected confounding factor that sound be taken into account: wealthier
by an independent, nondiscriminatory judiciary, whose countries are both more likely to be democratic and to be
decisions are enforced and respected by other centers of the heavier users of both the Internet and the cell phone. We
power. gathered gross domestic product (GDP) data from 2000 to
10) The rule of law protects citizens from unjustified deten- 2006 from the International Monetary Fund. The GDP data is
tion, exile, terror, torture, and undue interference in their purchasing parity adjusted to be directly comparable between
personal lives not only by the state but also by organized countries.
nonstate or anti-state forces. Diamond notes that country size is highly correlated with
regime type: “countries with populations under one million are
Like Dahl, Diamond includes procedural aspects of a demo- much more likely to be both democracies and liberal democ-
cratic regime (points one through three) and he enshrines what racies. Two-thirds of these countries are liberal democracies,
he considers essential communication requirements in points while only 30 percent of countries with populations over one
four through seven. These latter points can be characterized as million are. Among the larger 150 countries, only half are
approximating a “freedom of expression” or “political voice” democracies, while 70 percent of the small countries are. The
aspect to democracy. We found the World Bank governance countries with populations over one million are about twice as
metric of “Political Voice and Accountability” to represent likely as small states to have an electoral authoritarian regime
Diamond’s notion well in that it measures “perceptions of the and half again as likely to have a closed authoritarian regime.”
extent to which a country’s citizen’s are able to participate in [64], [65]. Because of this, we included population in our
selecting their government, as well as freedom of expression, models to control for country size. The population data for
freedom of association, and a free media” [66]. Like all the 2000 to 2006 was also gathered from the IMF.1
World Bank Governance metrics, it was built from surveys
and other sources of data within each country. These sources B. Limitations of The Data
give an idea of “freedom of belief, opinion, discussion, speech, The ICT data is gathered from the International Telecom-
publication, assembly, demonstration, and petition” present in munications Union (ITU). The ITU requested the number of
the country, although they do not measure the proliferation Internet and cell phone users from each country. This raises
of channels of communication that Diamond enunciates. The a host of questions about the reliability of the data since it
World Bank governance indicators metrics are based on 35
data sources some of which yield “subjective or perceptions- 1 Both the IMF GDP and population data are available at http://www.imf.
based data” including that from “household and firm survey org/external/pubs/ft/weo/2008/01/weodata/
9
is self-reported by the country. Perhaps the country has an activity for 40 of the countries most actively engaged in
incentive to under or over-report? It is plausible that some repressing internet activity [68], [69]. Although not as extreme
countries have more reliable data collection mechanisms in as the case of the Internet, cell phone use can be restricted
place than others. With one report per country per year, the by the government as well, and will differ from country to
data are highly granular, but they do seem to follow steady country. Zuckerman gives several examples of government
trends, and steadily upwards in ICT adoption. crackdowns on mobile phone use: Belarus’s reported shutdown
The World Bank did not calculate governance metrics for of their SMS network in March 2006, reports of Ethiopian
2001. We carried out simple linear interpolation to provide a cell phone blocking during the 2005 election protests, and
data point for each country in this year. We made the decision Cambodian blocking of SMS for two days before their 2007
to do this since there are only six time points from which elections [7].
to interpolate, and only one, the year 2000, falls before our North Korea and Cuba were dropped from the study since
missing data. This creates data that are “too smooth” for year official data is not reported for these countries. It is generally
2001 since they have been created from our pre-existing data known that there is very little internet access in Cuba, and little
and this will make our results appear more precise than they to none in North Korea and these are both regimes with little
in fact are. We argue that this effects is minor since the World democracy. If we had been able to include these countries in
Bank data are themselves aggregated from a large number of our estimation of the models, this would likely have bolstered
sources, and thus less subject to noise than using a single our results.2 Details of the data cleaning and amalgamation
source would be. process are on the study website at http://www.stodden.net/
Other metrics of the level of democratic rights exist, such as ICTD.
the Freedom House “Freedom in the World” metric. Freedom Our population data was obtained from the International
House carries out an annual global survey of political rights Monetary Fund and contains a sparse number os missing
and civil liberties. We choose not to use this as a measurement values. The IMF has made estimates of their missing data
of democratic tendency for two reasons. We felt that the World to complete the dataset.3
Bank Governance Indicators could be well grounded in the
theory of democracy as measures of democracy. Secondly, the III. E MPIRICAL M ODELING
Freedom House measures have some quantitative limitations. Our data comprises a panel containing N different times
A certain amount of inertia is built into the measurements so series each consisting of T observations. The number of
that it is difficult for a country to move much from one year to countries, N , is 181, and T , the number of years in our study,
the next. For both political rights and civil liberties a country is 7. A fixed effects model of our democratic measures’ effect
is scaled from 0 to 7, giving only 8 possible outcomes for on ICT penetration follows:
a country. Combining these two factors leads to a database
that does not shift very much from year to year. In the years ICTit = β0 + β1 RLit + β2 V Ait + β3 P Sit +
of our study, 2000 to 2006, of the 193 countries surveyed by
Freedom House (after subtracting the 9 with missing values β4 P OPit + β5 GDPit + β6 M Fit +
for both political rights and civil liberties for the entire time γt T D + ξi CD + it
series), 105 had no change in their scores for political rights
and 86 had no changes in their civil rights scores. The average i = 1, . . . , N, t = 1, . . . , T.
variance of those that did exhibit some change from 2000 to
2006 was 0.48 for political rights and 0.32 for civil liberties.
In this paper we model the penetration of Internet or cell
This means the majorities of countries, if they changed at all,
phone usage per 100 inhabitants, ICT , as a function of the
changed by perhaps one point on the 8 point scale.
World Bank democratic measures (Rule of Law, Voice and
Having more years of data, including 2007, would improve
Accountability, and Political Stability)4 , country size, country
our modeling. At the time of this writing, the World Bank
wealth, and the male/female ratio in the country.
Governance metrics were not available for 2007.
ICT is one of “Internet Use” or “Cell Phone Use.”5 GDPit
The IMF estimated some of the population number for some
is the per capita Gross Domestic Product for country i at
of the countries. It is likely this has the effect of providing
year t, adjusted for purchasing power parity.6 P OPit is the
population data that is smoother than it would otherwise be.
Note also that both ICT measures, the Internet and Cell Phone 2 Countries with missing values also tended to support our hypothesis:
use, are measures per 100 inhabitants. We emphasize that this Afghanistan, Bhutan, Comoros, Kiribati, Serbia, St. Kitts and Nevis, Timor-
must be carefully noted in interpretation of the regression Leste, and Tonga. Cuba, Iraq, Montenegro, and North Korea simply did not
furnish enough data for inclusion in the study.
results, since we use population as an independent variable. 3 A precise explanation of their data interpolation procedure was not readily
We also note that even though the ITU collects Internet use available. See http://www.econstats.com/weo/V023.htm
4 Available at http://info.worldbank.org/governance/wgi/
statistics for each country, what it really means to use the
5 The data used in this study is available at http://www.itu.int/ITU-
Internet can vary by country due to filtering, censoring, and
D/ict/informationsharing/
other restrictions on access. The OpenNet Intiative at the 6 The data are available at
Berkman Center for Internet and Society monitors the filtering http://www.imf.org/external/pubs/ft/weo/2008/01/weodata/index.aspx
10
population of country i at year t and M Fit is the male/female the same test was performed on the World Bank variables, 54,
gender ratio.7 62, and 60 of 181 rejected nonstationarity for Rule of Law,
β0 is the intercept term for country i, and γt T D and ξi CD Voice and Accountability, and Political Stability, respectively
are a time effect and a country effect, respectively. The effect (about one third of the data). Although there some evidence
of time is controlled for by dummy variables: T Dt is 1 for year of stationarity, it appears that the majority of the time series
t and 0 otherwise. Similarly CDi is a dummy variable that is included in this study are nonstationary, and it’s potential
1 for country i and 0 otherwise, controlling for the differences biasing of coefficient estimates is a concern.10 The typical
between countries. Finally, it is a disturbance term with remedy is the difference the data to remove the nonstationarity.
distribution N (0, σi2 ), which we assume to be uncorrelated Running the regressions in the above equation allowed us to
across country cross sections. In this study there are N = 181 carry out tests on the residuals directly to evaluate the level of
countries and T = 7 years. autocorrelation. There are two regressions to be run in this
study, modeling Internet penetration and cell phone usage.
A. Autocorrelation in Panel Data
Typically the Durbin-Watson test with one lag is used to test
A panel regression model of this type is subject to possible for autocorrelation in the structure of the regression residuals
autocorrelation between subsequent observations because of and the regression with Internet use as a dependent variable
the time series components. In a regression model as described was found to have autocorrelation present, and the cell phone
above, it is possible to ’discover’ what Granger and Newbold penetration regression was not, with Durbin-Watson values of
[70] termed “spurious” relationships between the variables. 1.86 and 2.01 respectively.11
That is, tests of significance on estimated coefficients may The Durbin-Watson statistic is created by calculating d =
PT
indicate a significant result, when in fact none is present. (t −t−1 )2
t=2P
t 2 , where t is the tth residual from the regres-
Granger and Newbold suggest economic time series data may 1 t
sion. It follows that 0 < d < 4. A value of 2 indicates
be especially prone to autocorrelation since they tend to be
no autocorrelation. To test whether the Durbin-Watson test
non-stationary, in that it is not uncommon for the process
statistics could be considered equivalent to 2, the test in the
generating the data to depend on the time it was sampled.
R statistical software package was used [71].
For example, economic time series are commonly subject
The adjusted R2 values were 0.9246 and 0.9199, seeming to
to seasonal or cyclical effects. Time series data that are
fit the Internet regression squarely into Granger and Newbold’s
not stationary will violate the assumptions of least squares
area of caution: a low Durbin-Watson statistic and a high
regression since the variance of the error term will depend
R2 value, and also casts some suspicion on the cell phone
on time and thus introduce a bias into coefficient estimation.8
regression. Granger and Newbold offer that until “a really
Granger and Newbold describe a high R2 value and a low
satisfactory procedure is available, we recommend taking first
Durbin-Watson statistic as warning signs that the estimated
differences of all variables that appear to be highly autocorre-
model may be yielding spurious results.
lated.” (p. 8.) We carried out this operation on both the Internet
Thus it is important to determine whether the time series
and cell phone use regressions because of the high R2 values
data in this study are nonstationary. Our data are sampled
and the nonstationary data in both regressions, even though the
yearly suggesting they may avoid annual cyclical effects.
Durbin-Watson statistic did not suggest autocorrelation among
We analyze 181 countries from 2000 to 2006 (with 2001
the cell phone regression errors. The plots of the residuals
interpolated for the World Bank variables) and thus have 7
for both regressions indicate possible heteroskedasticity. This
values in each time series. As Granger and Newbold mention,
suggests running the following differenced model:
finely sampled time series tend to exacerbate the cyclical
effects and thus nonstationarity in the data.
As is typical, the Augmented Dickey-Fuller test was used to ∆ICTit = β00 + β10 ∆RLit + β20 ∆V Ait + β30 ∆P Sit +
assess nonstationarity in each of our time series. Since Internet
and Cell phone use are increasing rapidly for the vast majority β40 ∆P OPit + β50 ∆GDPit + β60 ∆M Fit +
of the countries in our study, we measured the autoregressive
structure of each time series as stationary around a trend line, γt0 T D + ξi0 CD + it
and modeled with one lag. The null hypothesis is that the
data are nonstationary. The test was applied to the ICT and i = 1, . . . , N, t = 2, . . . , T.
World Bank data at the country level. For the Internet and
mobile phone data the test rejected the null hypothesis of
nonstationarity for 40 and 35 of 181 countries respectively, Running this model for differenced Internet penetration and
meaning that for around 20% of the Internet and cell phone differenced mobile phone use did not improve the Durbin-
penetration time series, nonstationarity is not evident.9 . When
10 Note that a combination of nonstationary time series may in fact be
7 The gender ratio data was obtained from the Census Department’s Inter- stationary. This is termed cointegration.
national Database at http://www.census.gov/ipc/www/idb/tables.html 11 The p-values for the Durbin-Watson test are generated via a bootstrapping
8 For a mathematically precise explanation see [70], p. 2.
method and can fluctuate. In this case the p-values were 0.022 for the Internet
9 The tests were performed at the 10% level regression and 0.962 for the cell phone regression.
11
TABLE I TABLE II
S UMMARY S TATISTICS FOR 2000 DATA S UMMARY S TATISTICS FOR 2006 DATA
TABLE III
R EGRESSION C OEFFICIENT E STIMATES FOR G LOBAL I NTERNET
12
Watson statistics. They became 2.16 and 1.92 respectively. P ENETRATION
The adjusted R2 values were reduced to 0.3686 and 0.4059.
Variable Estimate Standard Error p-value
Both Durbin-Watson statistics reject the null hypothesis of
Rule of Law 0.9018 1.1193 0.4207
no autocorrelation at the 5% level. A common method of Voice & Accountability 0.9122 0.8514 0.2844
controlling for autoregression is using a 2-stage least squares Political Stability -0.3783 0.5345 0.4793
approach [74], [75]. In the first stage, the autoregressive GDP 0.0004 0.0002 0.0972*
POP -0.0488 0.8123 0.9521
structure in the residual is estimated using a model postulating Gender Ratio 0.8212 0.9189 0.3718
that the autocorrelation has a single lag structure, specifically:
Variable Estimate Standard Error p-value Rank Country 2000 Country 2006
Rule of Law -1.4120 2.0333 0.4876 1 United Arab Emirates 43.98 United Arab Emirates 118.51
Voice & Accountability -2.458 1.5946 0.1236 2 Qatar 19.90 Qatar 109.6
Political Stability 2.2823 1.0044 0.0233** 3 Kuwait 21.74 Kuwait 91.49
GDP -0.0002 0.0005 0.7061 4 Oman 6.63 Maldives 87.88
POP 0.1225 1.4327 0.9319 5 Bahrain 30.61 Bahrain 122.88
Gender Ratio 4.8839 1.7342 0.0050*** 6 Saudi Arabia 6.40 Oman 69.59
7 Maldives 2.83 Saudi Arabia 78.05
TABLE V 8 Bhutan 0 Bhutan 9.77
T OP 10 M ALE /F EMALE R ATIOS , 2000 AND 2006 9 Jordan 7.72 Jordan 74.4
10 Djibouti 0.04 Grenada 44.59
Rank Country 2000 Country 2006
1 United Arab Emirates 212.3 United Arab Emirates 218.5
2 Qatar 198.1 Qatar 202.5 confounding factors.
3 Kuwait 150.3 Kuwait 152.3
4 Oman 131.3 Maldives 127.3 The World Bank measure for political stability is also
5 Bahrain 129.9 Bahrain 126.7 significantly positively correlated with increased cell phone
6 Saudi Arabia 125 Oman 124.7 use. This finding suggests that political instability is related
7 Maldives 117.3 Saudi Arabia 120.5
8 Bhutan 112.3 Bhutan 111.0 to the mass diffusion of cell phone usage. In other words,
9 Jordan 109.9 Jordan 110.2 an increase in cell phone availability could increase the
10 Djibouti 107.1 Grenada 108.1 perceived likelihood that the government will be destabilized
or overthrown by unconstitutional or violent means. In their
statistical analysis, Mansfield and Snyder find that the process
as the gender ratio increases by about 5, cell phone usage per of democratization itself is indeed a destabilizing one [76].
100 inhabitants will increase by one phone. Throughout the “Certainly, the virtues of working democratic structures do not
years studied, roughly 7 of the 10 countries with the highest translate into a carefree path to the stabilization of democracy”
male/female ratio each year were located in the middle east, [77]. This finding also supports the arguments presented by
and the ratios at that end of the distribution dwarfed the other Rosenau and Johnson, as well as Shirky, who opine that the
countries’. As displayed in Table VI, it is plausible some of Internet can be used as a tool by civil society to organize
these values are extreme enough to have a large impact on the collectively to effect political change [25], [78].
regression fit, although why this did not occur in the internet
regression is not clear. It is also possible the very high gender B. Modeling ICT Penetration as a Function of Democratic
ratio values represent a ’middle east effect’ since a number Tendency Among the Most and Least Affluent Countries
of those countries are highly represented in the top 10 gender Examining ICT penetration for different strata of wealth
ratio values. This implies that the coefficient on the gender may help isolate effects that are characteristic of those groups.
ratio variable could represent a high growth in cell phone use Since wealth is a driver of investment one would expect
in the middle east. Table VII gives the 2000 and 2006 cell GDP to play a role in the country’s readiness and ability to
phone data for these countries. The coefficient indicates that adopt new communication technologies [79], [80], [81]. As
as the rate of change in the proportion of men increases, so established in the literature we also found increases in GDP
does the rate of change in cell phone use. to be associated with increases in Internet use. We choose
Interestingly, the greater political stability and the lower the to examine ICT penetration in both the top and bottom 20%
perceived threat of violence, the greater cell phone penetration. of countries by 2006 GDP more closely. Our focus on these
This may represent infrastructural stability if associated with groups, in particular the bottom quintile, is driven by Mansfield
political stability and thus a measure of investor’s confidence. and Snyder’s work theorizing the instability of emergent and
It is not clear why this factor would not therefore also be transitional regimes and the existence of the global digital
associated with an increase in Internet use. Perhaps cell phones divide [82].
are easier to proliferate than access to the Internet and so a Table VII lists the countries that fall into each of these
smaller increase in political stability encourages cell phone groups.
increase before Internet increase. As in the previous section we fit a model with an ICT
As shown in Tables I and II, the average cell phone penetration measure as the explanatory variable, and measures
penetration in 2000 was about 16 phones per 100 inhabitants of democratic tendency and controls as independent variables
and in 2006 it was about 53, nearly a three-fold increase. This for a panel regression over years 2000 to 2006. For the
is a high rate of increase but, notably, the countries listed top wealthiest quintile of countries our model did not yield
in Table VI (those with the highest male to female gender statistically significant results for Internet penetration. Table
ratios) had much higher than average growth in cell phone VIII gives the coefficient estimates. Since these countries are
penetration. This is quantified in the significant coefficient in exceptionally wealthy and relatively stable politically it may
the regression in Table IV, while allowing for the included not be a surprise that GDP is not a driver of Internet use, and
13
TABLE VII TABLE IX
T OP AND B OTTOM 20% OF C OUNTRIES BY GDP IN 2006 I NTERNET P ENETRATION , B OTTOM 20% OF GDP (2006)
Top 20% Countries Bottom 20% Countries Variable Coefficient Standard Error p-value
Qatar Zimbabwe Rule of Law 0.3702 0.4081 0.3660
Luxembourg Congo (Dem. Rep.) Voice & Accountability -0.8115 0.3656 0.0281**
Brunei Darussalam Liberia Political Stability -0.2583 0.1822 0.1585
Norway Burundi GDP -0.0002 0.0014 0.9053
Singapore Guinea-Bissau POP -0.0309 0.2784 0.9117
United States Afghanistan Gender Ratio 0.4026 0.3235 0.2155
Ireland Sierra Leone
Switzerland Niger TABLE X
Hong Kong, China Central African Rep. C ELL P HONE P ENETRATION , T OP 20% OF GDP (2006)
Kuwait Ethiopia
Iceland Malawi
Canada Eritrea Variable Coefficient Standard Error p-value
Netherlands Mozambique Rule of Law 2.354 6.161 0.7029
Austria Togo Voice & Accountability 4.012 4.736 0.3984
Denmark Rwanda Political Stability 5.071 1.887 0.0612*
United Arab Emirates Uganda GDP -0.0008 0.0005 0.1071
Sweden Myanmar POP -2.896 6.483 0.6558
Australia Mali Gender Ratio 2.8300 4.3690 0.5182
Belgium Madagascar
United Kingdom Guinea
Finland Comoros
Germany Tanzania political turmoil and it may be the case that countries with
Japan Nepal higher Voice and Accountability rankings have been reluctant
France Burkina Faso to permit the growth of the Internet in their milieu.
Bahrain Lesotho
Italy Bangladesh Table X gives the coefficient estimates from the panel
Spain Gambia regression for cell phone penetration for the top 20% of
Taiwan, China Haiti wealthiest countries. There is a statistically significant effect in
Greece Zambia
Cyprus Ghana the World Bank metric of Political Stability: greater Political
New Zealand Sao Tomé & Principe Stability is associated with an increase in cell phone use per
Slovenia Benin inhabitant. This result seems intuitive as political stability is
Israel Kenya
Bahamas Senegal historically associated with greater investment in communica-
Korea (Rep.) Cambodia tions infrastructure and is consistent with our earlier regression
Saudi Arabia Chad on cell phone use. This suggests the wealthiest countries may
Czech Republic Côte d’Ivoire
be driving the correlation between political stability and cell
phone penetration. None of the other variables were found
TABLE VIII
I NTERNET P ENETRATION , T OP 20% OF GDP (2006) to have a statistically significant relationship with cell phone
penetration in the wealthiest countries.
Variable Coefficient Standard Error p-value Among the poorest countries, growth in Voice and Ac-
Rule of Law -0.3709 5.4055 0.945 countability had a statistically significant negative effect on
Voice & Accountability 4.771 4.0981 0.246
Political Stability -1.0344 2.2760 0.650 growth in cell phone penetration. Although consistent with
GDP 0.0003 0.0004 0.450 the Internet penetration regression results for this group of
POP 1.6791 5.5724 0.764 countries, the paradox remains as to the increase in per capita
Gender Ratio -1.7979 3.7568 0.633
cell phone use as Voice and Accountability decreases. It is
plausible citizens desire newer forms of ICT when Voice
and Accountability is restricted. Another explanation may be
nor are the measures of democratic tendency, even though our that although not all countries in the bottom quintile are
subsets contains countries with varying levels of autocratic autocratic, a significant proportion are and cell phone use may
control. facilitate the mobilization, organization and coordination of
Among the least wealthy quintile we fit the same panel resistance against autocratic rule. Interestingly, the voice and
regression model as above. Table IX gives the regression coef- accountability metric is not a significant predictor globally, yet
ficient estimates. Our predictors did not yield highly significant is significant for types of ICT among the poorest countries.
coefficient estimates with the exception of the World Bank
voice and Accountability metric. Voice and Accountability V. C ONCLUSIONS AND F UTURE R ESEARCH
is negatively correlated with Internet penetration: implying This paper is the first to our knowledge that uses recent
that when countries notch up in the Voice and Accountability Internet and cell phone use data in an empirical study of
ranking, the use of the Internet increases. This seemingly their relationship to democratic tendency. Previous studies uses
paradoxical finding may be explained when note that our measures of Internet use that ended in 1993 [26], [38]. Best
analysis is restricted the the lowest quintile of country in and Wade’s data reached only to 2002. This paper is also the
wealth. These countries experience disproportionately greater first to the best of our knowledge that measures ICT diffusion
14
TABLE XI
C ELL P HONE P ENETRATION , B OTTOM 20% OF GDP (2006) used globally than the Internet, so this may be a function of
cell technology’s earlier foothold than the Internet’s. As Inter-
Variable Coefficient Standard Error p-value net diffusion catches up to that of cell phone, the democratic
Rule of Law 0.3702 0.4081 0.3660
Voice & Accountability -0.8115 0.3656 0.0281**
metrics may be found to be predictors of this diffusion. This
Political Stability -0.2583 0.1822 0.1585 is not necessarily inconsistent with the populist thread in the
GDP -0.0002 0.0014 0.9053 literature as, according to Bimber, other scholars believe that
POP -0.0309 0.2784 0.9117
Gender Ratio 0.4026 0.3235 0.2155
the Internet may have a “transformative potential” because
of the “one-to-one interaction among citizens and between
citizens and government” it creates [12].
A paradox is created for the populist school in the finding
as a function of democratic indicators. Previous research has that “Voice and Accountability” is negatively correlated with
focused on whether ICT use predicts democratic measures. cell phone diffusion. This can be interpreted in favor of the
These results support Bimbers assertions that the structure argument that ICT diffusion can be centralized by government
of political power has not been revolutionized or transformed turning them into social control mechanisms [43]. As the rate
into a new epoch of democracy [12]. Evidence can be found of Voice and Accountability increases, the rate of diffusion
for both the populist thread in the literature and the notion that of ICTs decreases among the poorest and least developed
ICTs may act as an extension of the ruling class. countries. Rates of cell phone use increase globally with
We found a statistically significant positive relationship decreases in political stability, lending further support to the
between the rate of diffusion of the cell phone and the World thesis of ICTs as an extension of ruling class control.
Bank’s “Political Stability” measure capturing perceptions This paradox may be resolved is we consider Page’s view
regarding the likelihood that a government will be destabilized in 1995, that ICT diffusion may be still too nascent and
or overthrown by unconstitutional or violent means. Political insufficient to generate a well functioning system of political
Stability continues to have a significant positive relationship communication and deliberation [48]. It appears we are still
with the rate of cell phone use among the most affluent 20% too early to expect a close relationship between a vibrant
of countries, but that relationship does not hold among the public sphere and ICTs globally. As Bimber states, the “central
wealthiest countries. The “Voice and Accountability” indicator theoretical problem for the populist claim is the absence of a
which measures perceptions of the extent to which citizens clear link between increases in information and increases in
are able to participate in selecting their government, as well popular political action” [12]. This paper provides evidence
as freedom of expression, association and of the media, of the existence of this problem and the need to develop our
was a significant negative predictor of the rate of cell and understanding of this dynamic further.
Internet diffusion among the poorest quintile of countries. The It would be interesting to tie this research more closely to
diffusion of ICT access did not have any significant influence development, in line with Sen’s reasoning that “Developing
on “Rule of Law” while the increasing availability of cell and strengthening a democratic system is an essential compo-
phones were shown to have no influence on any of the three nent of the process of development,” could extend the empiri-
World Bank indicators. cal analysis in a fruitful direction. This might mean specifically
Recall that the Rule of Law metric measures perceptions of testing whether the order in which political and civil rights are
the extent to which individuals have confidence in and abide extended as a country emerges from an autocracy affects the
by the rules of society—in particular the quality of contract rate of development. This could provide a setting in which to
enforcement, property rights, the police, and the courts as well test the “Lee Thesis,” that political rights should be withheld
as the likelihood of crime and violence. This measure was orig- until economic development is achieved.
inally included in the analysis based on Diamond’s research on It may be valuable to explore empirical issues further.
the characteristics of democracy. However, the findings here Modeling the autocorrelation structure with more than one lag
suggest that the increase in “Rule of Law” perceptions does may help reduce autocorrelation. It would also be interesting
not influence ICT penetration, either positively or negatively. to test for cointegration among these variables. Certainly
In other words, perceptions regarding the “Rule of Law” may documented feedback loops exist between our independent
be framed and influenced by factors other than widespread variables, such as GDP and measures of democracy, and taking
ICT use. this explicitly into account may improve the modeling [83],
The populist school of thought believes ICT diffusion will [26]. It is also plausible that feedback loops exist between
decentralize access to communication and information while ICTs and democratic measures and future modeling could
increasing citizen access [12], while Hill and Hughes claim accommodate this. Further research into the modeling aspects
that perhaps the Internet will bring about “a wider demo- could estimate models including variables that control for
cratic revolution in the world” [22]. In our modeling Internet the different manifestations of cell phone and Internet use in
diffusion was not predicted by our measures of democracy, different countries. A more comprehensive model might ex-
implying this revolution has not yet arrived. The rate of cell plore possible non-linearities: whether countries with low ICT
phone diffusion was predicted by higher rates of the “Political adoption rates have different patterns of democratic tendency
Stability” metric. Recall that cell phone are much more widely than those with high adoption rates.
15
[64] L. Diamond, “elections without democracy: Thinking about hybrid [76] E. Mansfield and J. Snyder, “Conceptualizing and measuring democ-
regimes,” Journal of Democracy, vol. 13, no. 2, April 2002. racy:evaluating alternative indices,” International Organization, vol. 56,
[65] ——, Developing Democracy: Toward Consolidation. John Hopkins no. 2, pp. 297–337, 2001.
University Press, 1999. [77] H. Barnes, “The contribution of democracy to rebuilding postconflict
[66] D. Kaufmann, A. Kraay, and M. Mastruzzi, “Governance matters vii: societies,” The American Journal of International Law, vol. 95, no. 1,
Aggregate and individual governance indicators 1996-2007,” World pp. 86–101, 2001.
Bank, Tech. Rep., 2007. [78] C. Shirky, Here Comes Everybody: The Power of Organizing without
[67] A. Sen, Development as Freedom. Oxford University Press, 2001. Organizations. Penguin Press, 2008.
[68] “Opennet initiative.” [Online]. Available: http://opennet.net/ [79] C. Argevou, “Ols bias in a nonstationary autoregression,” Information
[69] R. Faris and N. Villeneuve, Access Denied: The Practice and Policy of Technology for Development, vol. 8, no. 1, pp. 15–28, 1998.
Global Internet Filtering. MIT Press, 2008, ch. 1. [80] B. Oyelaran-Oyeyinka and K. Lai, “Internet diffusion in sub-saharn
[70] C. Granger and P. Newbold, “Spurious regressions in econometrics,” africa: A cross-country analysis,” Telecommunication Policy, vol. 8,
Journal of Econometrics, vol. 2, pp. 111–120, 1974. no. 1, pp. 507–527, 1998.
[71] “R.” [Online]. Available: http://www.r-project.org/ [81] J. Gibbs, K. Kraemer, and J. Dedrick, “Environment and policy factors
[72] C. Granger, “Some properties of time series data and their use in shaping global e-commerce diffusion: A cross country comparison,” The
econometric model specification,” Journal of Econometrics, vol. 16, pp. Information Society, vol. 19, no. 1, pp. 5–18, 1993.
121–130, 1981. [82] N. Kshetri and N. Dholakia, ENCYCLOPEDIA OF INFORMATION
[73] ——, “Co-integration and error correction: Representation, estimation SCIENCE AND TECHNOLOGY, 2nd edition. Idea Group Publishing,
and testing,” Econometrica, vol. 2, pp. 251–176, 1987. 2008.
[74] T. Amemiya, “The nonlinear two-stage least squares estimator,” Journal [83] P. Norris, Digital Divide: Civic Engagement, Information Poverty, and
of Econometrics, vol. 2, pp. 105–110, 1974. the Internet Worldwide. Cambridge University Press, 2001.
[75] R. Cumby, J. Huizinga, and M. Obstfeld, “Two-step two-stage least
squares estimation in models with rational expectations,” Journal of
Econometrics, vol. 21, pp. 333–355, 1983.
17
i
The analysis includes two studies of fishermen but excludes
farmers
ii
The term ‘value system’ is sometimes used interchangeably
with ‘industry value chain’. This review uses the Porter
nomenclature to distinguish between the intra- and extra-
enterprise systems
27
Abstract—Claim Mobile is a platform designed to support a fraud and transformed supply-chain management for the E-
project that subsidizes healthcare by reimbursing health service choupal project [6]. While health information is critical to the
providers in Uganda for treatment of patients with sexually improvement of healthcare in developing regions, financing
transmitted infections. As with many development projects, the
Uganda Output-Based Aid (OBA) project involves a number of healthcare also remains a significant unsolved problem. Can
stakeholders: the service providers, the project implementers, we take lessons from e-Choupal and apply them in the
the financiers, and the Ugandan government. Design of an healthcare sector? The design of usable, reliable, and fraud-
appropriate solution requires meeting the various and conflicting resistant tools for management of these aid flows is an area
requirements of all of these stakeholders. In this paper we detail with potential for very significant impact.
the rapid design and testing of a pilot implementation of a
mobile and web-based system for processing claims forms, based
on two prior field visits to Uganda. Based on a comparative However, in the case of healthcare, the financial models are
device study, semi-structured interviews, health clinic surveys,
and a brief deployment, we affirm the selection of the mobile very different from commercial markets – financing of health-
phone as a platform from the health clinic perspective, and care typically comes through transnational aid agencies like
further suggest that effective design for development requires the World Bank and International Monetary Fund (IMF), and is
more than addressing requirements of the the “users” of the implemented by non-governmental organizations (NGOs) and
mobile phones but also all the other stakeholders involved, who the local government. Since the NGOs are typically experts in
may have conflicting requirements.
Index Terms—mobile phone, ICTD, health, participatory de-
health, not technology, data processing is often outsourced to
sign, Africa, HCI third-party information technology (IT) vendors. Relationships
between the vendors, the NGOs, the local governments, and
the transnational aid agencies are not always smooth - and
I. I NTRODUCTION
limitations in communications infrastructure means that the
Mobile phones are frequently touted as being the appro- information flows between them are scattered at best.
priate and sustainable platform for rural healthcare in Africa.
They are relatively cheap, durable, consume less power than
In this paper we suggest that the “closed loop model”
laptops and desktops, and incorporate a battery that makes
generally used by researchers in deployments of mobile health
them more amenable to use in places with intermittent or no
applications does not map onto the financial and political
power. Commonly proposed uses are for data collection [1],
realities of the mainstream of healthcare provision in Africa,
[2] and decision support for rural health workers [3], [4]. Some
and limits the ability of pilot programs to increase their scale
projects also use mobile devices as a platform for information
and impact. We describe an innovative, IT-based, NGO-run
dissemation as well as data gathering [5]. However, these are
healthcare access program in Uganda, and our experiences
all generally “closed loop” systems in which researchers are
designing and deploying Claim Mobile, a mobile-phone based
able to control all aspects of the system design and operation,
system intended to address inefficiencies and help the program
focusing their research primarily on the rural health workers
scale to additional districts. We argue that in addition to
that will be using the mobile phones.
addressing the needs of the primary users in the system, the
Other applications have even more potential for large-scale health workers, our design must consider the requirements,
impact. In the agricultural sector, we have observed how motivations and concerns of the other stakeholders: the IT
the introduction of transparent market prices and subsequent vendors, the NGOs, the government, and the aid agencies.
hiring of “middlemen” to purchase from farmers has reduced Our designs must consider the larger order ramifications of
how we may positively and negatively impact both the “users”
Manuscript received October 1, 2009. This work was supported by the
Blum Center for Developing Economies and the U.S. National Science who will be generating the data, and the entities that will be
Foundation Summer Undergraduate Program in Engineering Research at engaged in managing and using the information in the resulting
Berkeley (SUPERB) under Grant No. 0453604. database. Just as the e-Choupal project assimilated the mid-
Melissa R. Ho is with the School of Information at the University
of California, Berkeley, 94720, USA. (phone: +256 777 723 786; email: dlemen by hiring them as kiosk operators, we propose that we
mho@ischool.berkeley.edu). can design applications structured to accommodate conflicting
Emmanuel K. Owusu is with the Computer Engineering Department at stakeholder requirements, while also alleviating information
Iowa State University, 50011, USA. (email: kwame@iastate.edu).
Paul M. Aoki is with Intel Research, Berkeley, CA, 94704, USA (email: inequalities resulting from limitations in the system prior to
aoki@acm.org) the introduction of the information technology.
36
Mbarara, Uganda
• HIV prevalence: 10% of adult population (15-49
years)
• Syphilis prevalence: about 5-7% of adult population
• 1 in 4 households had at least one phone.
• 39% reported STI symptoms
• only 1/3 sought care
• 54% of respondents who sought any STI treatment
reported using private clinics.
Fig. 1. A HealthyLife voucher. The ‘M’ in the top left is a note written TABLE I
by the health service provider to indicate that the first client voucher on the S OME BACKGROUND STI STATISTICS ABOUT M BARARA , U GANDA [7].
left was given to a male client, and that therefore the partner vouchers on the
right should be given to a female client.
can take two weeks or more to move from the providers office
II. BACKGROUND to the management agency. The current data management
system requires all claims to be submitted on paper forms
Providing effective health care in poor countries is an to the management agency. At least another two to four
essential component to economic development and poverty weeks are spent reviewing each claim, cleaning data from
reduction. Unfortunately donors supporting this endeavor often improperly-completed forms, and verifying that the service
find that resources given are not matched by desired gains. took place among suspect claims. Two months or more can
The output-based aid (OBA) model of financing seeks to ad- go by before the provider is reimbursed for service provision.
dress this by paying healthcare providers directly for services In Uganda, private providers traditionally operate on a fee-
rendered instead of paying for the service provision up front. for-service model, receive prompt payment, and do not have
However, OBA program management is information intensive, a large operating margin. In many cases, payment is provided
necessitating much paperwork to track and reimburse payment prior to service. Delays in payment result in delays in pro-
claims. Smartphones (mobile phones with advanced features curement of replacement prescriptions and medical supplies,
such as the ability to run third-party software) have the often leading to a temporary hiatus in service. Encouraging
potential to alleviate this burden. In collaboration with a local provider involvement in the OBA program requires a great
NGO and their partnering IT vendor, we have proposed to deal of confidence on the part of the providers to participate. If
deploy a number of smartphones for use in an OBA project a system to shorten claims processing could be devised, more
based in Western Uganda, with dual goals of reducing claim providers could join the scheme and more patients could be
processing time and improving communication between the provided the life-saving STI treatment voucher subsidy.
health care providers and the OBA management agency. The remainder of this paper details the system we are cur-
The project is managed by the local branch of a multi- rently piloting, in which claims are submitted via Internet from
national NGO and a for-profit health insurance company, in a mobile phone directly to all the parties in the management
collaboration with the Ugandan Ministry of Health (MoH) agency. In addition to describing our user studies and how this
and Ministry of Finance (MoF). The project is primarily has informed the design of the system, we discuss the problem
funded by an aid agency based in Europe, with additional of negotiating conflicting stakeholder requirements. We find
funding for the expansion coming from a separate transna- that in projects with multiple stakeholders, the introduction of
tional funding agency. Together, they run a voucher program a system may disrupt balances of power, particularly around
called HealthyLife, which treats sexually transmitted infec- the flow of information and money. As a result, the design of
tions (STIs), reimbursing providers for the diagnosis and full this system, in order to secure positive support from all parties
course of treatment only after the patient is seen. This program involved, must carefully balance stakeholder incentives.
was implemented in response to the high burden of sexually
transmitted infections in Uganda, and began in July 2006 III. M ETHODS
in four districts of southwestern Uganda: Mbarara, Ibanda, The research described here involved an iterative process of
Kirihura and Isingiro (See Table I). field research and prototyping. The fieldwork and deployments
Patients buy treatment vouchers in pairs, one for the client have been done over the course of three visits to Uganda:
and a second one for the client’s sexual partner (See Figure 1). an initial two-week visit in Summer 2007 to establish a
Each voucher is good for one consultation (generally including relationship with the project, in which we also conducted a
a lab test to diagnose the STI) and three follow-up visits. survey of the clinics in the program; a followup visit for
During the consultation, the provider completes a claim form three weeks in November 2007; and a five-week pre-pilot
recording the client’s demographics, the examination and deployment in August-September 2008. During all three visits
laboratory results, a diagnosis and details of the course of we conducted semi-structured interviews with the various
treatment prescribed (See Figure 5). Completed claims forms stakeholders, and directly observed claims form entry and
are sent to the voucher management office in the city of processing. When given permission, we did audio and video
Mbarara, the main urban center of Western Uganda. Forms recording of interviews and user study activities. In all, we
37
A. Clinic Surveys
The clinic survey was conducted in conjunction with a
larger survey of available infrastructure at contracted clinics. Fig. 2. This diagram illustrates the flow of money and information between
selected stakeholders in the OBA project, both currently, and how it will be
We asked 14 questions, assessing familiarity with computers once the Claim Mobile system is fully deployed.
and mobile phones, but primarily gathering feedback from the
health clinics on the claims process (e.g., how long it takes
them to fill out the paper forms, and what their priorities be conducted in May 2009, with mobile phone-based claims
might be for improvement of the process). We also collected processing expanded to 8-10 additional clinics.
various documents from the management agencies regarding
the performance of each clinic, including all available financial IV. S TAKEHOLDERS
reports on processed claims, and in November, we returned to In this section we describe the funding, management and
seven of the clinics to do in-depth surveys and to follow-up service provider organizations to which we alluded in Sec-
on the survey findings. tion II. Fundamentally, all stakeholders want to improve STI
treatment and reduce the prevalence of STIs. Each stakeholder
B. Rapid Development and Pre-pilot Deployment also has a financial interest in the success of the overall project
Initial prototyping occurred in early 2008, and we returned - staying afloat for the health service providers, and staying
to Uganda in Summer 2008 to do a three-stage pilot deploy- within the aid agencies’ target budgets for the management
ment, first testing the functionality of our software, second agency partners. The discussion highlights the ways in which
reviewing the proposed claims process with the management the various stakeholders have competing as well as common
agencies, and finally taking the phones to the health clinics to interests. While we will detail several stakeholders in this
test the mobile phone interface in the field. During this time section, there are six key stakeholders: the aid agencies who
we also conducted another round of semi-structured interviews fund the OBA project, the financial management agency
to gather information on changes in the claims submission (FMA) which receives the funds from the government and
process (for example, claims processing had moved from disburses them, their program management office (PMO) in
Mbarara to the national capital, Kampala). We did iterative Mbarara which runs the program and interacts directly with
development based on feedback from the various stakeholders, the health service providers, the technical management agency
trying out features as they were suggested, and developing (TMA) that manages the claims processing, the health service
new tools as seemed merited by findings in our interviews. providers (HSPs), and us, the mobile platform developers
To gain a more in-depth understanding of health clinic life, (MPDs). Figure 2 illustrates some of the relationships between
we stayed overnight for three days in the rural health clinic, these entities which we will describe in detail in the remainder
thereby supplementing the the semi-structured interviews with of this section, based on qualitative fieldwork and document
direct observation of actual practice. analysis.
The primary purpose of this last field visit was to con-
duct a pre-pilot demonstration, using the mobile phones to A. Aid Agencies
submit actual claim data to the management agency, have it As the funder of the HealthyLife STI treatment program,
reviewed, and have the management agency provide feedback the involvement of the European aid agency is more than
to the health clinics via the mobile phones. We simulated apparent. Their role in the management of the program is
the proposed process, physically following the claims forms more supervisory – a consultant goes to Uganda at irregular
from the time the patient comes into the health clinic, through intervals to help with planning of the program, and they do
the preparation of the claims summary forms, physically some monitoring. They also have commissioned another non-
transporting the forms to the management agency where we profit, affiliated with a North American university, to conduct
observed the claims approval, and data entry into the existing an evaluation of the program. Ultimately, however, they control
database. We simultaneously had the service providers submit the flow of money to the financial management agency, which
the claims form via Claim Mobile, enabling the management then pays the IT vendor to handle the technical aspects of the
agency to provide feedback to the service providers through operation.
the system. The pre-pilot is still operational, with mobile In the past year, the European aid agency has worked with
phones remaining in the two clinics, and the full pilot will an additional transnational aid agency to fund the expansion
38
of the project into additional districts. While they may not for the PMO, and one for the TMA. However, the copy that
have a direct impact on the information processes in the remains in the PMO does not have the voucher number, a
project itself, the funders’ internal actions have direct impact critical piece of information, and with stacks of hundreds of
on the project as a whole. In one example, a delay in claims per month, the information is not in a format actually
payment to the European aid agency resulted in a delay in accessible to the program office until the TMA sends back
payment to the two management agencies. As a result the IT claims summaries. However, even this is stripped-down and
vendor ceased processing of claim forms until payments were only includes the value of the claims, without any patient or
received. However, not only did the voucher program grind diagnostic information from the claims.
to a halt, but reimbursement to the participating providers for This poses a problem for the PMO staff’s interaction
patients already seen was delayed as well; the management with the health clinics. They lack sufficient information to
agencies ended up with a backlog of claims forms to process, effectively counsel and train the clinics, and often feel like
exacerbating the length of time it takes to process claims and they do not know what is going on with the program because
further delaying payment for services. they do not have access to the claim data for the long claims
processing cycles.
B. Financial Management Agency (FMA)
In the version of Claim Mobile developed in early 2008,
The local NGO partner that acts as the financial manage- we intended to make the claims process more efficient by
ment agency (FMA) is the Uganda office of a multinational enabling the mobile phones to submit claims directly to the
non-profit sexual and reproductive health organization with a database (originally co-located in the PMO, now located in
goal of reducing unintended pregnancies and unwanted births Kampala). In the Summer 2008 interviews it became clear that
through family planning and other methods. Their role in having the claims data bypass the PMO staff would deny them
this project is to receive the funds from the aid agencies even more of the information they need in their interactions
via the Ugandan government, using these funds to pay the with the clinics. Based on this realization, we discussed the
health clinics and to pay for other program costs, including possibility of an intermediary application, a website in which
the database software development and management. The main the project coordinator in the PMO would be able to view
office in Kampala runs this program (as well as others) and claims as they are submitted, as well as any status updates.
manages several clinics throughout Uganda, one of which was Furthermore, Claim Mobile could facilitate another of the
a participating clinic in the HealthyLife program until Summer project coordinator’s key roles in the OBA ecosystem: as the
2008. In addition, they have a program management office primary interface between the health service providers, the
(PMO) in Mbarara which is directly in charge of coordinating project coordinator would also be able to send messages to the
the the OBA project. In the claims process, the FMA disburses service providers through Claim Mobile, either as individual
payments to each of the service providers, based on claim messages, or broadcast announcements.
reports from the technical management agency (TMA).
Fig. 3. This is a sample summary sheet prepared for one health clinic, showing a partially paid claim (QC11=Wrong consultation fees), and another claim
quarantined for having the wrong voucher. While these summary reports can be informative, most clinics are not familiar with the quarantine codes, and they
often don’t reach the health clinic for several months after the original claim has been submitted, often too late for the clinic to rectify any errors indicated
on the report.
(a doctor) who “vets” the paper claims1 . They then produce software vendor. Again, while this situation is being resolved,
two reports: a summary report for all clinics, and an itemized the TMA asserts its control over the project by processing the
report (see Figure 3) for each clinic detailing the status of each claims, but refusing to pass on the summary reports to the
claim, as well as any quarantine codes (Table II) or rejection FMA. While the project continues running, and the service
reasons for any partially paid or rejected claims. providers continue to see patients, this introduces additional
In addition, when required, the TMA produces reports delays into the claims process, and frustrates the health clinics,
(based on the information in the database) for the FMA, the whose payments are delayed without explanation.
European aid agency, and the aid agency’s evaluating partners.
E. Service Providers: Health Clinics/Hospitals
Although these reports were not part of the original specified
mandates for the operation, they have proved necessary for Service Providers are selected on the basis of a number of
the program’s external evaluation, and there has been much factors (e.g., services offered, capacity, personnel, geographi-
friction over the work involved in the creation of reports. cal location). In one respect, they are the origin of the primary
The relationship between the TMA and the FMA in this information in the claims management process, producing
program is highly contentious. While initially they were equal the claims records, which are then used to determine reim-
partners in the program, both reporting directly to the Euro- bursement. At the same time, as is perhaps typical, they are
pean aid agency, changes in funding have led to a situation in information-poor, because they are not given tools to use this
which the TMA reports to and is paid by the FMA. On top of information effectively. At the point of claims submission, they
this, the funding for the expansion of the program has been are no longer agents in the process, and must wait passively
delayed a number of times, from October 2007 to April 2008, for both payments and any feedback reports produced from
and again to September 2008. While the TMA has received the information in their claims.
some payment, both the TMA and FMA have been operating Code Description
QC01 No indication of date of treatment
without pay (but with promise of pay) since April 2008, just QC02 No indication of time of treatment
QC03 Wrong visit type: Consultation or follow-up, etc
to keep the program running while the aid agencies work out QC04 Wrong demographic information: no age and name of client
QC05 Wrong/No syndrome, no diagnosis
the details of the new grant and the expanded program. This QC06 Wrong Clinical examination / not applicable to OBA
QC07 Wrong/Poor diagnosis
is part of the reality of dealing with aid-funded projects – QC08 Wrong investigation/poor lab reporting
QC09 Wrong drugs prescribed/invalid treatment
unexpected delays in funding are common, and projects are QC10 Over prescription: more than enough
QC11 Wrong consultation fees
subject to the vagaries of arbitrary rebudgeting. While the QC12 Wrong patient status: cured or not cured
QC13 Next date of visit: wrong or not filled in
FMA is often powerless to address the issue, in this case, QC14 Wrong voucher attachment/interchanged vouchers on followup visits
QC15 Unclear claim/uncharged claim and treatment contradicts other visits
the TMA often chooses to respond by cutting off program QC16 Partner treated on client form
QC17 Exceeded VMUS ceiling limit
access to the database, ceasing claims processing and refusing QC18 Treated syndromically and asymptomatically
QC19 Unclear/wornout/blank attached vouchers
requests for reports, until their problems have been resolved. QC20 Claim without patient thumbprint
QC21 No voucher attachment
Perhaps in response to these database shutdowns, but of- QC22 No doctor’s signature
QC23 Diagnosis contradicts clinical examination
ficially as part of the aid agency’s project policy and the QC24 Used drugs not on OBA list
QC25 Undercharged/overcharged drugs, double lab charged
Ugandan government’s policy on software developed for QC26 Patient free/normal from STI or cured not allowed for next visit
QC27 Diagnosis not catered for by project
government-funded projects, there is an expectation that the QC28 Follow-up contradicts previous visits
voucher, and has the patient sign and fingerprint the form, at and a revised version of the claim form. In addition, all of
which point their participation in the claims form process is the codes and tables in the web application database also
complete, until they return for a follow-up consultation. For the include mappings to their equivalents in the TMA database,
follow-up, the service provider checks recovery progress and so the information can easily be transferred between the two
prescribes additional medication if necessary. In some clinics, databases.
patients are given a copy of the claim form, which they are
directed to keep and bring back for the follow-up. However, A. Claim Mobile
most clinics do not depend on the patient copy of the claim The web application is designed primarily with three func-
form, and just go back through their time-ordered record book, tions in mind: claims submission, feedback/communications
finding the prior consultation manually. Sometimes patients process, and in-clinic claims management.
either accidentally swap vouchers with those of their sexual Claims Submission: This is the bulk of where the service
partners, fraudulently give their own voucher or the partner’s providers will spend their time. In this case, we adopt common
voucher to someone else, or simply choose to go to a different design strategies such as (1) using pre-filled checkboxes to
clinic for follow-up. Claims are quarantined or rejected if any reduce the amount of required text entry, (2) limiting answers
of these potential errors are detected, but not until the claim to valid options to reduce coding errors (see QC01,02, 04-09,
has been processed by the TMA, and the fraudulent patient has 27 in Table II), (3) downloading logistical data such as drug
already been treated. Since the original voucher is attached to prices into the application to eliminate pricing errors, and (4)
the submitted claim, the clinics do not always have a way of calculating dependent values such as expected claim amount
verifying these external aspects of voucher validity. Although to eliminate arithmetic errors and save time. However, we
their direct involvement in the claims process is minimal, it is must counter-balance potential fraud by also introducing cross-
their identity that is often contested in the vetting process. checks that are not clarified explicitly. That is, to encourage
G. Mobile Platform Developer (researchers) accurate clinical reporting (as opposed to clinical reporting
that has been “fiddled” to make electronic claim submission
As ICTD researchers, we are of course also stakeholders
more convenient or favorable), providers are allowed to submit
the claims management process – initially as outside observers,
inconsistent claims but are warned that they should clarify any
later as designers interested in using technology to measurably
discrepancies from normal OBA treatment protocol.
improve the process, and finally as researchers interested in
watching the mechanisms by which the process changes over Closing the Feedback Loop: Based on our primary finding
the course of the project. From an outside perspective, our role from the clinic surveys and follow-up interviews, we have
is most allied with TMA, the technical partner in the project; also included the ability for the clinics to send queries to the
however, since the aid agencies and FMA are interested in management agencies about particular claims and to receive
replicating the mobile device system in other OBA projects, live updates on the a claim’s status (e.g., whether it has been
there is a vested interest in the new technology from other approved, the amount for which a claim has been approved,
stakeholders as well. and explanations why the full amount may not have been
approved – see Figure 6). Any changes to a claim’s status
V. D ESIGN are included in this annotation audit trail, and anyone with
Claim Mobile is a two-part system, including a web-based access to the claim can respond to and receive queries.
PHP/MySQL application and a Java-based mobile application
running on Palm (GarnetOS) phones. For the pilot program,
the web-based application has a single level for all users, but
the final implementation will be tiered, having appropriate
access levels for service providers, management agency users,
medical advisors, etc. Both the web and the phone applications
require user login to protect patient data.
The two applications are paired, designed such that the
phone-based application uploads claims to the web site, and
downloads configuration information (drug lists, status feed-
back, claim form backups) from the web site. Eventually,
the web application will also connect to the TMA database, Fig. 6. On this screen the service provider can view the current status of their
sharing the cross-checked and validated claims form data claim, as well as any annotations or feedback from the management agency
made in response to their queries.
directly so the TMA’s staff do not have to do redundant data
entry.
To facilitate end-user training, both of the applications In-Clinic Claims Management: In the phone application,
are based on the original claim form and largely retain the the service providers can also link between consultations and
same structure, titles, and information. Figure 5 illustrates follow-up visits, as well as between client and partner visits,
some of the mappings between the phone-based user interface so they can easily check to see if the valid voucher is being
42
Fig. 5. From paper form to mobile phone: a mapping of the mobile phone interface equivalents for each section of the claim form.
partner associated with that particular voucher. This allows the inconsistency is in lab reporting – some lab tests require a
viewer to correlate treatments, lab tests, and diagnoses across value to indicate the result, and where not included, the data
visits. entry clerk just changes the lab test in the database to one
which does not require a result. This is an error, which never
C. The Modified Claims Process gets communicated back to the service providers because only
In the modified claims process, the service providers con- errors which accompany a payment change are reported in the
tinue to complete and submit the claim forms. However, in claim summary sheet.
order to receive faster payment, as well as the immediate cross- The 18 claims from the urban clinic spanned dates from
checked feedback from the phones on claims completion, they February 16, 2008 through August 25, 2008. During the
also enter the data on the mobile phones, submitting each process of simultaneously entering some of the claims into
claim to the web application as it is completed. Prior to claims Claim Mobile with the service provider, we were able to
submission the status of the claim is “Unsubmitted (1),” after identify some problems: missing personal information, missing
which it can progress through a number of stages. The service next visit date, and wrong consultation fees. However, not
provider can verify that a claim has been successfully received all claims were entered into and reviewed via Claim Mobile,
if the claim status has been updated to “Under Medical Review and, as can be seen from Figure 3, three paper claims were
(2)” or “Under Administrative Review (3)” for medical advisor submitted with wrong consultation fees, an error that would
review or database cross-checking (validation of voucher) not have occurred with an electronic submission. In addition, a
respectively. If a digital claim has been verified, the TMA will fourth claim was submitted with the wrong voucher. We were
set its status to “Preliminary Approval (4)” until the paper able to catch this while entering the claim into the mobile
form with the voucher has been received. Once preliminary phone, noting that the voucher number did not match the
approval has been received, the clinic can be paid. If no consultation type, but too late to change the submission and
voucher is received, or the wrong voucher is attached to the retrieve the correct voucher. As a result, the claim has been
form, then the preliminary payment is subtracted from the next quarantined until the correct voucher is given to the program
month’s payment until the error is resolved. In the meantime, office in Mbarara.
the service provider can view status updates as they are made
to the web application and sent to the mobile phones, and can
send annotations on each claim, which then appear in the web
application and in the status update window (Figure 6) when
they are received.
VI. P RE -P ILOT R ESULTS
Having detailed the claims submission process, our findings
with respect to the various stakeholders in the OBA project,
and the design of the Claim Mobile system, we now discuss Fig. 9. Rural clinic staff entering data from claim forms into two of the
phones.
some specific results from our user studies.
The pre-pilot demonstration was designed to last one claims
cycle, following one week’s worth of claims (submitted in With regard to the digitally-submitted claims, we spent
parallel through Claim Mobile and on paper) for two clinics about a day training the staff in the rural clinic on how to
through the claims submission process. A total of 35 claims submit the claims, and returned later to spend another day in
were submitted to the web application, including the full training. They were very enthusiastic, and although only one
complement of 18 claims from the urban clinic, 12 out of person was actually responsible for submitting claims, they
the 86 paper claims from the rural clinic, and 5 additional were all training each other (Figure 9). However after the the
claims from the urban clinic following the pre-pilot study. researchers’ departure they have still not submitted any claims.
The claims from the rural clinic spanned August 9, 2008 It is unclear whether this is from technical difficulties or lack
to August 27, 2008. We observed three patient consultations of time to enter the claims into the phone. The urban clinic
during our two visits to this clinic, as well as the preparation has continued to submit claims, with five new claims arriving
of the summary sheets for the 86 claims, taking careful note in the two weeks since the pre-pilot study.
of what the service provider verified on each form. Notably, An interesting outcome from our observation of the claims
although “syndrome” is a required field (see QC05 in Table II), review process is that there are many errors that are made
it was left blank in almost all of the claim forms. In one that affect the quality of the data, but are never communi-
case, the drug was entered correctly, but with the wrong cated to the service provider, in part because they have no
reimbursement value, and in another case, a drug was entered, attached financial consequences. The annotation feature (see
but no reimbursement was claimed either in the subtotal or Section V.A) enables attachment of quarantine codes to any
the total. At the time of the claim approval process, they claims that were in error without affecting the payment of the
were not reimbursed for the drug, because it had not been claim, providing feedback to the service providers on how to
claimed in the amount, although it had been listed. Another better complete the claims in the future.
44
One concern that emerged from this proof-of-concept study, B. Information Poverty
however, was with the reliability and the speed of Internet In addition to trying to address delays, we have also tried
access in the TMA office where the claims processing was to address information asymmetry and information poverty
occurring. Accessing individual claims took a long time, and within the system, identifying where stakeholders are disad-
the online database was completely inaccessible when the vantaged by lack of information, or lack of tools with which
Internet connection was down, which occurred infrequently to use the information.
but noticeably often. It may not be desirable for the claims This past year’s move of the data processing from Mbarara
submission process to introduce a dependency on Internet to Kampala especially has further exacerbated the gap between
connectivity where it is unreliable. the people that have the information and the people that
Unfortunately, the financial and claim review aspect of the can make use of it. While limited remedies have been made
pre-pilot was halted early due to administrative and political to rectify the situation, with a paper-based process, these
reasons, the result of which is that payments in the OBA remedies have been ineffectual, leaving the PMO without
program as a whole have been halted; so, while the technical access to necessary claims data, including voucher numbers.
feasibility of the system has been demonstrated, the logistical Through extensive stakeholder interviews, in particular with
details are still in process. We found that while the TMA’s the program office in Mbarara, we have identified the need
database entry staff were enthusiastic at the prospect of for a transparently accessible database, with the ability to
spending more time reviewing claims and less time doing just generate reports based on the submitted claims data. While
data entry, their participation in the pre-pilot was limited by control over access to the database is a key means by which the
two factors: the press of other claims that still needed to be TMA asserts itself in the OBA project, this practice is highly
processed, and pressure from the TMA to be secretive about disruptive to the OBA program, causing deep difficulties for
the data being processed until certain political issues had been the health clinics and the program office, rather than affecting
sorted out. the financiers or the FMA.
The initial design of Claim Mobile, reflected the paper-
VII. D ISCUSSION based process, and directly submitted claims from the mobile
A. Understanding Delays in the System phones in the health clinics to the TMA, bypassing the PMO
entirely. In response to our findings, we developed Claim
Delays can occur in a number of places in the claims pro- Mobile Web as a means of re-engaging the staff of the PMO
cess, not all of which can be accounted for by the introduction in the mobile claims process. The integration of the new web
of mobile phones. However, there are three key bottlenecks: application database is specifically meant facilitate resolution
1) the delay in the health clinic between when the health of information gaps, not only sharing the information with the
clinic sees the patient and when the claim is submitted, 2) people that can make use of it, but also giving them the tools
the time it takes to process the claims, entering each on into they need to make sense of the information.
the database, and 3) administration of feedback to the health Likewise, for the mobile-phone application, we also specif-
clinics, especially in case of errors. ically do not design one-way system in which claims data is
Claim Mobile is able to address all three of these cases by going out and only money returns. Instead, the claims data
1) encouraging providers to submit claims as they see patients, created in the clinic is also used within the clinic to help them
2) reducing the data entry burden through the use of digital improve patient care, as well as the accuracy of future claims.
claims, and 3) eliminating the possibility of a number of errors,
and providing a digital feedback mechanism to supplement the C. Related Work
infrequent in-person feedback. There have been a number of recent technical projects on
However, another source of delay is the administrative and the use of ICTs for healthcare in Africa [8], [9], [10], [5],
political dynamics by which program administration halts, and specifically on mobile devices for healthcare in Africa [3],
although health distributors continue to sell vouchers, and [11], [12]. However, many of these projects are design-focused
health clinics continue to see patients. During these times and technology-driven, reflecting on designing a working
payments are delayed unexpectedly for undetermined lengths technological solution to complete a particular task, rather than
of time, as can be observed from the early termination of our reflecting on the role of the technology in the system and how
pre-pilot study. Understanding this particular delay is key: the various solutions or approaches might affect social processes
TMA halts the program by withholding information, specif- within the system.
ically claim reports, from other stakeholders in the system. Braa describes two action research projects to deploy the
This is possible because the database is owned and controlled Health Information System Program (HISP) in Cuba [13]
entirely by the TMA. What happens when another outside and in South Africa [8], using Actor Network Theory (ANT)
database is introduced, with independent control? In this case, to think about how human and non-human (e.g. documents,
the data was not so much the key as the data entry staff that events, software, standards) interact. He specifically addresses
were responsible for approving the claims and validating the the challenges of designing for the multiple levels of entities
vouchers. At the same time, it is unclear where Claim Mobile involved in district health information systems, able to com-
Web fits in with the political strategy of the TMA. pare deployments across Mozambique, India, South Africa,
45
and Cuba. However, these entirely government-based contexts the fieldwork and setting up this pilot, as well as their honest
are much more hierarchical than the highly disparate multi- feedback on this paper.
organizational context described here.
R EFERENCES
VIII. C ONCLUSIONS AND F UTURE W ORK [1] Karen G. Cheng, Francisco Ernesto, and Khai N. Truong. Participant
and interviewer attitudes toward handheld computers in the context of
There is clearly much additional work to be done, in which hiv/aids programs in sub-saharan africa. In CHI ’08: Proceeding of the
the lessons learned from this pre-pilot study will be applied twenty-sixth annual SIGCHI conference on Human factors in computing
in the development of a new version of Claim Mobile for a systems, pages 763–766, New York, NY, USA, 2008. ACM.
[2] Cynthia Casas and William LaJoie. Voxiva: Case Study. December
full pilot in Spring 2009. 2003.
The outcomes from this study were three-fold. Firstly, the [3] Brian DeRenzi, Neal Lesh, Tapan Parikh, Clayton Sims, Werner Maokla,
choice of the mobile phone as a platform was affirmed by Mwajuma Chemba, Yuna Hamisi, David S Hellenberg, Marc Mitchell,
and Gaetano Borriello. E-imci: Improving pediatric health care in low-
the health clinics, for reasons of battery life, design for income countries. In CHI ’08: Proceeding of the twenty-sixth annual
readability, portability (susceptibility to theft), and ease of data SIGCHI conference on Human factors in computing systems, pages 753–
entry. Where we were concerned about introducing “qwerty” 762, New York, NY, USA, 2008. ACM.
[4] E.S. Berner and M.J. Ball, Editors. Clinical Decision Support Systems:
keyboards to novice users, our fears were alleviated, and Theory and Practice. Springer-Verlag, 1998.
all of our users assured us that “we can learn,” which they [5] Henry Lucas. Information and communications technology for future
did, quickly. Secondly, the mobile platform is not a sufficient health systems in developing countries. Social Science & Medicine,
66:2122–2132, May 2008.
solution for this program, and alone has the potential to exac- [6] ITC - e-Choupal. http://www.itcportal.com/sets/echoupal frameset.htm.
erbate information asymmetries between the stakeholders. To [7] 2006 Venture Strategies and Mbarara University population survey. http:
address this issue, we complement the mobile phone platform //www.oba-uganda.net.
[8] Jørn Braa and Calle Hedberg. The Struggle for District-Based Health
with a web application. However, Internet accessibility issues Information Systems in South Africa. The Information Society, pages
may require further development to enable local hosting and 113 — 127, 2002.
synchronization of the web application [10], [14]. Our final [9] Tessa Tan-Torres Edejer. Disseminating health information in developing
countries: the role of the internet. BMJ (British Medical Journal), pages
point is related – we consider the plethora of stakeholders 797—800, 2000.
in this project, and note that as technology providers we are [10] Rowena Luk, Melissa Ho, and Paul M. Aoki. Asynchronous remote
not coming in as naturally neutral players. Our projects are medical consultation for Ghana. In CHI ’08: Proceeding of the
twenty-sixth annual SIGCHI conference on Human factors in computing
necessarily disruptive, and equally potentially disrupted by systems, pages 743–752, New York, NY, USA, 2008. ACM.
other dynamics within the program as a whole. As a result it [11] Tapan Parikh. Position Paper: Mobile Phones may be the Right Devices
for Supporting Developing World Accessibility, but is the WWW the
is necessary for us as researchers to position ourselves and our Right Service Delivery Model? In W4A at WWW2006. ACM, 2006.
designs carefully, making sure to take into account the needs [12] Tapan Parikh and Edward D Lazowska. Designing an architecture for
of all of the stakeholders, and not just our primary users. delivering mobile information services to the rural developing world. In
Proceedings of WWW2006. ACM, 2006.
IX. ACKNOWLEDGMENT [13] Jørn Braa, Ola Hodne Titlestad, and Johan Sæbø. Participatory Health
Information Systems Development in Cuba the Challenge of Addressing
We thank Ben Bellows for his contributions to this project, Multiple Levels in a Centralized Setting. In Proceedings of Participatory
and feedback on the paper. In addition, we would like to Design Conference 2004. ACM, 2004.
[14] Michael Demmer, Bowei Du, and Eric Brewer. Tierstore: A distributed
thank all the staff at our partnering organizations and in the file-system for challenged networks. In Proceedings of File and Storage
health clinics for their assistance and patience in conducting Technologies (FAST), 2008.
46
III. METHODOLOGY
B. Ethnography
Our research is the result of longitudinal qualitative and
Ethnographic results are from the CAICT project as well as
quantitative work done as a component of a larger project on
earlier fieldwork conducted by the first author during a six-
the effect of information technology on society. The Central
month residence in the region in 2000. Fieldwork data
Asia Information and Communication Technology Project
collection methods follow standard ethnographic format for
(CAICT) is a multi-year study of ICTs. The project goals
participant-observation and include field notes and
include investigation of how ICTs diffuse within societies,
photography.
how cultural issues affect technology usage, and how patterns
In addition, interview studies have been conducted with
of trust and confidence in media and institutions change over
various groups to further investigate issues relevant to survey
time as technology diffuses in diverse communities.
findings. Qualitative data reported in this paper are drawn
A. Survey from multiple separate studies conducted in 2005 and 2006
The project includes a yearly survey of four countries in including interviews with Internet users, interviews with
Central Asia: Kazakhstan, Kyrgyzstan, Tajikistan, and mobile phone users, interviews with computer gamers, and a
Uzbekistan. The quantitative results presented here are based design ethnography. In addition, ethnographic observations
on the nationwide survey of 1000 respondents in each of these are based on field notes collected during separate data
four countries age 15 and older, administered in 2006, 2007, collection trips in 2000, 2002, 2003, 2004, 2005, and 2006.
and 2008, for a total of 1000 respondents per year per country,
and a total of 12,000 respondents overall. Survey sample was IV. BEYOND STEREOTYPES OF GAMES AND GAMERS
based on census information for age, gender, ethnicity, and Before addressing the specific ways in which games
geographic location as released by each country‟s government. provide an important pathway to ICT engagement, it is
The survey includes multiple urban and rural sampling important to deconstruct some of the stereotypes of games and
48
gamers. Recent research has begun to establish that gamers in
developed countries are not all male and not all teenagers [33-
34], that in fact the audience for games is diverse in terms of
gender and age [1][22-24]. Studies of complex multi-player
gaming environments, for example, demonstrate that although
male players outnumber female players, it is those female
players who actually spend more time online playing [25-26].
Games themselves come in many flavors, including the violent
shooting games that gain so much press attention, but they
also include puzzle games and simulations. Game systems
such as the Nintendo Wii or games with alternative input
modes like Rock Band or Dance Dance Revolution have
similarly changed the cultural activity of games and broadened
the audiences to which they appeal [27].
While not identically diverse, gamers internationally also
Figure 2: Photograph of Starcraft competition
belie some stereotypes. While gamers are still more likely to
announcement outside Kazakh game club
be male and under 30, in our research population, female
respondents and those over 30 were as likely to play games as
they were to use the Internet, See Figure 1. This finding
30%
indicates that depending on the population segment, gaming is
25%
either a more used pathway to ICT use than the Internet, or it
is at least equivalent. 20%
10% 15%
9%
8% 10%
7% 5%
6%
5% 0%
Kyrgyzstan Kazakhstan Tajikistan Uzbekistan
4%
Female 8% 18% 5% 8%
3%
Male 17% 26% 18% 19%
2%
1% Figure 3: Male and Female Gamer Populations in Central
0%
Male Female Under 30 Over 30 Asia, 2008
Gamers 8% 5% 9% 4%
Internet users 5% 5% 6% 4% Internet cafes vary in quality of equipment and furniture.
Some businesses will have recent equipment, flat panel LCD
Figure 1: Demographics of gamers vs. Internet users screens, and glossy office furniture; others have older PCs,
CRT monitors, and plastic chairs. Often the ones the feel more
V. SETTING THE STAGE: GAMING CONTEXT IN CENTRAL ASIA like a business center are located in the center of the city and
Public gaming centers in Central Asia meet many of the cater to a mixed clientele. However, businesses located in
preconceptions of computer game culture, but they are also neighborhoods similarly can serve a diverse clientele,
nuanced environments. The crowds of young boys that are although not necessarily at the same time. And games exist in
seemingly ubiquitous are reminders of gamer stereotypes, a context of ICT diffusion overall, with Internet growth
although our survey results indicate that women do indeed creeping slowly and mobile use skyrocketing, see Figure 4.
play games (see Figure 3), although women tend to play at
home rather than in public cafes. 70%
Many of the popular game titles are familiar, but the style of 60%
play and the mechanics of getting the games to work in 50%
different infrastructure environments are unfamiliar. 40%
Counterstrike (CS) and Starcraft are literally everywhere, and 30%
kiosks on the streets sell countless numbers of CS mods, the 20%
cd covers in English and Russian. Starcraft competitions can 10%
be found in many of the capitol cities in the region, see Figure 0%
6, and the World Cyber Games are a coveted destination (and Mobile Users Computer Users Gamers Internet users
Central Asian countries have placed quite high in the World 2006 20% 25% 13% 9%
500.00
VII. SURVEY FINDINGS: LESS ADVANTAGED USERS MORE
400.00
LIKELY TO GAIN INTRODUCTION TO ICTS THROUGH GAMES
300.00
As discussed in the previous section, games can provide a
200.00
mechanism by which youth can be drawn into ICT training
centers, and games also provide a revenue stream for owners 100.00
Abstract—83% of the world population does not have access information that caters to a wider section of the population,
to Internet. Therefore there is a need for a simple and affordable thus leaving out the details for specific villages. One of the
interaction technology that can enable easy content creation and common methods of providing local information (such as a
dissemination for this population. In this paper, we present the
design, development and usage pattern of a VoiKiosk system health-related warning) is done through announcements using
that provides a voice-based kiosk solution for people in rural loud speakers in vehicles. Figure 1 shows one such vehicle
areas. This system is accessible by phone and thus meets the being used by a government healthcare official to announce
affordability and low literacy requirements. We present usability an upcoming health camp in the village.
results gathered from usage by more than 900 villagers during
four month of the on-field deployment of the system.
The on-field experiments suggest the importance of locally
created content in their own language for this population. The
system provides interesting insights about the manner in which
this community can create and manage information. Based on the
use of the system in the four months, the VoiKiosk also suggests a
mechanism to enable social networking for the rural population.
I. I NTRODUCTION
The total rural population in 2005 is about 51% worldwide.
If we look at less-developed nations, then this number is
even larger, at 57%, which translates to more than 3 billion
people. For India, the percentage of rural population is even
higher, 71%. Based on the projections in [1], this number is
not likely to change fast in the coming years (the percentage
yearly change from rural to urban is projected to be around
TriC
1 Pyr.mea.IT IBM India Research Lab June 24, 2008
1%). Even so, most of the HCI research has been focussed Fig. 1. Announcement of a government organised health camp.
on the interaction with computers, to which this population
does not have access. Therefore it is extremely interesting to There is a lack of enough information sources that can create
look at sections of this large population base and identify the sufficient locally relevant content such as doctor visit organised
information and communication needs, the means to achieve by the government, change in school timings, delay of the train
these needs, and study the interaction modalities of this service, list of movies in the village community, etc. If there
society. is a system that can enable this community to create their own
A majority of this 71% of rural population in India is content, the information and communication ecosystem can be
illiterate or at least textually-illiterate [2]. Therefore though significantly improved.
they can count currency and make phone calls using numbers, Since not much is studied about the rural population, to
the ability to read or write is very limited. The means of build an information system that can be used by rural popu-
interaction of this society is either completely voice-based or lation to create and access content, it is extremely important
paper-based. Some people (esp. the money lenders, shopkeep- to follow the following design steps:
ers) make paper notes of their business. Most people (such 1) Perform a needs-study to identify the information needs
as farmers) interact using voice, and get information through of this population. It is important to understand what
radio, television and other public announcement methods does a community consider as information. To enable
in villages. News papers continue to be another source of easy access of information, a meaningful categorisation
accessing information. is also needed.
However the common problem across these information 2) Follow a participatory design process to develop a low-
sources for these villagers is the lack of locally relevant con- cost, low-literacy solution. Since a majority of this rural
tent. Most news papers, radio and television programs provide population in India lives below USD 2 per day [3], it
57
is important that the solution is affordable. Moreover, also been done on improving the usability for mobile Internet
the low literacy considerations also need to be taken interaction. In [14], the authors highlight the various usability
in developing the end-user interface of this system. issues by studying the different critical comments by mobile
Participatory design process is important to study a Internet users. Despite all these solutions, the usefulness of
population that has not been exposed to interaction Internet information for people in developing countries is very
modalities beyond paper, radio and TV. limited due to the fact that the content that is relevant for
3) Observe usage pattern of the solution for the rural a villager is not available on the Internet [15]. This paper
population. Since not much research has been performed positions itself in the space of creation and dissemination of
in enabling local content creation, the usage pattern of locally relevant content and is supportive of the statement:
this population will be interesting to observe and can access to relevant information is key to the development of a
provide important insights. society.
In this paper, we follow these three principals to build
B. Alternate Models for Local Information Systems
VoiKiosk – a system for increasing reachability of information
kiosks in developing regions. A VoiKiosk is a voice based Community Radio systems have been studied as a alterna-
service available on the Telecom network [4]. As opposed to tive solution to provide information in rural areas. In [16],
PCs, the phone penetration in rural India has been significant the authors present a study that evaluates the effectiveness
and continues to grow. Also based on field studies and litera- of 10 community radio stations in different rural communi-
ture reviews semi-literate and illiterate people are more com- ties in South Africa and report the popularity of programs.
fortable with speech-based interfaces to access information A sustainability study of the community radio service is
services [5]. Leveraging the increased mobile penetration and presented in [17] for Nepalese radio stations. Extensions to
comfort with speech based interfaces, we present this alternate community radio service to involve audience participation
model to create and host voice-based kiosks (VoiKiosks). have been presented in [18]. Among other things, the success
We identify the information needs of villagers (Section III) of community radio lies in the availability of relevant content
in south of India by visiting several villages and talking for the rural communities.
to villagers and officials of an NGO that operates in these The CAM system [19] provides an interface to capture
villages for more than seven years. Based on the needs, we local data using the mobile phone camera. It has been applied
implement the VoiKiosk system (Section IV) and involve a for microfinance data capture [20] and for procurement and
limited number of users and the NGO (Section V) in the quality control in a rural cooperative [21]. Using camera phone
participatory design phase. We deploy the system live in one to capture paper content is another approach to create local
village and gather important usage statistics by four months content that can be used in several application domains. We
of live field deployment of the system (Section VI). We derive the importance of relevant content presented in the cited
conclude the paper by discussing important insights gained work and use these concepts in this paper.
during the entire process of needs-gathering, participatory
C. Speech and Mobile based Systems for Low-literacy Users
design, prototype testing and usage analysis.
Since textual literacy is a major issue in developing coun-
II. R ELATED W ORK tries, especially in rural areas, several interesting information
Several studies have suggested that rural communities have systems for this community are speech based. The community
very different information and communication needs and radio systems mentioned above are completely based on audio
patterns [6] [7]. Moreover, it has also been studied that information. Studies from a spoken dialog system for rural
technological innovation opportunities in rural areas have not India [22] show that users are able to navigate the dialog
been studied in great depth [8]. In [8], the authors present field system irrespective of their literacy levels. A speech-based
studies suggesting that use of a local stakeholder contributes system for providing health information to low-literate users
significantly to a better design of the technologies for rural has been presented in [23]. This paper reports that there is a
communities. While there are significant design challenges need to educate the users by creating a video to explain and
for the rural community, at the same time, the potential of demonstrate the system for a new user.
technology to improve their lifestyle is huge [9]. This forms Information systems based on low-end mobile devices
the motivation for the work presented in this paper. provide an affordable alternative to PC-based systems. A
system for providing information to farmers using mobiles
A. Information Access and Dissemination has been presented in [24]. The authors in [7] conclude that
In the current world of globalization, access to information exploring mobile social software holds significant potential as
is key to the development of a society [10]. The increasing an information tool in rural areas. The MobilED [25] and the
amount of information on the WWW [11] is a reflection of this AudioWiki [26] are recent solutions that address this space
fact. Not surprisingly, therefore, there have been several efforts by providing information to this population through mobile
to provide this information to people in rural areas. There are devices.
solutions in the network space to provide low-cost Internet The work presented in this section leads us to believe that
access access for rural areas [12], [13]. Significant work has creation and dissemination of locally relevant content through
58
speech interface using mobile devices can be used as an The NGO has a Village Coordination Officer (VCO) for
effective system to provide information needs of the rural every village. A VCO is usually a woman who has had 12
population in developing countries. System level technology years of formal education. The VCO can only understand
development work in this domain has been presented in [27] local language, Telugu. For every 3-5 villages, the NGO has
and [28] where the authors present technologies for content positioned a Nodal Coordination Officer (NCO) to oversee
creation and linking. A solution based on these technolo- activities in these villages. The NCO can understand English
gies [29] proposes to organize the unorganized urban poor and is a degree holder in any subject. The NCO can operate
businesses. In this paper, we extend the concepts of the the computer. Every village has a Gram Vikas Samitee (GVS)
World Wide Telecom Web [30] and apply them for providing that is a committee of 18 members from the village who
information solutions to the rural population in developing have volunteered to work with the NGO on the 18 different
countries. modules.
Village Population Families Mobiles Main DTMF Will pay? Information Required Remarks
Oper-
ators
Vandaram 2292 450 150 Airtel/ Depends Not Program schedule of Ashwini, In- Yes for
BSNL/ on initially. formation categorised on Modules, pilot
Idea/ value Pesticides, Medicines, Aqua infor-
Voda- of mation, sports updates, railway en-
fone infor- quiry.
mation.
Juvvala Palam 3800 850 600 Airtel/ Voice is same as Community news, agriculture info, Yes for
Tata/ better above NO to train, no to Electricity, doc- Pilot
BSNL/ tor visit, Teleconferencing infor-
Voda- mation announcement, panchayat
fone meetings, ankur channel program
schedule. Hub village for 4 lakh
population in nearby villages.
Cherukumilli 4047 1000 500 Airtel/ Voice After information about village, Pan-
Idea realising chayat meetings, GVS meetings
the value,
they
may. Not
initially.
Ibhimvaram 4600 3000 Idea/ Voice Yes Tractor renting, electricians, etc. GVS
Airtel panchayat meeting, national fish very in-
rates, V-agri, V-agri, v-agri terested
for
pilot.
related information such as distance education class schedule, utes, sports achievements, etc.
village committee meting schedule and minutes. Other infor- It was clear from the on-field interactions that a phone-
mation needs were based on agriculture information, commu- based information portal that enables interaction through voice
nity news and crop rates. For the Juvvala Palam village, people will be ideal from the end-user perspective. Such a system
were interested in knowing about electricians, mechanics and will ensure that the end-users will not need to learn any new
different rental options for farming equipments. Except for interaction modality nor a new device. Given the significant
Ibhimavaram, other villagers mentioned that the villagers may cell-phone penetration, a cell-phone based system can be used
not pay for the service initially, but once they can see value to fulfill the information needs of these villages.
from the information, they have the capacity to pay for the
information. IV. P ROTOTYPE IMPLEMENTATION AND TESTING
Based on these meetings, we identified the following cate- Of the 10 different categories of information identified in the
gories for which the information is required by villagers: previous section, we built a prototype of the VoiKiosk system
for four main categories. A VoiKiosk is a specific instance of
• V-Agri: Provide agriculture and aquaculture advice to a VoiceSite that can be used by villagers to create and access
farmers on cell-phones. locally relevant content. A VoiceSite can be thought of as a
• Job work: items and type required by merchants to that parallel to a Website, but which can be accessed by dialing
villagers can get business. a phone number and information can be listened rather than
• Availability of transport. being read or seen. Creation of a VoiceSite is made easy by
• Health: Mostly static info such as doctor timings, date the VoiGen system [27] through which anyone can call up the
for the eye camp, student camp VoiGen system and interact with it through voice. This can
• Aqua prices: Traders can upload prices of fish, prawns at enable any illiterate person to create her VoiceSite. Such a
which they will purchase. system enables easy local-content creation. All information in
• People: Static information about electricians, mechanics, the VoiceSite is stored as audio messages that are recorded by
carpenters of that area with phone numbers making a phone call to the system.
• Entertainment: Name of movies in the nearby theater. Since a VoiKiosk is a VoiceSite for the entire village,
• Ashwini program schedule different type of users can update content in different sections.
• General info about the village: Population, availability of We explain the type of users and their interaction with the
health center, schools, famous local food, etc. VoiKiosk in the next section. The kiosk operator (typically a
• Community news: GVS meeting timings, meeting min- NCO or a VCO) navigates through the VoiKiosk application
60
Create: Operator
to configure it to offer various services for the village. Figure Welcome
Message
Listen: All
Advertisement
Health
Message
Health
VoiKiosk Message Agri/Aqua-culture
Ashwini Advertisement
Health Advice Ad
Center
Message
Kiosk-operator uploading VoiGen Health
Schedule
Message Ad
information on VoiKiosk
Center section where a kiosk operator can post the latest Add an agri expert
Browse registered experts
news related to the distance education program. Back to go to service menu
• The Kiosk Operator has the permission to modify any cognitive load on the users is less if they are told what all they
content on the VoiKiosk. He can call the VoiKiosk from can say to the system.
his cell-phone and the system identifies his caller-id and
provides the right permissions.
• Experts are allowed to post expert comments on agricul-
ture and aquaculture for specific farmers. They dial the
phone number of the farmer for whom they have to post
a message. Experts can not post comments in any other
category.
• Farmers call in the system to listen to any message from
experts. They can listen to information in other categories
too.
• Villagers can post an advertisements, or listen to any other
general information in the VoiKiosk.
A kiosk operator is usually identified for the VoiKiosk. The
kiosk operator has permissions to add different categories, and
also to add and delete information in any category. We have
developed a kiosk template for a group of villages in Andhra
Fig. 5. Initial prototype testing with the farmers.
Pradesh.
A. Participatory Design Modifications From the experts perspective, they wanted that any advice
We developed the prototype system in consultation with the they post to the farmers should also be archived in the NGO
NGO and then took this to the field to get feedback from the IT infrastructure. The NGO believes that this database of
four different type of end-users. Since the Kiosk Operator has specific advices can act as a knowledge resource in the future.
the maximum number of options available on the VoiKiosk, we We created a FTP site so that all the content in the V-Agri
started our testing with 2 Kiosk Operators. The initial design section would be pushed to the NGO server at midnight. The
of the system was such that when the system expected a user duration of any message to be posted on the VoiKiosk was
input, there would be a beep sound at the end of the system restricted to 10 seconds in the initial design of the system.
utterance. However the beep sound was present only when The system automatically detects silence and identifies the
the system used to record user voice and not at the time when end of the utterance. However if there is some noise in the
the system would do a speech recognition on the voice. Since background, then the system is not able to detect the silence
this fact was not known to the users, they did not know when and the message recording continues. This can result in user
to start speaking. Sometimes they would respond too early frustration, both during message recording as well at the time
and sometimes they were too late. This resulted in increased of listening. The other option was to ask the users to press
frustration and so to make sure that they are speaking in the a specific key to explicitly let the system know the end of
interval when the system accepts their input, they kept on utterance. But this would involves a key input, which we
repeating the same utterance, such as “exit, exit, exit”. We wanted to minimize, given the profile of target users. So we
modified the VoiKiosk interface so that after every system had fixed the duration to 10 seconds. However this duration
input, there was a beep sound. The users could then figure was too less to record some of the messages from the experts.
out that they are always supposed to wait for a beep sound Since experts were more advanced users in terms of the device
before they respond. This improved the system performance familiarity, they could use the keypad while talking over the
significantly. phone. So we increased the duration to 30 secs for the Experts.
There was a system prompt that asked the following from They now have to press the # key to indicate the end of the
the users: “Do you want information on Health or Agriculture utterance in case of background noise.
or Ashwini Center Schedule or you want to about the people The participatory design process improved the system sig-
in this village?”. We expected that people will respond by nificantly. With the first version of the system, the Kiosk
saying either of Health, Agriculture, Ashwini Center or People. operator used to take about 10 minutes to reach the Ashwini
However the users used to answer Yes or No. So we had Center part of the VoiKiosk and then update a message. Now it
to change the prompt to “Please say Health to know about takes him (and any novice user) less than 2 minutes to update
health related information, or say Agriculture to ...”. Although a message.
these descriptive prompts are too lengthy and not so natural,
VI. U SABILITY OF VOI K IOSK
but this community preferred to listen to prompts that clearly
tell the user what to say. This was a very good insight from The VoiKiosk was tested with the villagers for a period of
these participatory design experiments. Users in village have four months. We present the usage statistics of the VoiKiosk
more patience and prefer simple descriptive system prompts system by analysing the data through two approaches. The
as compared to more natural sounding prompts. Perhaps the first approach presents the analysis by looking at the data as a
62
TABLE II
whole. This analysis is presented to demonstrate the increasing S UMMARY OF DIFFERENT SECTIONS VISITED BY USERS .
acceptability of the VoiKiosk system in the village. In the
second approach, we present the results on the backdrop of Description Number Percentage
Agriculture 1640 8%
the changes that were made in the system over time. This will Health 2383 11.6%
provide more details about the effects of the usability of the Ashwini Center 1749 8.5%
system with improved interactions. In this section, we will Advertisements 7492 36.5%
present the two results in more detail.
A. User Background and Village Demographics while there used to be about 50 calls per day at the start of
The pilot was deployed for villagers in the Juvvala Palam the pilot, the last week of the pilot has witnessed more than
village in South India. This village has a population of about 300 calls every day. The interesting thing to note is that the
4000 people that form about 850 families. About 70% of these NGO did not actively train the villagers to use the system nor
families have a cell-phone. The male/female ratio is 100/94 in did they advertise this service aggressively. We had started
this village. The main occupation of this village is agriculture, with a group of about 30 users who were initially contacted
especially paddy crop. Transportation is the main business to start using the system and provide us feedback. The caller
for this village. People rent their trucks, tractors and smaller base increased from these 30 users just by word of mouth
vehicles for a few days to people in nearby villages. These are and has reached even the neighbouring villages. We have had
used by farmers in their field and for a specific social function villagers from the neighboring villages calling and creating
such as a wedding. The village a health center where the doctor their advertisements on the VoiKiosk.
visits for two hours every day. There is one bank in the village 700
where about 10% of the families have a bank account. The
average monthly income for a family is roughly about 2000 600
rupees (USD 50). All cellphone users are comfortable dialing 500
Number of calls per day
24 Hours of day
The number of calls have been increasing since the pilot (starting from 12:00 midnight)
was launched. The graph in Figure 6 hints at the increased
acceptability of the VoiKiosk system over time. On an average, Fig. 7. Hourly breakup of number of calls in the day.
63
There were a total of 2528 advertisements recorded by behind this was that by listening to the advertisements, a
463 different users. Many people recorded their advertisement novice user will know what to speak in her advertisement.
multiple times over the period of four months. Although one When the number of advertisements increased beyond 20, we
person can record only one advertisement, people often call noticed that the users were still patiently listening to them
again either to improve their advertisement, or to record a and then would post their advertisement. This usage pattern
fresh advertisement. One caller had recorded his ad 62 times. provides (at least) one of the following two insights about
Though most callers had recorded it only once. The graph in the users:
Figure 8 shows the different callers and the number of times
they have recorded their advertisements.
This population is more patient and could listen to a long
70 list of information over phone as compared to what has been
observed with the western population’s experience with
60
spoken dialog systems.
Number of advertisement recordings
50 And/OR
The villagers ascertain certain value to posting an
40
advertisement and so are ready to wait for a long time in
30 order to get a chance to post.
20 When the number of advertisements increased beyond about
40, we provided option to jump to the next advertisement
10
by pressing a key on the phone keypad. Perhaps the users
0 were comfortable with navigating the VoiKiosk system by
1 18 35 52 69 86 103 120 137 154 171 188 205 222 239 256 273 290 307 324 341 358 375 392 409 426 443 460
Unique callers
then, and therefore they were able to learn the keypad
navigation. It should be noted at this point that during the
Fig. 8. Advertisement update frequency of callers. needs-gathering phase, the users had clearly mentioned that
navigation through keypad will not be easy and so a speech-
input method should be provided to navigate the VoiKiosk.
C. Improvements with Changes in Interface
Even then, the average time spent per-call was higher (104
During the four months of the live pilot, we made two seconds) after Modification-1, it reduced to 75 seconds when
changes to the system: we provided option to record without users having to listen to
1) Enabled easy navigation of advertisements by providing all advertisements. This is a clear insight that users were able
options to jump to next advertisement through keypad to interact more efficiently with the VoiKiosk system when we
input. provided keypad-based shortcuts for faster navigation.
2) Enabled easy recording of advertisement by allowing
users to record without them having to listen to any VII. D ISCUSSIONS
advertisement. The advertisement section attracted the most attention, and
The initial system was live for 70 days. After the first change, was accessed the most. Initially, we had a couple of users
the system was live for 21 days and then after the second (a mechanic and a truck owner) upload personal (classified-
change, it was live for 30 days. Table III shows the number of type) advertisements to the VoiKiosk. Inspite of these initial
calls, average call time, number of advertisements and unique straightforward examples, the villagers soon found some very
callers in the three situations. interesting and innovative uses for this section:
• An eighth grade student announced himself and his
TABLE III
C HANGE IN V OI K IOSK USAGE WITH SYSTEM CHANGE . I NITIAL S YSTEM mobile number.
IS THE FIRST VERSION THAT WAS DEPLOYED . M ODIFICATION -1 IS THE • A man uploaded his profile for matrimonial purposes (to
SYSTEM THAT ENABLED NAVIGATION OF ADVERTISEMENTS .
M ODIFICATION -2 ALLOWED EASY RECORDING OF ADVERTISEMENTS .
invite marriage proposals). He made several attempts till
he was satisfied that all the details he wanted to capture
System State No. of calls Average time Unique callers were recorded.
Initial System 6239 79 335
Modification-1 4437 104 227 – Another person created a ”response advertisement”
Modification-2 9820 75 412 commenting on the above profile!
• A pair of young parents recorded a message in their
It is interesting to note that over a period of these four child’s voice for the child’s grandparents to hear.
months, the number of advertisements have been increasing. • A politician posted a thank you note after winning a local
though they may have never heard of social networking on the the system over a period of four months. Users made about
Web, the need for social networking is natural and compelling, 20000 calls to this VoiKiosk system. We present the usage
and (c) they independently discovered the message/response- pattern obtained by different modifications in the VoiKiosk
to-message interface as well as voice mail (as obvious as these over these four months. We provide a detailed description of
might sound to us). how some of the services were used beyond what they were
Thus the advertisements section evolved to a message-board designed for. This leads to an interesting insight into the social
by innovative use of the interface by the villagers. networking applicability of the VoiKiosk system.
As was mentioned in the previous section, in an earlier The increasing use of the VoiKiosk system for the different
incarnation of the VoiKiosk, a villager had to listen to all the purposes leads us to believe that a voice-based mechanism for
advertisements before posting his own, but that did not seem local content creation is a very powerful interaction modality
to act as a deterrent. The villagers appear to be extremely to provide information and communication technologies in
patient in listening to all the information and spending a lot rural areas.
of time with the system, which may be a reflection of culture
or the lack of alternatives, most likely both. ACKNOWLEDGMENT
Of the 20499 calls received on the VoiKiosk system, only We thank Byrraju Foundation for supporting VoiKiosk in
2532 were for creating the content. Creating the content in Andhra Pradesh villages and for providing valuable insights.
the VoiKiosk is either by recording of an advertisement by
a villager or by recording of other messages by the Kiosk R EFERENCES
Operator or the agriculture Expert. Therefore, more than 87%
of the calls were for accesing the content on the VioKiosk [1] “World Urbanization Prospects: The 2007 Revision Population
system while only about 13% calls were made for creating Database,” http://esa.un.org/unup/.
[2] Ministry of Finance, Government of India, “Economic
the content. This indicates the usefulness of the content that Survey 2001-2002,” http://indiabudget.nic.in/es2001-
was being created on the VoiceSite. 02/chapt2002/chap106.pdf.
In the previous section, we have provided more details [3] Population Reference Bureau, “World Population Data Sheet,”
about the advertisements rather than the agriculture, health and no. ISSN 0085-8315, Aug 2006.
ashwini center. This is because the other three categories were [4] S. Agarwal, A. Kumar, A. A. Nanavati, and N. Rajput,
“Voikiosk: Increasing reachability of kiosks in developing re-
mostly information upload from one specific person (either
gions,” in Proc. Intl. Conf. on World Wide Web (WWW), China,
the Kiosk Operator or the Expert), and the access was to the April 2008.
villagers. Therefore these services, although important, were [5] T. S. Parikh and E. D. Lazowska, “Designing an Architecture for
different from the advertisements because the latter had user Delivering Mobile Information Services to the Rural Developing
generated content. By providing the ability for all villagers to World,” in Proc. Intl. Conf. on World Wide Web (WWW), May
2006.
create content, VoiKiosk becomes a much more participatory
[6] S. Seshagiri, A. Sagar, and D. Joshi, “Connecting the bottom
platform for rural areas. of the pyramid: An exploratory case study of india’s rural
Although the statistical analysis proves the acceptance of communication environment,” in Proc. Intl. Conf. on World Wide
the VoiKiosk system by the villagers, the next step is to Web (WWW), China, April 2007.
get usability feedback directly from the users. We plan to [7] B. Kolko, E. Rose, and E. Johnson, “Communication as
talk to a sample of VoiKiosk users and seek their feedback information-seeking: The case for mobile social software for
developing regions,” in Proc. Intl. Conf. on World Wide Web
in terms of the usability of the system and the value of (WWW), Canada, May 2007.
the content. We also plan to provide more critical services [8] D. Ramachandran, M. Kam, J. Chiu, J. Canny, and J. L.
such as daily crop prices to registered farmers and medication Frankel, “Social dynamics of early stage co-design in developing
reminders to patients through the VoiKiosk system. It will also regions,” in Proc. CHI, USA, April-May 2007.
be interesting to observe the usage pattern of a different village [9] C. K. Prahalad, “The Fortune at the Bottom of the Pyramid:
Eradicating Poverty Through Profits,” in Wharton School Pub-
and identify the similarities and differences and question lishing, 2004.
whether they reflect the demographics of that village. [10] E. Brynjolfsson and L. Hitt., “Paradox lost: Firm-level evidence
on the returns to information systems spending.” Management
VIII. C ONCLUSION Science, vol. 42, April 1996.
In this paper, we present a information kiosk system for [11] “Netcraft Web Server Survey,”
http://news.netcraft.com/archives/web server survey.html.
use in rural areas to create and access locally relevant content.
[12] K. Matthee, G. Mweemba, A. Pais, G. V. Stam, and M. Ri-
We identify the information needs in rural areas by talking to jken, “Bringing internet connectivity to rural zambia using a
villagers and the NGO. We develop VoiKiosk – a voice-based collaborative approach,” in Proc. Intl. Conf. on Information and
system that can be accessed by a phone to provide a cost Communication Technologies and Development (WWW), India,
effective solution that is easy to use by the less-literate people December 2007.
in rural areas. We test the initial prototype with the villagers to [13] U. Saif, A. L. Chudhary, S. Butt, N. F. Butt, and G. Murtaza,
“Internet for the developing world: Offline internet access at
evolve a participatory design of the final system. The system modem-speed dialup connections,” in Proc. Intl. Conf. on In-
was then deployed live in a village in South India and we formation and Communication Technologies and Development
captured user statistics for 4 months. More than 900 users used (ICTD), India, December 2007.
65
[14] G. Buchanan, S. Farrant, M. Jones, H. Thimbleby, G. Marsden, [23] J. Sherwani, N. Ali, S. Mirza, A. Fatma, Y. Memon, M. Karim,
and M. Pazzani, “Improving mobile internet usability,” in Proc. R. Tongia, and R. Rosenfeld, “Healthline: Speech-based access
Intl. Conf. on World Wide Web (WWW), Hong Kong, May 2001. to health information by low-literate users,” in Proc. Intl. Conf.
[15] R. Heeks, “Information Systems and Developing Countries: on Information and Communication Technologies and Develop-
Failure, Success, and Local Improvisations,” The Information ment (WWW), India, December 2007.
Society, vol. 18, March 2002. [24] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana
[16] E. R. Megwa, “Bridging the Digital Divide: Community Radio’s Unwired: Replacing PCs with Mobile Phones in a Rural Sug-
Potential for Extending Information and Communication Tech- arcane Cooperative,” in Proc. Intl. Conf. on Information and
nology Benefits to Poor Rural Communities in South Africa,” Communication Technologies and Development (WWW), India,
Howard Journal of Communications, vol. 18, October 2007. December 2007.
[17] F. Hussain and R. Tongia, “Community radio for development [25] “MobilED Initiative,” http://mobiled.uiah.fi/.
in south asia: A sustainability study,” in Proc. Intl. Conf. on [26] T. Leinonen, F. Aucamp, and E. Sari, “Audio Wiki for mobile
Information and Communication Technologies and Development communities: information system for the rest of Us,” in Work-
(WWW), India, December 2007. shop on speech in mobile and pervasive environments, Mobile
[18] S. R. Sterling, J. O’Brien, and J. K. Bennett, “Advancement HCI 06, September 2006.
through Interactive Radio,” in Proc. Intl. Conf. on Information [27] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A.
and Communication Technologies and Development (WWW), Nanavati, “Voiserv: Creation and delivery of converged services
India, December 2007. through voice for emerging economies,” in WoWMoM’07 Pro-
[19] T. S. Parikh, “Designing an Architecture for Delivering Mobile ceedings of the 2007 International Symposium on a World of
Information Services to the Rural Developing World,” Ph.D. Wireless, Mobile and Multimedia Networks, Finland, June 2007.
Dissertation, University of Washington, 2007. [28] S. Agarwal, D. Chakraborty, A. Kumar, A. A. Nanavati, and
[20] T. S. Parikh, P. Javid, S. K., K. Ghosh, and K. Toyama, “Mobile N. Rajput, “HSTP: Hyperspeech Transfer Protocol,” in ACM
Phones and Paper Documents: Evaluating a New Approach for Hypertext 2007, UK, September 2007.
Capturing Microfinance Data in Rural India,” in Proc. CHI, [29] A. Kumar, N. Rajput, S. Agarwal, D. Chakraborty, and A. A.
Canada, April 2006. Nanavati, “Removed for blind review,” in Proceedings of the
[21] Y. Schwartzman and T. S. Parikh, “Using CAM-equipped Mo- World Wide Web, April 2008.
bile Phones for Procurement and Quality Control at a Rural [30] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. Nana-
Coffee Cooperative,” in MobEA V, Mobile Web in the Developing vati, “Removed for blind review,” in SIGCOMM Workshop on
World, Canada, May 2007. Networked Systems for Developing Regions, Japan, Nov 2007.
[22] M. Plauche and M. Prabaker, “Tamil market: a spoken dialog [31] “Byrraju Foundation,” http://www.byrrajufoundation.org.
system for rural India,” in Proc. CHI, Canada, April 2006.
66
Abstract—While mobile phones have found broad application In migrating mobile data collection from PDAs to cell
in reporting health, financial, and environmental data, there has phones, a critical issue is ensuring the accuracy of data entry.
been little study of the possible errors incurred during mobile In the context of healthcare, an errant entry may prevent life-
data collection. This paper provides the first (to our knowledge)
quantitative evaluation of data entry accuracy on mobile phones saving treatments from reaching patients, or may cause the
in a resource-poor setting. Via a study of 13 users in Gujarat, prescription of unnecessary treatment that is costly and dan-
India, we evaluated three user interfaces: 1) electronic forms, gerous. In financial applications, entry errors may jeopardize
containing numeric fields and multiple-choice menus, 2) SMS, the economic standing of communities that are already very
where users enter delimited text messages according to printed poor. Due to the importance of this issue, several researchers
cue cards, and 3) voice, where users call an operator and dictate
the data in real-time. have studied the error rates incurred as PDAs are deployed
Our results indicate error rates (per datum entered) of 4.2% in developing regions. As detailed later (in Table II), the
for electronic forms, 4.8% for SMS, and 0.45% for voice. These error rates are generally less than 2% (i.e., 2 errors per 100
results caused us to migrate our own initiative (a tuberculosis entries) in programs where users received at least an hour
treatment program in rural India) from electronic forms to voice, of training [12], [24], [4]. However, in the context of mobile
in order to avoid errors on critical health data. While our study
has some limitations, including varied backgrounds and training phones, studies of data accuracy are distinctly lacking. The
of participants, it suggests that some care is needed in deploying closest work is by Parikh et al., where a hybrid system of paper
electronic interfaces in resource-poor settings. Further, it raises forms and camera-equipped mobile phones has demonstrated
the possibility of using voice as a low-tech, high-accuracy, and error rates of less than 1% [28]. For standalone data collection
cost-effective interface for mobile data collection. on low-end phones, we are unaware of any previous study
I. I NTRODUCTION with a rigorous evaluation of data accuracy. This research
Mobile devices have shown great promise for improving the opportunity is highlighted in Table I.
efficiency and effectiveness of data collection in resource-poor In this paper, we provide a quantitative evaluation of data
environments. Compared to a traditional process that relies entry accuracy using low-cost mobile phones in a resource-
on paper-and-pencil forms with subsequent transcription to a constrained environment. We evaluate three practical user
computer system, mobile devices offer immediate digitization interfaces for entering data on a mobile phone: electronic
of collected data at the point of survey. This allows for fast forms, SMS, and voice. Electronic forms consist of numeric
and automated data aggregation. It also improves adherence fields and multiple-choice menus, and can be implemented in
to complex or context-dependent questionnaires, as the device Java or a native phone platform. The SMS interface requires
determines which questions should be answered or skipped. users to send a structured SMS messages to a server, with
The benefits of mobile data collection have been demon- logical fields separated by delimiters in the message. The voice
strated mostly in the context of personal digital assistants interface represents a normal telephone call, with a live human
(or PDAs) [31], [10], [8], [2], [32], [12], [24], [9], [16], operator that enters the data into a centralized spreadsheet.
[4], [3], [15]. Given the recent explosion of mobile phones We evaluated these interfaces in a study of 13 health workers
around the world, there is growing excitement in extending and paramedical staff over a month-long period in Gujarat,
the successes achieved on PDAs to a phone-based platform. India. Each participant was trained and evaluated on all of the
While high-end phones provide the same capabilities as PDAs, interfaces. We focus on the collection of health data relevant
low-end phones lack features such as high-resolution displays to tuberculosis (TB), as we anticipate deploying an electronic
and touch-screen capabilities. To empower the full population system in a real TB treatment program. The data in this paper
of nearly 4 billion mobile phone subscribers [26] with the represent only simulated patient interactions.
capabilities of mobile data reporting, it will be important to Our results indicate an error rate of 4.2% for electonic
establish usable interfaces that are portable to inexpensive forms, 4.5% for SMS, and 0.45% for voice. These represent
phones, and there have been a number of recent efforts in the fraction of questions that were answered incorrectly; as
this space (see for example [13], [1], [22], [25], [7]). each patient interaction consisted of eleven questions, the
probability of error somewhere in a patient report is much
Manuscript received September 22, 2008. higher. For both electronic forms and SMS, 10 out of 26
Somani Patnaik and Emma Brunskill are with the Massachusetts Institute reports (38%) contained an error; for voice, only 1 out
of Technology (email: somanip@mit.edu, emma@csail.mit.edu). During this
research, William Thies was affiliated with both the Massachusetts Institute of 20 reports (5%) contained an error (which was due to
of Technology and Microsoft Research India (email: thies@microsoft.com). operator transcription). As detailed in Section VI, error rates
75
SATELLIFE [15]
DataDyne EpiSurveyor [31]
Cell-Life in South Africa [13]
EpiHandy [10]
Jiva TeleDoc in India [1, p.42]
Infant health in Tanzania [32]
Other programs Pesinet in Mali [22]
e-IMCI project in Tanzania [8]
Malaria monitoring in Kenya [25]
Respiratory health in Kenya [9]
Voxiva Cell-PREVEN in Peru [7]
Tobacco survey in India [16]
Ca:sh project in India [2]
TABLE I
P REVIOUS WORK IN EVALUATING THE ACCURACY OF MOBILE DATA COLLECTION IN THE DEVELOPING WORLD .
are distinctly higher for health workers than for hospital staff, of each interface (Section IV). We describe the setup of our
though this difference may also be influenced by variations in user study (Section V) and the results obtained (Section VI),
our training environment. and we discuss the implications (Section VII). We conclude
We were surprised and alarmed by these results. In our own in Section VIII.
treatment program, our original intent was to utilize electronic II. R ELATED W ORK
forms. However, we consider it to be an unacceptable risk
As summarized in Table I, there have been several initiatives
that 38% of submitted forms – containing critical health
to apply PDAs and cell phones for mobile data collection in
information – may contain errors. For this reason, we have
the developing world. While a fraction of the studies on PDAs
overhauled our plans and will implement a treatment program
includes an experimental analysis of the error rate incurred, we
using voice rather than forms or SMS. While the cost of a
are unaware of any study which systematically measures the
live operator may be prohibitive in many countries, in India it
accuracy of data entry on a cell phone. This is the principal
proves to be very cost-effective. The increased cost of a human
novelty of our work.
operator is more than compensated by the decreased cost
Lane et al. provides a review of nine randomized controlled
of voice-only handsets, voice-only cellular plans, decreased
trials that compare the effectiveness of PDAs and paper
training time, and decreased literacy requirements for health
forms for data collection [21]. Six of the trials reported entry
workers. We offer a more detailed analysis in Section VII.
accuracy, with varying results: two studies found PDAs to be
While the results of this study have changed our own ap- more accurate than paper [20], [29], three studies found the
proach to implementing mobile data collection, we caution the accuracy to be similar with both methods [17], [23], [36], and
reader in extending the results of the study beyond its original one study found that paper was more accurate [35]. None of
context. In particular, we are focused on the scenario in which the trials were in the context of the developing world (they
users have limited cell phone familiarity and there is limited took place in North America and Europe).
time to perform training. If either of these variables changes, Previous studies of PDA entry accuracy in the developing
it may be possible to implement high-accuracy mobile data world are summarized in Table II. In cases where workers
collection with electronic forms or SMS. Also, while the error received at least an hour of training, error rates are under 2%
rates that we report on mobile phones are 3-8x higher than (i.e., 2 errors per 100 questions). As early as 1991, Forster et.
those previously reported for PDAs, our data are unable to al evaluated the use of PDAs for a malaria morbidity study in
distinguish whether this difference is due to the devices, or the Gambia [12]. Employing secondary-educated workers who
due to other aspects of the study demographics, training, and received five days of training, they report error rates between
evaluation. A future study could address this question directly 0.1-0.6% and argue that the PDAs offer improved accuracy
by evaluating both phones and PDAs in the same context. and efficiency over paper forms. Missinou et al. employed
Despite these limitations, our study is the first (to our PDAs in a clinical study in Gabon, employing four clinicians
knowledge) that evaluates data entry accuracy on mobile who had no prior PDA experience and received 8 hours of
phones. Based on our results, we submit only that electronic training [24]. They report a 1.7% rate of discrepancy between
forms and SMS may need further validation before gaining PDAs and paper forms, and note that clinicians preferred
widespread deployment in accuracy-critical applications, and the PDAs. Blaya et al. found that error rates improved from
that voice may deserve more attention as a high-accuracy and 1.3% (with paper forms) to 0.37% (with PDAs) in reporting
low-cost means of data collection. tuberculosis bacteriology data in Peru1 [4]. The authors also
The rest of this paper is organized as follows. We start by argue that PDAs are cost-effective [5].
reviewing related work on mobile data collection (Section II). 1 Blaya et al. reports errors per form, rather than errors per entry [4]. Via
Then we consider the tradeoffs between electronic forms, personal communication with the author, we determined that there were an
SMS, and voice (Section III) and detail our implementation average of 7.5 entries per form, yielding the error rates quoted here.
76
Higher error rates have been reported in the case of prior to data entry. Users that lacked prior camera or computer
self-administered surveys, when limited training is possible. experience were trained to a level of comfort within 5 to 15
Bernabe-Ortiz et al. evaluate the use of PDAs for surveys of minutes. A separate study measures error rates of 1% or below
sexual behavior in Peru [3]. To protect patient privacy, the using the CAM system [28]. This represents an interesting and
PDAs were intended for use by actual subjects, rather than by useful design point, especially in cases where paper forms are
health workers. As only some subjects had finished secondary already ingrained into the workflow. We focus on solutions
education, and subjects received only 2-3 minutes of training, that are independent of any paper workflow, and which do not
the authors observed a 14% discrepancy between electronic necessarily require a camera-phone (while Java-phones often
and paper forms. However, the error rate was substantially have cameras, our SMS and voice solutions are suitable to the
lower for subjects who had finished secondary schooling. most inexpensive phones).
Additional programs have applied PDAs for data collection While electronic forms have been widely deployed, there
in the developing world, but have not provided a rigorous are fewer solutions that rely on user-constructed SMS mes-
analysis of entry accuracy. SATELLIFE uses PDAs for dis- sages for mobile data collection. One example is a system
seminating and collecting medical information in numerous from Dimagi, Inc. which monitors water treatment plants in
countries [15]. There are anecdotal reports that the PDAs India [34]. We are unaware of other systems which rely on
improved data quality [19], and the benefits of decreased error a cue card (as we do in our evaluation) for submitting a
rates were estimated on a five-point scale [6]. Users of the structured SMS message to a server.
system have also rated its usability [11]. However, we are un- Others have considered broader issues in the contextual
aware of a quantitative assessment of the error rates incurred. design of user interfaces for data collection in the developing
DataDyne EpiSurveyor [31] has been widely deployed for data world. Examples include interface design for Auxiliary Nurse
collection in Sub-Saharan Africa; while it has been argued that Midwives in India [14] and a methodological framework for
the system is more accurate than paper forms [30], we are evaluating health devices [18]. Our focus is on assessing the
unaware of a controlled study. EpiHandy also provides tools entry accuracy for a range of standard interfaces.
for deploying electronic forms on PDAs and has been deployed
in South Africa, Uganda, and elsewhere [10]. PDAs have III. U SER I NTERFACES
also found application for gathering infant mortality data in Three of the central modes on a cell phone that can be used
Tanzania [32], for pediatric care (as part of the e-IMCI project) to perform data collection are voice, SMS and an electronic
in Tanzania [8], for assessing respiratory health in Kenya [9], forms application. Data collection performed by voice can be
for surveying tobacco use in India [16] and for maternal and further split into systems that link the data collector with a live
child health (as part of the Ca:sh project) in India [2]. These operator, those that connect to an automated interactive voice
studies lack formal evaluations of entry accuracy. response system, and those that allow the user to record a
Cell phones have also found broad application for mobile message. We focus our discussion around live voice operators,
data collection in the developing world. Cell-Life employs SMS and electronic form based systems, and examine some
electronic forms on mobile phones to improve TB and HIV of the strengths and weaknesses of these various approaches.
treatment in South Africa [13], [33]. Electronic forms are We use SMS to refer to data collection systems that involve
also used by Jiva TeleDoc for improving rural healthcare in information entered by a structured text message: in particular
India [1, p.42], and by Pesinet for monitoring infant health we assume that the information is entered by following a small
in Mali [22]. Mobile phones with forms are also being used cue sheet with a flowchart that directs the collector how to
to monitor malaria in Kenya [25]; while PDAs were also enter the data. To our knowledge, using cue cards to guide
piloted, the authors note that phones are more intuitive due data entry by text message has not been done previously.
to worker familiarity. Voxiva’s Cell-PREVEN uses interactive In contrast, electronic forms (particularly on personal digital
voice response and voice recording to monitor adverse events assistants) have been widely used. In this paper, we use the
amongst sex workers in Peru [7]. We are unaware of any term “electronic forms” to denote any external application that
quantitative evaluation of entry accuracy in these projects. can be placed on a phone, and that automatically guides the
To avoid the complexities of navigating electronic forms, user how to enter data, through the use of text, menus or other
the CAM framework offers a hybrid system in which paper tools. In a voice operator interface, the user simply calls a live
forms are used for organization while phones are used for operator, who asks the user a series of questions to elicit the
data entry [27]. Each field on the paper form is annotated information needed. Figure 1 illustrates each interface as used
with a barcode, which is recognized by a camera on the phone in our particular experiment.
77
1 2 3
Electronic Forms Interface
General Strengths
Easy patient identification
Ongoing cost is low (SMS or data plan)
Can store visits when connectivity is poor
a) General Weaknesses
Requires programmable phones
Requires basic literacy skills
Hard to alter survey questions
Hard to enter in free-form notes Electronic form screenshots
Application can be deleted by user 1. Start application
2. Select the patient
Our Results: Accuracy & Efficiency 3. Select "new visit"
We measured 4.2 errors per 100 entries 4. Enter the patient's temperature
The average interaction was 99 seconds 5. Select severity of patient’s cough
4 5
Fig. 1. The three user interfaces evaluated in this paper: a) electronic forms, b) SMS + cue card, and c) voice.
78
In general, there are a variety of factors that affect the data. High quality data can perhaps be characterized by two
choice of a data collection interface. These may be loosely simple criteria: whether or not the data is intentionally faked
categorized into operation, effectiveness and cost. Figure 1 by the user, and the accuracy of data that is not intentionally
supplements the below discussion by summarizing some of faked (which is the focus of this paper).
the strengths and weaknesses of each interface. Intentionally faked data can lead to incorrect conclusions
and potentially lead to significant misallocation of resources
A. Operation when interventions are based on false data. There may be an
We use “operation” to refer to factors involved with the incentive to fake data when users are busy and collecting real
general infrastructure of the data collection system. Initially data is time consuming, due to the data recording itself or
there is the investment of time to set up the system, and transportation time to reach the source of the data (such as
then train the workers who will be performing data collection visiting remote patients). Unfortunately in SMS systems it is
on the system. For voice or SMS interfaces, the set up quite easy to fake data, particularly for cell phone savvy users
time for workers is minimal: each worker must simply be that can copy and paste prior SMS messages. Faking electronic
provided with a phone, if he or she does not currently own forms is slightly harder as it requires the user to sequentially
one. However, electronic forms require that the application be fabricate data across an entire form. It requires the most effort
downloaded onto the phone, which requires either an Internet- for users to fake data while speaking on the phone, as the
enabled phone in an area of good connectivity, or specialized operator can always ask a new question to try to ascertain if
development tools and an external computer. the user is fabricating the data.
Training time for each application is an open issue, and is Voice also has the benefit that it is easy for users to convey
one of the factors we investigate further in this study. Worker additional information (not included in the original survey),
education and worker cell phone familiarity are likely to affect whereas it is more challenging to spell out text using the
how easy it is to set up each user with an interface, and train keypad, particularly in other languages which may or may not
them how to use it. We expect that a voice interface requires be supported on a given phone. Voice is also likely to have
the least amount of education and background to get users fewer operational risks: users may accidentally delete the form
equipped to start performing data collection. In particular, a application, or forget their SMS cue card, but since an operator
voice interface does not require that its users be literate. can always call a worker directly, the voice system is fairly
System coverage and reliability are also critical factors to robust. Voice also makes it easy for users to correct previous
ensure good data collection. Voice calls have priority over visits, by simply calling back the operator. This is also easy
SMS, and there is the possibility of lost SMS messages. to do by modifying and resubmitting a saved electronic form.
The delivery mechanism with electronic forms can vary: both However, it is also important to consider the speed of data
GPRS and SMS can be used. GPRS has the advantage that entry, how much the user likes the interface, and the accuracy
there is an acknowledgment of whether the data was sent; of data entry. To our knowledge there are no prior studies
however not all locations have coverage. From the user side, comparing the accuracy and speed of data entry using SMS,
voice appears to be the most reliable and has the most far electronic forms and voice. Since we regard these as some of
reaching coverage; however, this also requires that there exists the most critical factors in choosing an interface, this is a large
a sufficient number of operators so that users can always reach motivation for our current study.
a person when they call. If this is not always possible then
there may be a reliability penalty as users may have to call C. Cost
back later (or wait for the operator to return their call).
In addition to reliability, a good system should enable One of the other important considerations is cost: the most
some degree of flexibility. Despite good initial prototyping, beautiful, user-friendly, accurate interface may still not be
it may sometimes be important to be able to modify the practical if the cost overhead is too high for the particular
data collection interface, fix an error, improve usability, or problem. Costs consist of fixed one time costs as well as
add or remove information to be collected. If users have an ongoing marginal costs.
Internet enabled phone and are always working in areas of For all three interfaces users must have a cell phone. An
high connectivity, then updating an electronic form system is electronic form requires a programmable phone (such as a
quite feasible. However, if this is not the case, then users must Java-enabled phone or Windows phone) but both SMS and
reprogram their phone using the same specialized tools needed voice applications can be used with any phone. The ongoing
for initial set up. SMS is similarly challenging to update since cost for an SMS phone depends on the rate per message
a new cue card must be distributed to direct the user to enter which is typically quite low. An electronic form can send data
the data. In contrast, voice is trivial to update, as the operator using SMS or through a data plan; typically SMS is cheaper
can simply ask a new set of questions. depending on the amount of data that is being collected.
Voice minutes are frequently more expensive than SMS. But
B. Effectiveness most importantly, voice has the ongoing cost of the salary of
In any data collection effort, one of the key considerations the operator, which is an additional overhead not shared by
is the effectiveness of the program at obtaining high quality electronic forms or SMS.
79
IV. U SER I NTERFACE I NSTANTIATION The electronic forms underwent several design iterations,
including gathering feedback from a 3-day session with 22
The prior section discussed some of the general factors
health workers in Bihar, India, prior to this study. Based on
important to consider when designing and selecting a data
feedback from the workers in Bihar, we choose to employ
collection interface. We now discuss the context for our data
hybrid English/Hindi menus for some of the forms, since some
collection effort and the interfaces we evaluated.
medical terms are easier to understand in English, but others
A. Domain context are easier to understand in Hindi. We also changed from using
multi-select lists (with a checkbox per symptom) to using
Soon the authors, along with other collaborators, intend to individual yes/no questions.
conduct a trial that examines whether increased information Figure 1a shows a series of screenshots of the form interface
and monitoring can improve heath outcomes and adherence used for the present study. The Java application can be set
during tuberculosis treatment in Bihar, India. Treatment will up to either relay this information via SMS or GPRS. This
be conducted by having tuberculosis patients regularly visit distinction is important for cost considerations but does not
health workers and receive drugs as part of a directly observed affect the interface testing considered here.
therapy (DOT) strategy. During these visits, health workers
will collect data about their patients and report this information C. SMS implementation
by mobile phone back to a central office. This information For the SMS interface we designed a cue card that instructs
will be aggregated and analyzed to inform doctors and the the worker how to record information about the patient into a
trial manager about which patients may need to be visited, text message; Figure 1b displays a subset of the cue card used.
for example, if a patient is not improving or is experiencing All information is coded numerically; this is done to reduce
adverse side effects. To support this effort we need a user the amount of cell phone familiarity necessary, as well as to
interface that enables fast and accurate data collection. increase the speed of data entry. Participants enter in data as
The data collected during a patient visit will include both prompted by the cue card and then send the text message at
identification and health status information. The worker will the end of the interaction. The final part of the cue card as
enter in information to identify both the worker name (done displayed in Figure 1b shows a sample text message.
only once at the start of treatment, in the case of forms
and SMS) as well as the patient name. In addition, the D. Voice implementation
health worker will record the patient’s current temperature, For the voice interface the worker calls a live operator.
weight and pulse, as well as the presence or absence of The operator asks the worker a series of questions about the
seven symptoms: night sweats, chest pain, loss of appetite, patient’s health, which prompts the worker to ask the patient
nausea, coughing with blood, yellow eyes and fatigue. These that question. This means that workers interact simultaneously
symptoms were chosen based on advice gathered from tuber- with an operator and a patient; we are unaware of previous pro-
culosis health experts. The worker will also record whether the grams that have taken a similar approach. Figure 1c displays
patient’s current cough is absent, rare, mild, heavy or severe a sample interaction. The live operator confirms answers with
with blood. the worker; this adds to the length of each call but is done
The trial intervention is centered around the hypothesis that to increase accuracy. This can be particularly important when
better, more frequent data collected about patients can improve the phone connection is poor or there is background noise.
tuberculosis health outcomes and therefore high quality data
collection is critical. However, even if an interface encourages V. S TUDY M ETHODOLOGY
high quality data, it is still essential that such a data collection The user study took place in the Surat and Bharuch districts
method also be easy to use and affordable in order for such of the Indian state of Gujarat during July and August of 2008.
an intervention to have widespread applicability. Originally
we were planing to use electronic forms for data collection. A. Participants
However, since there appeared to be a dearth of literature As detailed in Table III, the study participants consisted of
in evaluating mobile data collection accuracy, we decided to six community health workers and seven hospital paramedical
evaluate the accuracy, speed and usability of three mobile staff. The community health workers were associated with the
phone interfaces. The results of this evaluation influenced our Dahej public health center; five of the paramedical staff were
choice of an interface for use in the treatment program. at the Reliance Tuberculosis hospital; and the remaining two
paramedical staff were at the dispensary of the Sardar Vallab-
B. Electronic forms implementation hbhai National Institute of Technology. The study participants
We created a Java application which provides a sequence of were recruited through contacts of the first author.
electronic forms that guide the worker to request information Initially, we had hoped to perform the study entirely with
from the patient. The worker identification number is encoded community health workers, as they are often the primary
once into the phone and is included with each recorded agents of remote data collection (including in our upcoming
visit. The worker has to either enter numeric data or make tuberculosis treatment program). However, this turned out to
a selection from a multiple-choice menu to encode symptoms. be infeasible because some community health workers were
80
unable to travel to the Dahej public health center for training six community health workers completed only one interaction,
and testing, and it was not feasible for us to travel to each while others completed two interactions (we did not anticipate
worker’s home. This prompted us to recruit participants from that voice would become a focal point of this study until
two other centers. There were also some logistical challenges halfway through our experiments).
in performing the studies due to adverse weather conditions The lag time between training and testing was exactly
and the bomb blasts occurring in July 2008 in the Surat area. one day for seven of the participants, and ranged between
The education level of the health workers ranged from 10 half a day and two days for the remaining participants. All
to 12 years, while the education of the hospital staff ranged participants received a brief refresher and supervised entry
from 10 years to a B.A. degree. The average age of the study session immediately prior to testing.
participants was 26.4 years (range 19-35). Seven participants
owned a cell phone, four participants had used but did not own VI. R ESULTS
a cell phone, and two participants had never used a cell phone The results of the user study are detailed in Table III. We
previously. Eleven of the participants were native Gujarati present both the accuracy of data entry, as well as the time
speakers and all spoke Hindi. needed to interview patients and report the data.
On average, electronic forms and SMS offered comparable
B. Training error rates of 4.2% and 4.5% per entry, respectively. The
Participants were trained by at least two trainers in small voice interface proved to be approximately 10x more accurate,
groups of at least two. Initially, examples were presented with an error rate of 0.45% per entry. While only one out of
on a whiteboard and participants were instructed to practice thirteen participants performed perfectly on both the forms and
entering in the data on either electronic forms or as an SMS SMS interfaces, twelve out of thirteen participants performed
using the cue card. After this stage, a paper with a set perfectly on voice. A Student’s two-tailed, unpaired t-test
of example patients was handed out, and participants were revealed that voice had a significantly lower error rate than
instructed to practice entering in this data. In the final stage, electronic forms (p < 0.01) and SMS (p < 0.01); no significant
participants were instructed to practice role playing patient– difference was found between the error rates of electronic
worker interactions with each other. forms and SMS (p = 0.84).
Participants received variable amounts of training, ranging It is important to note that our results indicate a bimodal
from 45 minutes to 8 hours, depending on their experience and distribution of error rates: participants 7-13 performed notably
availability. The longer training sessions were not necessarily better than participants 1-6. While there are many compound-
more effective, as they were performed in larger groups. While ing differences between these participants, including the man-
it would have been desirable to achieve more uniform training, ner in which we conducted training, we refer to them by their
this was difficult given the logistics of transportation and occupation in order to simplify the discussion; participants 1-6
worker schedules. Prior to the completion of training, all are health workers while participants 7-13 are hospital staff.
participants had completed at least two perfect interactions As summarized in Table III, health workers exhibited an error
on both electronic forms and SMS, and at least one perfect rate of 7.6% for forms and 6.1% for SMS, while hospital staff
interaction on the live operator mode. exhibited an error rate of 1.3% for forms and 3.2% for SMS.
Throughout the user study, we employed Motorola L6i In addition, the only voice error occurred with health workers.
cell phones for training and testing. This is the cheapest Unfortunately, our data are insufficient to explain the dif-
Java-enabled phone from Motorola (the source of our current ferences observed between these two groups of participants.
development tools) that is available in India; see Appendix On average, the hospital staff were older, more educated, and
A-1 for a cost analysis. All interfaces and related tools (cue more likely to own a cell phone than the health workers. It is
cards, etc.) were presented in Hindi, and the mobile phones plausible to suspect that these factors contributed to the higher
used had dual Hindi menus. accuracy achieved by hospital staff. However, due to logistical
reasons, our training procedure also differed between the two
C. Testing groups: health workers were trained in a large group for 6-8
Participants were tested in pairs, alternating who was being hours, while hospital staff were trained in small groups for
tested on data entry, and who was playing the fake patient for 1-2 hours. Our trainers were also somewhat more experienced
that data point. The order of the interfaces was randomized: when working with hospital staff, as health workers were
for a given participant pairing, the order of voice, SMS, and trained first. We re-iterate, however, that training continued
electronic forms was alternated. For the voice interface, the until all participants were able to complete two perfect trials
first author acted as the operator and was located outside of on forms and SMS, and one perfect trial on voice.
the room testing was being conducted in; however, there was To better understand the error rates observed using each
always an additional person associated with the experiment interface, we tabulate the exact sources of error in Appendix
inside the room at all times with the participants. A-2. We classify errors by their entry type (numeric, multiple-
During testing, each participant performed two complete choice, yes/no). We also inspect whether each error could be
patient–worker interactions (in the role of the worker) for each detected, by a trained eye, using the submitted data only;
of the forms and SMS interfaces. For the voice interface, the in the future, such errors could potentially be flagged or
81
TABLE III
R ESULTS OF THE USER STUDY. A LL PARTICIPANTS WERE EVALUATED ON TWO INTERACTIONS WITH THE FORMS INTERFACE AND TWO INTERACTIONS
WITH THE SMS INTERFACE . T HE COMMUNITY HEALTH WORKERS (1-6) WERE TESTED ON ONE INTERACTION WITH THE VOICE INTERFACE , WHILE THE
PARAMEDIC HOSPITAL STAFF (7-13) WERE TESTED ON TWO INTERACTIONS . AVERAGES AND STANDARD DEVIATIONS ARE SHOWN AT BOTTOM .
automatically fixed using self-correcting forms. Finally, we forms and SMS). One factor that contributed to the slower
tabulate whether each error is potentially dangerous (e.g., entry rates using voice was the cellular coverage in our study
a severe cough reported as a mild cough would prevent a area; the connection between participants and the operator was
physician from delivering needed care). highly unreliable. The audio quality was frequently degraded
Electronic forms witnessed errors in each entry type; only beyond recognition, and calls were occasionally dropped and
three of the twelve errors are evident from the values sub- re-started. While many resource-poor environments have ex-
mitted, while five errors may be dangerous. Surprisingly, cellent cellular coverage (including the area of Bihar that
eight of the errors were due to numeric entry problems on we are planning to target with our treatment program), the
the electronic forms. Two errors were due to a mis-placed weak coverage in our study area nonetheless reflects a realistic
decimal point in the temperature entry; while our interface hazard of voice in some environments.
automatically places the decimal point if needed, the user In addition to quantitative results, we also solicited qual-
failed to enter the right number of digits in the temperature. itative feedback from each participant, asking them to rank
The SMS interface also witnessed errors in each entry type; the interfaces by their order of personal preference. The
out of thirteen errors, eight are detectable and seven may be forms and SMS interfaces were most popular amongst the
serious. Three of the errors could perhaps be averted with a participants, with each receiving six votes as the most popular
revision of the SMS cue card: to indicate the absence of a interface. Only one participant preferred the voice interface
patient cough, many participants entered the code “0” rather to the others. This feedback is indicative of the poor phone
than the desired (though perhaps less intuitive) value of “1”. connections experienced during the trial; many found voice to
Unlike the forms interface, workers sometimes entered the be frustrating due to the bad call quality. We were surprised
wrong patient identity when using SMS. that any participants preferred the SMS interface, given the
The voice interface witnessed only a single error for the relatively cryptic message that is produced in the end; however,
entire duration of the trial. We consulted a videotaped record participants noted that fewer keys are required under SMS
of the interaction in question (we taped one interaction for each than under electronic forms (which requires scrolling and
participant), and found that the error was incurred by the op- selection). We also note that 8 of the 13 participants preferred
erator in translating the participant’s report into a spreadsheet. the interface on which they demonstrated the fastest entry time.
While such transcription errors could indeed occur in practice,
it is encouraging that the participants were not responsible for VII. D ISCUSSION
any errors on the voice interface. In addition to the factors examined in our experiment, cost
While the voice interface offered the lowest error rates, it is a critical variable for selecting a data collection interface.
also led to the longest entry times. Electronic forms and SMS For the purposes of our own decision making with regards to
averaged 1:39 and 1:37 per interaction, respectively, while the selecting an interface for our tuberculosis treatment program,
voice interface required 2:20 on average (1.43x higher than we performed a simple cost analysis. Details are provided in
82
Appendix A-1, but in summary, the expected cost for data worker to a live operator if the patient symptoms entered are
collection for each patient during his/her treatment is US worrisome. We look forward to exploring solutions for han-
$7.89 using electronic forms, US $4.59 using voice, and US dling these different tradeoffs, and considering IVR solutions,
$2.99 using SMS2 . These results show the cost of voice is as part of our future work.
competitive with the cost of the other two interfaces. Though
SMS is slightly cheaper, in order for tracking patient symptom VIII. C ONCLUSION
status to be helpful, it is essential that the reported data be Given the widespread excitement in using mobile phones
close to error-free. This data will be used to guide doctor for collecting and analyzing data in the developing world, it
intervention, and faulty data may lead to unnecessary visits is important to establish that the data entered on these devices
or worse, missed visits when a patient is sick. The voice meets the strict accuracy requirements of health, finance, and
interface had close to perfect accuracy and was significantly other applications. In this study, we provide a quantitative
more accurate than SMS or electronic forms. Voice also allows evaluation of data entry accuracy on mobile phones using
for additional, unscripted information to be easily collected, electronic forms, SMS, and voice interfaces in a resource-poor
and provides a social dimension to the health worker’s job. setting.
We anticipate that this social dimension could potentially lead Our results indicate that, within the context of our study, the
to higher performance and a lower turnover rate amongst error rates for electronic forms (4.2% of entries wrong) and
workers, since talking to an operator is likely to increase the SMS (4.5% of entries wrong) may be too high to deploy these
worker’s feeling of being supported and integrated in a larger solutions in a critical application. In contrast, the accuracy of
project. Voice also allows for verification to be performed the voice interface was an order of magnitude better (0.45%
easily: operators can simply request the worker to verify the of entries wrong), with only a single error observed across all
data entry just given, which can be particularly useful for trials. This result has influenced us to overhaul our plans for
unusual entries. In addition, a voice interface can be replicated an upcoming tuberculosis program in Bihar, India, to switch
very easily in other contexts– no special software or cue cards to a voice-only interface. Employing a voice interface requires
need to be developed, and any cell phones can be used. While the employment of an operator, and may not be cost-effective
voice requires longer entry times for workers, this represents a in all countries. However, in India, the cost of this operator
very small fraction of their overall working day. For all these is more than compensated by the lower cost of voice-only
reasons, we have now decided to use a voice interface for our handsets, voice-only cellular plans, decreased training time,
upcoming tuberculosis treatment program. and decreased literacy requirements on health workers.
Despite the many advantages of voice, there are still several While this study provides an initial data point for the
challenges that must be addressed in practice. In our upcoming accuracy of data collection on mobile phones, further research
treatment program, workers will be actively examining and is needed to distinguish the factors that are responsible for the
collecting data from patients and must report this information errors observed. In the case of electronic forms, we observed
back to an operator. Calling the operator and keeping him error rates that are 3-8x higher than previously measured
on the line as the worker examines the patient may lead to on PDAs. Our data are insufficient to diagnose whether this
a slightly awkward social interaction. Another more general difference is due to the devices themselves (screen resolution,
challenge for voice interfaces is how to handle scenarios in touch screen vs. keypad, etc.) or due to other aspects of the
which a user calls and the operator line is busy. One potential evaluation (worker education, training duration, etc.). A future
solution for these two challenges is to have the worker write study could address this question directly by evaluating PDAs
down the data on paper and then call the operator. This and mobile phones in the same focus group. However, it is
introduces an additional opportunity for transcription errors not our goal in this paper to prescribe the optimum device for
but has the side benefit of creating a paper trail that may mobile data collection. Rather, we aim only to highlight that
be used for later verification. To handle missing calls the there exists at least one context in which electronic forms and
operator could be responsible for calling back workers, or SMS may be too error-prone for large-scale deployment in
workers could leave a message that would be transcribed by an accuracy-critical application. In this same context, there
the operator. is evidence that a low-tech alternative (voice) provides an
An alternative solution to these challenges would be to use accurate and cost-effective solution.
an interactive voice recognition (IVR) system. IVR could also
IX. ACKNOWLEDGMENTS
be useful when there is very frequent data collection or when
each survey questionnaire is very long. Hybrid live-operator- We are very grateful to the many people who graciously
IVR systems are also possible, such as directing the worker facilitated and participated in our user study. We thank the
initially to an IVR system, but automatically transferring the hospital staff at the Sardar Vallabhbhai National Institute of
Technology in Surat, and also at the Reliance TB Hospital in
2 We use Motorola phones for the electronic forms due to our current set Hazira. We thank Joshnaben Godia of the Taluka Development
of development tools. Moving to the cheapest available Java-enabled phone office in Vagra, Gujarat, and especially Suprava Patnaik for all
would decrease the forms cost to $5.39. However, in practice the cost of voice
phones can also be reduced by leveraging existing phones in the community. her help with our study. This work was supported in part by
The cost of voice remains competitive with forms in most practical scenarios. the MIT Public Service Center.
83
R EFERENCES [22] Mali: Mobile service helps bring down infant mortality, 2007.
Balancing Act News, Issue No. 364, http://www.balancingact-
africa.com/news/back/balancing-act 364.html.
[1] Achieving the Health Millennium Development Goals in Asia and the [23] J. S. McBride, R. T. Anderson, and J. L. Bahnson. Using a hand-held
Pacific: Policies and Actions Within Health Systems and Beyond. United computer to collect data in an orthopedic outpatient clinic: A randomized
Nations Publications, Dec. 2007. trial of two survey methods. Medical Care, 37(7):647–51, July 1999.
[2] V. Anantraman, T. Mikkelsen, R. Khilnani, V. S. Kumar, N. R. Machi- [24] M. A. Missinou, C. H. O. Olola, S. Issifou, P.-B. Matsiegui, A. A.
raju, A. Pentland, and L. Ohno-Machado. Handheld computers for rural Adegnika, S. Borrmann, D. Wypij, T. E. Taylor, and P. G. Kremsner.
healthcare: Experiences from research concept to global operations. In Short report: Piloting paperless data entry for clinical research in Africa.
Workshop on Development by Design, 2002. Amer. Journ. of Trop. Medicine and Hygiene, 72(3):301–303, Mar. 2005.
[3] A. Bernabe-Ortiz, W. H. Curioso, M. A. Gonzales, W. Evangelista, J. M. [25] Mobilizing against malaria. Expanding Horizons (by Nokia), pages 12–
Castagnetto, C. P. Carcamo, J. P. Hughes, P. J. Garcia, G. P. Garnett, 13, Feb. 2007.
and K. K. Holmes. Handheld computers for self-administered sensitive [26] Nokia CEO message, 2008. http://www.nokia.com/A4942317.
data collection: A comparative study in Peru. BMC Medical Informatics [27] T. S. Parikh. Using mobile phones for secure, distributed document
and Decision Making, 8:11, 2008. processing in the developing world. Pervasive Computing, 4(2), 2005.
[4] J. Blaya and H. S. F. Fraser. Development, implementation and prelimi- [28] T. S. Parikh, P. Javid, S. K, K. Ghosh, and K. Toyama. Mobile
nary study of a PDA-based bacteriology collection system. Proceedings phones and paper documents: Evaluating a new approach for capturing
of the AMIA Annual Symposium, 2006:4145, 2006. microfinance data in rural India. In CHI, 2006.
[5] J. A. Blaya, W. Gomez, P. Rodriguez, and H. Fraser. Cost and imple- [29] P. Quinn, J. Goka, and H. Richardson. Assessment of an electronic daily
mentation analysis of a personal digital assistant system for laboratory diary in patients with overactive bladder. BJU International, 91(7):647–
data collection. The International Journal of Tuberculosis and Lung 52, May 2003.
Disease, 12(8):921–7, Aug. 2008. [30] J. Selanikio. Advantages of using handheld computers for
[6] Cost effectiveness study report for the PDA data capture and transmis- data collection. DataDyne. http://www.datadyne.org/files/Brief-
sion. SATELLIFE, Sept. 2004. http://www.healthnet.org/coststudy.php. HandheldAdvantages.pdf.
[7] W. H. Curioso, B. T. Karras, P. E. Campos, C. Buenda, K. K. Holmes, [31] J. Selanikio and R. Donna. Datadyne brief. DataDyne.
and A. M. Kimball. Design and implementation of Cell-PREVEN: A http://www.datadyne.org/files/DataDyne brief.pdf.
real-time surveillance system for adverse events using cell phones in [32] K. Shirima, O. Mukasa, J. Schellenberg, F. Manzi, D. John, A. Mushi,
Peru. Proceedings of the AMIA Annual Symposium, 2005:176180, 2005. M. Mrisho, M. Tanner, H. Mshinda, and D. Schellenberg. The use of
[8] B. DeRenzi, N. Lesh, T. Parikh, C. Sims, W. Maokla, M. Chemba, personal digital assistants for data entry at the point of collection in
Y. Hamisi, D. S. Hellenberg, M. Mitchell, and G. Borriello. E-IMCI: a large household survey in southern Tanzania. Emerging Themes in
Improving pediatric health care in low-income countries. In CHI, 2008. Epidemiology, 4(1):5, 2007.
[9] L. Diero, J. Rotich, J. Bii, B. Mamlin, R. Einterz, I. Kalamai, and [33] D. Skinner, U. Rivette, and C. Bloomberg. Evaluation of use of
W. Tierney. A computer-based medical record system and personal cellphones to aid compliance with drug therapy for HIV patients. AIDS
digital assistants to assess and follow patients with respiratory tract in- Care, 19(5):605–607, 2007.
fections visiting a rural Kenyan health centre. BMC Medical Informatics [34] SMS data management. Dimagi, Inc., 2008. http://www.dimagi.com/
and Decision Making, 6(1):21, 2006. content/sms-data-management.html.
[10] T. Engebretsen. Acceptance of information technology by health re- [35] B. Tiplady, G. Crompton, M. Dewar, F. Boellert, S. Matusiewicz,
search projects in low-income countries: Intention to use and acceptance L. Campbell, and D. Brackenridge. The use of electronic diaries in
of using EpiHandy (IUAUE). Master’s thesis, Agder University College, respiratory studies. Drug Information Journal, 31(3):759–764, 1997.
Dec. 2005. [36] I. Walker, C. Sigouin, J. Sek, T. Almonte, J. Carruthers, A. Chan,
[11] Evaluation of the SATELLIFE PDA project, Feb. 2003. Bridges.org. M. Pai, and N. Heddle. Comparing hand-held computers and paper
[12] D. Forster, R. H. Behrens, H. Campbell, and P. Byass. Evaluation of a diaries for haemophilia home therapy: A randomized trial. Haemophilia,
computerized field data collection system for health surveys. Bulletin 10(6):698–704, Nov. 2004.
of the World Health Organization, 69(1):107–11, 1991.
[13] R. Fynn. Remote HIV/AIDS patient monitoring tool using 3G/GPRS A PPENDIX A-1. C OST ANALYSIS
packet-switched mobile technology. In Appropriate Healthcare Tech-
nologies for Developing Countries, pages 129–138, 2006. In our basic cost analysis we first assume that the treatment
[14] S. Grisedale, M. Graves, and A. Grünsteidl. Designing a graphical user pool is 1000 patients. In our treatment program each worker is
interface for healthcare workers in rural India. In CHI, 1997.
[15] T. Groves. SatelLife: Getting relevant information to the developing
responsible for 10 patients, so there is a total of 100 workers.
world. British Medical Journal, 313(7072):1606–9, 1996. Each worker must be equipped with a cell phone. Our current
[16] P. C. Gupta. Survey of sociodemographic characteristics of tobacco use development tools for electronic forms are tied to Motorola,
among 99,598 individuals in Bombay, India using handheld computers. and require a Java-enabled phone. The cheapest such phone in
Tobacco Control, 5(2):114–120, June 1996.
[17] R. N. Jamison, R. H. Gracely, S. A. Raymond, J. G. Levine, B. Marino, India is the Motorola L6i which is 75 US dollars. In contrast,
T. J. Herrmann, M. Daly, D. Fram, and N. P. Katz. Comparative study both the SMS interface and voice interface can be used on
of electronic vs. paper VAS ratings: A randomized, crossover trial using any cell phone, one of the cheapest of which is the Motorola
healthy volunteers. Pain, 99(1-2):341–7, Sept. 2002.
[18] D. R. Kaufman and J. B. Starren. A methodological framework for Motofone F3 ($26). Voice calls are slightly more expensive
evaluating mobile health devices. Proceedings of the AMIA Annual than text messages: a 3-minute voice call, which is longer than
Symposium, 2006:978, 2006. the average time in our experimental study, is about 3 rupees
[19] H. Ladd and N. Sewankambo. Uganda health information network:
Technical report (submitted to the International Development Research
(0.065 US dollars, Airtel carrier). SMS messages using Airtel
Centre). SATELLIFE & Uganda Chartered HealthNet, Dec. 2004. are 1.5 rupees per message (0.0327 US dollars). The average
http://www.healthnet.org/idrcreport.html. call length in our user study is 2 minutes and 20 seconds;
[20] S. O. Lal, F. W. Smith, J. P. Davis, H. Y. Castro, D. W. Smith, D. L. therefore conducting 100 calls would require slightly under 4
Chinkes, and R. E. Barrow. Palm computer demonstrates a fast and
accurate means of burn data collection. The Journal of Burn Care & hours. We therefore anticipate that a 100 call load would be
Rehabilitation, 21(6):559–61; discussion 558, 2000. reasonable for an operator working 8–9 hours per day, in order
[21] S. J. Lane, N. M. Heddle, E. Arnold, and I. Walker. A review of to include a liberal number of breaks. Our program design
randomized controlled trials comparing the effectiveness of hand held
computers with paper methods for data collection. BMC Medical involves each worker visiting each patient to record symptom
Informatics and Decision Making, 6:23, 2006. information every two weeks. At this rate a single operator
84
sent user’s intuitive way of finding locations and directions to guide the person. This came as a surprise to us, as we
in developing regions. We conducted an additional survey were expecting this community to be more reliant on maps.
with the aim of understanding the technological requirements 10 out of 24 subjects usually are able to give exact direction
for F OLKSOMAPS. In this survey, we primarily focused on to a known landmark using relationships as shown in table
understanding the ways in which users express directions and II. 13 subjects mentioned that they guide the person to the
location information. In addition, to provide the subjects with nearest big landmark and then guide the person by phone or
an idea of the system, we created a paper prototype that shows pick the person from the landmark. 14 out of 24 subjects felt
how a user would typically access the services and populate confident that people should be able to follow the way they
content to it. guide. Table II shows the most commonly used relationships
The paper prototype was an audio recording of system to describe proximity of a landmark to another one. Typically,
prompts and user responses. We created audio prototypes for while giving directions, such relations are used to relate a
two scenarios 1) User calls F OLKSOMAPS to find directions to less visible or a less known landmark with a more visible or
a particular location 2) User calls to add information about a popular one.
location These prototypes were created using audio recording Interestingly, 21 out of 24 subjects either use both time
and editing tools and were used to illustrate the concept of and kilometers as a measure of distance and only 3 subjects
F OLKSOMAPS to some users. claimed to use only kilometers to represent distance. 17 out of
We conducted this survey along with the working model 24 participants never use zip codes while 6 use them rarely.
survey with a total of 40 subjects - 16 from Non-ITsavvy and This study, coupled with insights obtained from the survey
24 from ITsavvy community. Here is a sample of the questions of current models, has been useful to design the ontology as
we asked: well as the output of F OLKSOMAPS- differentiating it from
• How do you give directions to people on the road or traditional map-based systems prevalent today.
friends? IV. T HE T ECHNOLOGY
• How do you describe proximity of a landmark to another
one? Based upon insights gained from the surveys we devel-
• How do you describe distance? Kilometers or using time-
oped F OLKSOMAPS – a community generated map system.
to-travel? F OLKSOMAPS is designed to be populated by end users for
their own consumption. This section presents the architecture
A. Non-ITsavvy Community and design of F OLKSOMAPS system while highlighting design
When asked about how they personally give directions to choices that differentiate it from the established notion of map
anyone who asks them, 12 out of 16 subjects said they make systems. They key differences are listed below.
use of landmarks to explain the direction to the destination. • It relies primarily on user generated content rather than
They use names of big roads to describe a location, and use data populated by professionals.
“near to”, “adjacent to”, “opposite to” relations with respect • It strives for spatial integrity in the logical sense and
to visible or popular landmarks to point the destination. 5 does not consider spatial integrity in the physical sense
subjects said they can provide exact directions within one as essential. For instance, information such as “Building
kilometer of the destination. 4 said that they usually give A is located near to circle J after taking first turn on the
directions up to the nearest landmark thereafter which people circle while arriving from location B.” is treated complete
will need to ask again. 9 subjects felt confident about guiding and correct for tracing path from B to A. In other words,
a person to the exact landmark. the direction and distance parameters are not specified
Interestingly, 6 of 16 subjects said that they use time (only) in precise terms. This is because, as evident from the
as metric to measure the distance between any two locations. surveys, the end users are not likely to specify physical
12 out of 16 subjects mentioned that they use either time data while populating geographical landmarks.
or kilometers. 3 subjects mention that they sometimes use • A visual representation is not essential to F OLKSOMAPS
“rough” distance measures in terms of kilometers. One subject which is important considering the fact that a large
did not have the notion of kilometer as a measure at all. segment of users in developing countries do not have
TABLE II access to Internet.
C OMMON R ELATIONSHIPS USED TO EXPRESS RELATIVE LOCATION • F OLKSOMAPS is non-static and intelligent in the sense
that it infers new information from what is entered by
Relationships Non-ITsavvy ITsavvy
A isNear B 93.75% 67% the users.
A adjacent to B 81.25% 45.8% • The user input is not verified by the system and it
A opposite B 75% 75% is possible that pieces of incorrect information in the
knowledgebase may be present at different points of time.
F OLKSOMAPS adopts the Wiki model and allows all users
B. ITsavvy Community to add, edit and remove content freely. From the estab-
Only 2 out of 24 subjects tell people to use maps to guide lished Wikis on the Web we expect that the community
them to their house. The rest either use landmarks on roads would actively remove or edit invalid content and keep
88
the maps up-to-date. However, to limit malicious intent, to query the system either when they are stuck on road
the system places two minor restrictions described in the looking for directions or before starting on a trip and would
next section. be pressed for time. Further, the user set also consists of
people who might be illiterate or semi-literate or not very
A. Conceptual Design ITsavvy. Considering these, we discuss three different modes
We use the notion of a landmark as the basic unit of of interaction that the system needs to support to cater to
representing nodes in F OLKSOMAPS. A location represents different user segments for different tasks.
more coarse grained geographical area such as a village, city, There are three main tasks that a user can perform with
country etc., in addition to also representing a landmark. The the system. First is to find a landmark/location by specifying
core knowledgebase of the system needs to capture few key its name possibly including some related information such
logical characteristics of locations that users are interested in as nearby places or enclosing area. Second, users can ask
specifying and making use of. These include the following: for tracing a path between two locations. Third, users could
• Direction i.e. the positioning of a location relative to add to the knowledgebase by adding information about a
another one. From the surveys we found out that users location/landmark that they know of. In addition, some or all
are comfortable with providing relative information such users may also be given the facility to edit or remove entries
as ‘towards left of’, ‘on the right side of’ etc. instead of from the knowledgebase.
absolute direction in the form of north, east, west, south We also consider three sets of users who would interact with
compass points. the system. On one extreme, the users of F OLKSOMAPS are
• Distance i.e. the measure of amount of space between ITsavvy people who can access it over the Web. F OLKSOMAPS
two locations. This can be represented as numbers along provides a web interface to these users for submitting queries
with units in which the distance is expressed. From the as well as to update its knowledgebase by adding new locations
surveys, we learnt that F OLKSOMAPS needs to consider and related information. On the other extreme, we have users
both time and metric units to represent distance. who are illiterate or semi-literate and cannot afford to have
• Proximity and Reachability i.e. representation of infor- high end devices but can use an ordinary low end phone
mation stating that one location is in close proximity to for voice communication. Studies done earlier [3], [4], [5]
another or is reachable from another respectively. suggest that a voice-based interaction works well for this
• Layer i.e. granularity of geographic area that a location user segment and for them F OLKSOMAPS supports a voice
name represents. It could be a division as big as a whole based interface for querying the system. The third segment
country or as small as a village. The notion of direction of our users lies between the two extremes and consists of
and distance from a location, are interpreted with respect mobile people with low end devices who are familiar with
to the layer that the location represents. In other words, SMS. F OLKSOMAPS allows SMS based querying and location
direction and distance could be viewed as binary operator updates in a constrained form for these users.
over locations of the same level. For instance, ‘is towards
C. System Architecture
left of’ would be appropriate if the location pair being
considered is <Libya, Egypt> or say <South Korea, Figure 1 shows the architecture of the F OLKSOMAPS sys-
Japan> but not if the pair is <Sheraton waikiki hotel, tem. As shown, users can upload content into the knowl-
Mexico> where Sheraton Waikiki hotel is in Honolulu, edgebase through an SMS interface, a web based interface
Hawaii. or through a voice interface. Similarly, the content delivery to
We model the knowledgebase for representing and storing the consumers also happens through these multiple interfaces.
these concepts in two parts. The first one makes use of Web The knowledgebase consists of an ontology and a graph
Ontology Language (OWL) 6 to model the categorical charac- database. An ontology is used as the primary repository of
teristics of a landmark, i.e. direction, proximity, reachability the location information. This is because the user generated
and layer. Use of a Semantic Web language to represent content cannot be expected to be complete. It is essential to
relationships between locations brings in the advantage that be able to infer facts not explicitly populated by users in
the system can reason on those and infer newer relationships order to have a pragmatic map system. The graph portion
not explicitly specified by users of the system. The second part of the database captures additional information that either
makes use of a graph database to represent distance between cannot be expressed appropriately in the ontology or needs
landmarks which is numerical data. The two modules are used to be processed differently. This includes numeric data such
in conjunction to generate answers to queries submitted by as distances between locations.
users to the system. The central block of the figure forms the core of the
runtime system of F OLKSOMAPS, acting as an intermediary
B. User Interaction between the consumers of the service and the knowledgebase.
The user interaction aspect is critical for the success of It consists of a module each corresponding to the tasks listed
F OLKSOMAPS. This is especially true since users would tend above, namely location insertion/removal, location finding and
path finding. It provides a similar interface to the users across
6 http://www.w3.org/TR/owl-ref/ interaction modalities for information upload and retrieval.
89
Browser
Location Finding
Voice
Voice
Module
is defined as complementary and disjoint to Location. This be-
Ontology
Ontology
API
Path Finder
SMS Module came necessary since OWL ontologies follow the open world
SMS
G/W
Location
Insertion Module
assumption8 which means that a relation not explicitly asserted
Graph
in the ontology being reasoned upon cannot be concluded to
API
Graph
Web
App
Location DB
Removal Module be false since it may be specified elsewhere.
<owl:ObjectProperty rdf:about="&commgis;partOf">
Fig. 1. The System Architecture <rdf:type rdf:resource="&owl;TransitiveProperty" />
<rdfs:domain rdf:resource="&commgis;Location" />
<rdfs:range rdf:resource="&commgis;Location" />
</owl:ObjectProperty>
In the next section, we provide some details of the design <owl:Class rdf:about="&commgis;Landmark">
<rdfs:subClassOf rdf:resource="&commgis;Location"/>
of F OLKSOMAPS knowledgebase. <rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource="&commgis;partOf"/>
<owl:allValuesFrom rdf:resource="&commgis;SubArea" />
V. K NOWLEDGEBASE D ESIGN <owl:allValuesFrom rdf:resource="&commgis;Area" />
<owl:allValuesFrom rdf:resource="&commgis;Village" />
to search for or add/update locations specified in user’s new entry is rejected. Also, this landmark is added only
query. within the context of its parent, i.e. it’s SubArea. If the
The connectedTo relation is a symmetric as well as a SubArea supplied does not exist, then also the landmark
transitive relation. It expresses the fact that two locations is rejected. In addition to the landmark name, the user
are reachable from each other via one or more paths can optionally specify other meta information such as
that can obtained from the ontology facts. Users may which all landmarks are located nearby and to which all
add connectedTo relation between location instances that landmarks is this new one is connected to.
they are familiar with. F OLKSOMAPS makes use of the 5) editLandmark() : This method allows you to search for a
ontology to infer new connectedTo relations based upon landmark or a relation instance and allows you to delete
these individual assertions. The Path Finder module relies it in the fashion similar to addLandmark().
on this information to compute paths from the individual In the next section, we describe our prototype implementa-
connections supplied by users. tion that we used to conduct user studies.
• Bootstrap Process
The F OLKSOMAPS system could be bootstrapped from VI. P ROOF - OF -C ONCEPT I MPLEMENTATION
existing databases to populate instances of location types
in the upper part of the ontology. Two such sources We have implemented a prototype of F OLKSOMAPS and
of data in the absence of a full-fledged Geographical deployed it at our lab. The prototype’s knowledgebase includes
Information System (GIS) system come from the Telecom the ontology module and does not have the optional Graph
Industry and the Postal Department. While postal depart- module. We used OWL to implement the ontology. We im-
ment is obvious, the telecom companies also maintain plemented the F OLKSOMAPS modules for finding a location,
database of various circles that they operate in. Given that finding a path and adding a landmark. The ontology API
mobile phones have penetrated into the remote rural areas used is JENA9 with Pellet reasoner10. We bootstrapped the
as well, the upper ontology can be populated from their system with data about New Delhi, the capital of India and
data as well. While the actual GIS data benchmarking initialized it with SubAreas located in a couple of Areas under
require significant efforts and cost on the field to map the South Delhi district. We implemented a Web based interface
spatial data, the telecom and post office data give a very as well as a Voice based interface for this prototype. The Web
good logical view of the locations. This complements our based interface supports all the implemented modules and is
system’s design goal of providing a logical view rather developed using Java Server Pages (JSP)11 . On the other hand,
than a spatial view to the users. the Voice based interface is accessible over a phone call and
supports find location and find path modules. It is developed
B. Knowledgebase API using JSPs and VoiceXML12 .
This subsection describes the API that we built for accessing Users are allowed to populate the F OLKSOMAPS system
the knowledgebase for finding path, location or for adding a with new landmarks and associate them to the SubArea
landmark. which they belong to. Additionally, users can also provide
1) findLocation() : This method allows a user to search for information about the landmark. This includes other landmarks
a location specified by its name. Optionally, extra infor- located near to the one being added, and other landmarks that
mation can be supplied which includes the landmark’s are connected to this by road etc. Figure 4 shows a partial
relationship with another landmark or its attributes. snapshot of the populated Folksomaps knowledgebase.
2) findPath() : This method takes source location name India New Delhi
criteria and returns a list of locations that represent the South Delhi
(District)
Bootstrapped Data
Plaza
User Populated Landmarks
base.
4) addLandmark() : This method allows the user to insert a Fig. 4. A partial snapshot of populated ontology
new landmark into the knowledgebase. It takes the name 9 http://jena.sourceforge.net/
of the new landmark and also its immediate parent, i.e. 10 http://pellet.owldl.com
its SubArea name. Optionally, its next parent, i.e. PostOf- 11 https://java.sun.com/products/jsp
Users are also allowed to query F OLKSOMAPS for get- website. In addition to the bootstrapped data, the knowl-
ting location information and directions. Figure 5 shows the edgebase also contained some pre-populated locations. We
screenshot of webpage for getting directions from a source to requested them to populate content into F OLKSOMAPS while
destination location. Figure 6 shows the flowchart for querying restricting the locations to a set of 6 Areas in South Delhi
through the Voice based interface. Error steps are not shown District. This was done so that the content populated does
in this flowchart. The results from F OLKSOMAPS reflect the not get thinly spread out and is relatively rich for querying
14
ways in which people would essentially give directions. . After populating some landmarks known to them, users
then queried the system for finding information about other
landmarks and travel directions to those. After this, we asked
them a few questions about their view of F OLKSOMAPS
system.
For the Non-ITsavvy subjects, we briefed them about the
purpose of the proposed system and gave an explanation of
the prototype. We then walked them through the voice based
interface, by querying for some location and requesting for
a path to that location from another. This was followed by a
question answering session. All subjects grasped the concept
fairly quickly and were able to see the benefits they could
derive from such a system.
B. Results of Survey for Non-ITsavvy subjects
We conducted a total of 22 surveys with Non-ITsavvy
participants using the voice interface. The set of interviewed
Fig. 5. Web UI showing the results of querying for directions
people consisted of porters, security guards, elderly people,
draughtsmen, waiters and service staff. The results are tabu-
Welcome to
lated below.
Folksomaps
Play
Get Get
Location
Src_Name Loc_Name
Details
Fig. 6. Voice UI flow for querying F OLKSOMAPS. Red dots indicate voice
recognition steps.
than asking someone on the street. Apart from convenience, cities, more than the road route from a source to destination,
this overwhelming response can be attributed to a couple of people from the underprivileged segment were more interested
reasons. First, asking on the street does not always work. On a in knowing the bus route numbers that could take them to
secluded road, there may not be anyone to ask. Several times their destination. Road routes serve well those people who
people end up giving wrong directions (possibly to hide their travel by their own vehicles but the underprivileged rely
lack of awareness or to not appear rude) leading to precious primarily on public transportation. Similarly, for rural areas
time spent traveling on a wrong route and recovering from that consist of remotely located towns and villages, what helps
it. Also, the point of view of street vendors who often get the underprivileged people is information regarding modes of
enquiries from passersby about directions is quite interesting. transportation (train, bus, boat, cycle-rickshaw, taxi etc.) to
They candidly admitted that during the course of the day take from source to destination, where to make a switch and
they get so many such queries for detailed travel instructions estimated travel time. Time tables of these public transport
that it is easy to get irritated and shrug them off. Second, an mechanisms are another important feature for this segment
important insight provided to us by our subjects was the sense that can become an essential part of F OLKSOMAPS.
of security that they would get with such a system. We were Key Insights : We realized that accuracy of the information
informed that even though asking for travel directions from is a key requirement and more the meta information available,
strangers on the street is an option, it exposes the enquirer merrier it would be for these consumers. Also, voice based
to criminal elements, who often take advantage of their lack interface is indeed a preferred mode for this user segment
of knowledge and use it to rob them, the activity sometimes over SMS and the calls to the system should to be free of cost.
resulting into a worse situation such as a murder. This is
C. Results of Survey of ITsavvy subjects
especially true for first time visitors to the city from remote
towns and villages or female citizens traveling at odd hours For ITsavvy segment, we conducted the survey with a total
of the day. of 15 subjects using the web based interface. We also told
Many people were willing to pay for the call even a them that the system has a voice based interface available
small premium over normal charges as they saw value to over a phone call and supporting similar API. The user list
having this information available to them at all times. Yet consisted primarily of software professionals apart from a
a majority of the subjects preferred the advertisement model couple of businessmen. As expected, the ITsavvy community
where an advertisement played in the beginning of the call had significant experience in using the current online maps of
pays for the entire call. This is understandable, given this cities in Indian metropolitan cities and were able to carefully
segment’s high sensitivity to cost. A few users suggested that evaluate our approach, considering the map services that are
the advertisement model was better since most of these users already operating in metropolitan cities. We try to capture
primarily have pre-paid SIM cards and often they do not have learnings from their feedback.
sufficient balance to make outgoing calls. TABLE IV
Most people were willing to upload content, though a few S UMMARY OF IT SAVVY U SER R ESPONSES TO F OLKSOMAPS
refused as they were hesitant due to not owning a phone. Questions Yes No
Almost everyone preferred the voice based interface over Would you access it for directions? 93% 7%
SMS even though we demonstrated speech recognition errors Prefer F OLKSOMAPS over asking people? 87% 13%
Ready to pay for call (else want ads)? 67% 33%
during the study interviews. The primary reason for this cited
Will upload content? 87% 13%
by them was that many people are either not comfortable using Prefer Web for upload? 92% 8%
SMS or not comfortable using a mobile phone itself. However, Results need to be very accurate? 53% 47%
some users who were well versed with SMS preferred it over
voice. As is evident from the results (Table IV), most survey
In terms of accuracy of returned results, most people asked participants mentioned that they would like to use this service
for full accuracy while a very few were okay with minor and that it would certainly be more convenient than asking
mistakes. The need for strong accuracy is driven by the fact people around in the streets. Interestingly, a bulk of the
that most of these people either use public transport, or use a ITsavvy community did not stress on getting fine-grained
bicycle or even walk to reach their destination. The cost of a direction all the time. They were fine with getting high level
wrong input for them is huge compared to a person driving directions involving major landmarks.
in his own or rented vehicle. In fact, one of the main reasons Most people were fine with paying for the service when
for preferring a voice call over asking people for directions offered on phone. Most were also willing to upload content
was to avoid wrong directions. This is an important feedback into F OLKSOMAPS but preferred to do so over a web based
since we started with the assumption that we do not need strict interface as opposed to SMS or a voice based interface.
controls over the content and the wiki model would work. But Accuracy of responses was important for this segment as
the tolerance for incorrect information is low and we need to well, though not as strongly as for the Non-ITsavvy segment.
factor this in. As discussed earlier, this segment typically used their own
We also learnt that meta information is as important to Non- vehicles and need high level directions rather than precise
ITsavvy users as the landmarks themselves. For instance, in route.
93
A few participants pointed out that while voice-based access destination. Voice interface also takes care of the language
is good for interactive session, they would still prefer the barrier since content can be delivered in local language as
content is sent to them via SMS so that they can store it demonstrated in other systems [3]. Voice based interfaces are,
for future access, pointing out that it is easy to forget the however, constrained with the capability of speech recognition
directions if you just hear it. technology which is under slow but constant improvement.
However, this segment had other expectations from the sys- Our surveys indicate that most people would like to con-
tem. Some subjects mentioned that the system should adapt to tribute to F OLKSOMAPS knowledge base. However, in real-
the user’s request and have the ability to produce fine-grained life, there are several factors that provide impedence for a
details depending on whether the destination is a popular user to be an effective information producer. Reasons range
landmark, or an area or a house in a colony. Few others men- from users becoming busy, loosing interest after an initial
tioned that the directions provided by F OLKSOMAPS should surge, etc. In a live deployment of F OLKSOMAPS, one needs
take into consideration the amount of knowledge the subject to also consider pragmatic business models (such as bartering
already has about the area, i.e. it should be personalized based models, advertisements or incentives) using which an in-flow
upon user profile. One subject mentioned that the current of information can be sustained to keep improving the quality
prototype appeared more suitable for driving directions but of system responses.
not for pedestrians. Accuracy of results being of primary concern universally,
Just like the Non-ITsavvy community, the ITsavvy com- further research is needed to ensure that various modules of
munity also reflects the need for meta information on such a F OLKSOMAPS would guarantee correct and precise results
community-driven map. One subject mentioned that he would given that the data input by the users is correct in the first
be interested in uploading traffic information on routes and place. Also, as the knowledgebase as well as the userbase
would benefit from community uploading such information on grows, established scale-up techniques would have to be
the system. A few subjects mentioned that frequent changes applied for real life environments.
in road plans due to constructions should be captured by
such a system - thus making it more usable than just getting IX. R ELATED W ORK
directions. Two research areas that are very relevant for F OLKSOMAPS
Key Insights : While accuracy and convenience score with are the areas of research in use of IT for underprivileged
IT-Savvy population as well, this segment turned out to be in developing regions and semantic tools for geographic in-
more ambitious in terms of deriving benefits from such a formation systems. Apart from these, work in the area of
system. Based upon the feedback listed above, we learnt that intelligent user interfaces for masses in developing regions is
the user interfaces of F OLKSOMAPS needs to be rich and also relevant.
adaptive to the information needs of the user when considering There is a lot of literature on means to harness available
this community. It also appears to the authors that dynamic information and user generated content [11], [12], [13] to
and real-time information augmented with traditional services deliver useful services to underprivileged in developing re-
like finding directions and locations would certainly add value gions. [14], [15] talks about voice-driven technologies (e.g.
to F OLKSOMAPS. audio wiki) to capture user content from Non-ITsavvy masses.
The Neighbourhood Mapping [16]15 initiative proposes the
VIII. D ISCUSSION
involvement of school students to gather community input
F OLKSOMAPS we believe, presents a novel approach to- in the context of building maps. The project used PDAs
wards developing a self-sustaining map system, harnessing coupled with GPS to build an information repository that
community input, particularly targeted towards developing could be used of planning purposes. F OLKSOMAPS builds
countries, where there is a need for such a system. further along this direction and proposes to create alternatives
Reasoners using ontologies consume space and compute to well established solutions in developed countries, solely
power. [6], [7], [8] reports ways through which spatial reason- through user generated content. Specifically, it focuses on
ing can be made faster. This is an issue with F OLKSOMAPS creating a framework (exploiting ontological reasoning), where
as well. However, F OLKSOMAPS as a system compliments by geographic information can be captured, enriched, and
this body of work as its focus is not on improving reasoning funneled back to the masses - customized to the needs of
capabilities or address scalability needs of underlying ontology developing regions. To the best of our knowledge, we are
reasoners. Rather F OLKSOMAPS can benefit from this body not aware of any such community-driven map system for
of work by adopting the solutions suggested to improve the developing regions.
computational and reasoning efficiency. There has been considerable work on Place ontologies,
Given their preference towards voice based interface over retrieval and storage of geographical information using on-
SMS [9], [10], designing an efficient and user-friendly voice- tologies [8], [17], [18], [6], [7]. For example, [8] talks about
based user interface for the masses is important for F OLK - limitations of OWL to support spatial reasoning, integrity
SOMAPS . For example, while finding directions, user interface rules, and proposes a combination of spatial data-based store
should be designed in a way that facilitates users to specify the
level of detail they are looking for, varying from source to the 15 http://www.csdms.in/NM/
94
and ontology-based reasoning to better represent geographic [3] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A. Nanavati,
information and resources. [6] further critically evaluates “WWTW: A World Wide Telecom Web,” in ACM SIGCOMM Workshop
on Networked Systems For Developing Regions, Aug 2007.
ontology-based approaches towards geographic information [4] M. Plauche and M. Prabaker, “Tamil Market: A Spoken Dialog System
retrieval while [7] presents a spatially aware search engine, for for Rural India,” in Working Papers in Computer-Human Interfaces
semantic interoperability of distributed and heterogeneous GIS (CHI), 2006.
[5] T. S. Parikh, “Mobile Phones may be the Right Devices for Supporting
on the Internet. In contrast to this literature, F OLKSOMAPS Developing World Accessibility, but is the WWW the Right Service
does not focus on improving reasoning capabilities of OWL or Delivery Model?” in International Cross-Disciplinary Workshop on Web
address spatial and logical integrity issues of Place ontologies. Accessibility (W4A), Scotland, May 2006.
[6] A. Abdelmoty, P. Smart, C. Jones, G. Fu, and D. Finch, “A critical
Rather, we focus on designing concepts of an ontology that evaluation of ontology languages for geographic information retrieval
is suitable for capturing map information from communities, on the Internet,” Journal of Visual Languages and Computing., pp. 331–
keeping only logical integrity in mind, and by exploiting 358, August 2005.
[7] M. Kun and B. Fuling, “An Ontology-Based Approach for Geographic
currently standardized semantics of OWL. F OLKSOMAPS in Information Retrieval on the Web,” in International Conference on
a way is hence complimentary to this body of work and can Wireless Communications, Networking and Mobile Computing, Sept
benefit from it. 2007, pp. 5959–5962.
[8] A. Abdelmoty, P. Smart, and C. Jones, “Building Place Ontologies
A body of research focuses on intelligent user interfaces for the Semantic Web: Issues and Approaches,” in Proceedings of the
for developing regions [19], [10], [20] and infrastructures to 4th ACM workshop on Geographical Information Retrieval, Lisbon,
take IT services to the economically challenged and Non- Portugal, Nov 2007.
[9] “Africa:Impact of Mobile Phones,” Vodafone Policy Paper Series, Mar
ITsavvy masses in developing regions [3], [21]. F OLKSOMAPS 2005.
at its core can certainly benefit from user interface designs [10] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A. Nanavati,
to facilitate upload and download of map information. The “Voiserv: Creation and delivery of converged services through voice for
emerging economies,” in In Proceedings of the WoWMoM, Finland, June
system, by having multi-modal front-ends fits well with the 2007.
architectural principles outlined in [3], [21]. [11] R. Abraham, “Mobile Phones and Economic Development: Evidence
from the Fishing Industry in India,” in IEEE/ACM International Confer-
ence on Information and Communication Technologies and Development
X. C ONCLUSION (ICTD), Berkeley, USA, May 2006.
[12] K. Ramamritham, A. Bahuman, C. B. S. Duttagupta, and S. Balasun-
In this paper, we investigated feasibility of a community- daram, “Innovative ICT Tools for Information Provision in Agricultural
driven approach towards creating maps for developing re- Extension,” in IEEE/ACM International Conference on Information and
gions. Our system dubbed F OLKSOMAPS, has the potential of Communication Technologies and Development (ICTD), Dec 2006.
[13] P. K. Reddy, G. Ramaraju, and G. Reddy, “eSaguTM: A Data Ware-
providing an effective alternative to expensive map solutions house Enabled Personalized Agricultural Advisory System,” in ACM
using community input, making map-based services (finding International Conference on Management of Data, China, June 2007.
directions, finding locations and landmarks) available to peo- [14] P. Kotkar, W. Thies, and S. Amarasinghe, “An Audio Wiki for Pub-
lishing User-Generated Content in the Developing World,” in HCI for
ple in developing regions where such services are currently Community and International Development (Workshop at CHI 2008),
missing. F OLKSOMAPS builds on the current models adopted Florence, Italy, April 2008.
by users in developing regions and leverages their collective [15] L. Wang, P. Roe, and B. Pham, “An Audio Wiki Supporting Mobile
Collaboration,” in Proceedings of the ACM Symposium on Applied
knowledgebase thus overcoming the huge cost barrier in Computing (SAC), Brazil, March 2008.
developing such a system. It is specifically designed to provide [16] R. Mallick, H. Kalra, and D. Banerjee, “Infusing map
content that is intuitive for the users. culture through participatory mapping,” GIS@development.
http://www.gisdevelopment.net/magazine/years/2005/feb/infusing.htm,
We conducted a total of about 77 interviews in the process Feb 2005.
of evaluating a need for such a system and testing our proto- [17] J. Kay, W. Niu, and D. Carmichael, “ONCOR: Ontology and Evidence
type for verifying usability and utility of F OLKSOMAPS. Our based Context Reasoner,” in Proceedings of International Conference
on Intelligent User Interfaces (IUI), Honolulu, Hawaii, Jan 2007.
surveys suggest that the community is very receptive towards [18] G. Look and H. Shrobe, “Towards Intelligent Mapping Applications:
the concept of a community-driven map as that alleviates A Study Of Elements Found In Cognitive Maps,” in Proceedings of
some of the problems (reliance on people, security, inaccurate International Conference on Intelligent User Interfaces (IUI), Honolulu,
Hawaii, Jan 2007.
directions, etc) they face in day-to-day life. We intend to build [19] I. Medhi, A. Sagar, and K. Toyama, “Text-Free User Interfaces for
further upon this system and overcome its current limitations Illiterate and Semi-Literate Users,” in IEEE/ACM International Confer-
to bring it even closer to users’ expectation. ence on Information and Communication Technologies and Development
(ICTD), Berkeley, USA, May 2006.
[20] J. Sherwani, S. Tomko, and R. Rosenfeld, “Sublime: A Speech- and
ACKNOWLEDGMENT Language-based Information Management Environment,” in In Proc.
ICASSP, May 2006.
The authors would like to thank Amit Nanavati for suggest- [21] A. Kumar, N. Rajput, S. Agarwal, D. Chakraborty, and A. A. Nanavati,
ing the ’Folksomaps’ name. “Organizing the Unorganized - Employing IT to Empower the Under-
privileged,” in Proceedings of 17th ACM International Conference on
R EFERENCES World Wide Web(WWW), Beijing, China, April 2008.
A. Background
* Home Care can only be reached after the user has heard the Danger Signs of
the selected illness. For this reason, Task 2 is slightly more difficult than Task
1.
A. User Profile
During our user study, we had 33 caregivers who
participated, of which 27 tried both DTMF and ASR systems.
The caregivers were all female with the exception of one male
caregiver. The age of our users ranged between 22 and 61
years old with the average age of 34 years. The average
number of years of schooling amongst our users was 9 years
but 2 users had 0 years of schooling. All of the users could
read and write the local language Setswana and approximately
Fig. 8. Time since last Health Question by caregivers.
79% of them knew some English. In terms of occupation, 47%
were unemployed and of the 53% who were employed, the
majority were in low-income occupations such as cook, TABLE II
cleaner, house maid, hair dresser, or security guard. SUMMARY OF TECHNOLOGY OWNERSHIP AND USAGE BY
Caregivers reported that they visit Baylor between 1-3 times a CAREGIVERS.
month, with average travelling distance and time at 28 km and Technology Use Ownership
1 hour respectively, some travelling from as far as 130 km,
with average cost of travel at 18 Pulas (approx. $3 USD). The Mobile Phone 91% 91%
average waiting time at Baylor reported by the caregivers was
2.5 hrs, also travel time to Baylor ranges from 30 minutes to 3 Landline 30% 9%
hours; together, these represent a significant portion of a Computer 15% 3%
working day and are a substantial burden to the caregivers.
Caregivers were also asked what they usually do if they TV 76% 71%
have questions regarding the child’s health and when the last Radio 91% 91%
time was that they had such a question. One third of the
caregivers usually go the local clinic to resolve their health DVD/Video Machine 41% 38%
queries and another quarter go to Baylor for this purpose (Fig.
7). Forty percent of the caregivers had a query regarding the ATM 35% N/A
child’s health within the last 6 months, and another 21% had
more recent queries; in contrast, only 12 % could not
remember specific queries (Fig. 8).
102
B. Broad Usability Metrics and Observations
Fig. 11. Routing Time for completed tasks. Recall from Task Completion
Rates (above) that the Level 1 mean response is averaged over approximately Fig. 12. Usability Metrics for DTMF and ASR systems.
75% of our users, whereas Level 2 and Level 3 times are averaged over fewer
users, in some cases only one or two.
5) User Preference
Only two users correctly completed Task 2 with the ASR Systems were rated separately using the PARADISE [26]
system; both users took over 4 minutes to do so. With the framework after use. The Likert scores were found to be
DTMF system only one person completed Task 2, doing so in unreliable, however. Despite efforts to elicit honest, critical
less than 3 minutes. All users that correctly completed Task 1 feedback to the system (for example, we had a different
using the ASR system did so in about 60 seconds. For Task 1 person in a separate room conduct the post-study evaluations),
on the DTMF system, most users completed the task in just all caregivers gave the system the highest marks possible
over 60 seconds, but one user took over 6 minutes. across all categories and were hesitant to provide any
criticism. For those 27 (out of 33 total) caregivers who tried
both the DTMF and the ASR system, we were able to elicit
4) Other Usability Metrics feelings of preference for one system over the other (Table
III). Most caregivers (59%) preferred the DTMF system over
Results from our user study show that user interaction with the ASR one (19%) and 22% indicated no preference. Both
the ASR and DTMF system are very similar across a wide the DTMF and the ASR systems were judged to be the faster
range of additional usability metrics, illustrated in Fig. 12. Use system by those who preferred it. Our measurements of
of the Main Menu global (press '0' or say 'Simolola') and the routing time and task completion, however, show the systems
Exit global (press '9' or say "Fetsa"), were similar for both are comparable.
systems. There were almost the same average number of TABLE III
timeouts (when 4 seconds elapsed with no user response) and USER PREFERENCE FOR DTMF OR SPEECH INPUT
repeats (when user chooses to repeat an information node) for
the ASR and the DTMF system. On average, caregivers used Preferred Num Reasons Given Example Remark
the barge-in function one more time when using the ASR System ber
system but the total number of turns taken by users was
similar for both systems. Note, all the usability metrics are DTMF 16 Clearer "Its quick and the
means for each call, whether the task was completed or not. instructions (7) doctor gives you the
Faster to use (4) instructions, you
More private (2) just have to follow
them."
things." would not help the caregiver any further. In some cases, they
would ask to try again, which we allowed if time permitted.
6) Social Factor Correlations Most were very interested in the content and many referred to
the voice they heard as ‘the doctor’. Only one caregiver
We examined several social factors based on our users' recognized the voice of the SDS as the celebrity soap star
responses to questionnaires to see if they correlated with although most commented that “the ‘doctor’ explained very
either their interaction with the system, based on objective nicely”.
metrics, or their reported system preference (DTMF vs ASR).
During the interviews, all of our users enthusiastically
Employment and experience loading airtime on a mobile indicated that they would like to use the service again; many
phone were significant factors in the overall task completion said that it would be very valuable for educating themselves
rate (p=0.09 and 0.02, respectively). Those users who were and their family/friends on caregiving aspects for children
employed or who had experience loading airtime completed with HIV. A SDS such as OpenPhone could also serve as
more tasks during the study. Age, education, use of landlines, persuasion tool for caregivers trying to educate others, as
mobile phones, or ATM machines were not significant factors. explicitly reaffirmed by a caregiver “now I can tell them at
Only previous use of a landline was a significant factor in use home that the doctor (SDS voice) says the same thing
of system globals (p=0.01). Use of landlines was also a factor, (referring to a HIV related topic) that I’m telling them”.
along with loading airtime, in whether a user barged-in during
system use (p=0.1 and 0.1 respectively). Caregivers who use a
landline and load airtime were more likely to barge-in during IV. DISCUSSION
system use. No questionnaire responses were found to be
significant factors for system repeats or overall response time.
From our pilot study, we found that there were no
Employment, previous use of an ATM, and experience significant differences between task completion rates (ASR
loading airtime on a mobile phone were significant factors in only performed slightly better) or other usability metrics for
the overall correct response (p= 0.07, 0.09, and 0.03) 1 . both systems. This agrees with a number of previous studies
Caregivers who were employed, used ATM machines and in the developed world [12, 28, 29] where no major
loaded airtime had more correct responses to tasks during the differences were found in terms of performance. However,
user study. Age, education, and amount of mobile phone use subjectively the majority of our users preferred DTMF (59%)
were not factors in correct response. over ASR (19%), which is in contrast to formal studies in the
developed world [9]-[11] (where user preferences largely
Loading airtime was the sole significant factor in user
favour speech), but correlates with the observation that simple
preference of DTMF over the ASR system (p=0.1). Those
key-press replacement with keywords is generally not viewed
people who load airtime regularly preferred DTMF over ASR.
favourably. The users who did prefer ASR did not as in
Employment significantly correlated (p<0.1) with Overall
developed world studies comment on the aesthetics of speech
Task Completion and Overall Correct Response. Age and
input, that “speech is more entertaining or enjoyable” but
monthly mobile phone costs were not found to correlate
rather on the utility of speech “more accessible for older
significantly with user performance or user preference for
people or faster”.
either the DTMF or ASR system.
7) Other Observations Our finding that users’ employment and experience loading
airtime correlated with higher task completion but that
We observed based on body language and explicit remarks education level does not, indicates that technological literacy
that several caregivers were nervous at first, and then became is a more important factor in adopting new technology than
more relaxed during the first few minutes of the study. Many literacy itself. This may also contribute to the finding that
caregivers (and their children) showed signs of fatigue while ‘loading of airtime’ was the sole significant factor found in
trying the second system (sometimes an hour later). DTMF preference over ASR.
Caregivers sometimes had trouble understanding the task.
It is also interesting to note that our users who had minimal
They would often try to find information about ‘Nutrition’, for
exposure to SDSs (except loading airtime), were relatively
example, which was the topic of the demonstration, instead of
comfortable using our system for the first time, as indicated by
searching for information on their assigned task. Caregivers
their frequent, timely barge-ins. Also, our users noticed the
also often clarified with the interpreter what the task was,
value of speech (allowing hands-free operation,
what the keywords were, and what they should press or say
innumerate/older people being able to use it) and DTMF as
during their trial. The interpreter would nod or say 'yes' but
well (provides privacy). This highlights that even if users may
be technically inexperienced and unfamiliar with an ICT
1
The following categorical groups (yes and no) were divided into two application, they have valuable and sound judgment on the
groups and the resulting dependent values analysed: Employment, previous utility of the interface. Also, both DTMF and the ASR
use of an ATM, experience loading airtime, use of landlines, use of ATM
machines. A p-value of 0.1 and less indicates a significant difference between systems were judged (subjectively) to be the faster system by
the two groups, whereas a p-value higher than 0.1 indicate no significant those who preferred it, which indicates that time constraints,
difference. may also be of importance for low literacy users, in contrast to
105
earlier observations. The DTMF vs. speech input comparison interactive (perhaps with audio cues, getting intermittent user
could be improved in terms determining task performance by feedback) and conversational.
using a between-subjects experiment design; however this In addition, the nature of educational information services
approach would not reveal user preference. tends to be rather exploratory; where a user may peruse
various topics related to his/her general query (e.g. a user may
Our preliminary investigations indicated that a good fraction
want to, in general know about “Dealing with body fluids and
of caregivers were in the low and non–literate range. In our
infected waste” which has 4 topics all related to that option).
user study though, we encountered that many of our recruits
This in our case translated to some users struggling to find the
were semi- to low literate. Also, whilst our user numbers
exact menu option related to the task and exploring related
represent a significant sample size for a developing world user
sub-topics. One user even singled out that she would like to
study, it may in comparison to studies in the developed world
know the mapping of the menu options beforehand to help her
be on the ‘small’ end of the spectrum. The above-mentioned
locate the information she is looking for. This experience
issues emphasize some of the challenges faced in research for
highlighted challenges of using hierarchical menus [8] and the
the developing world; that the very users for whom an SDS
importance of paying attention to the taxonomy and
could be most useful may be the hardest for us to reach and
vocabulary of the system to enable easy navigation for the
also that user recruitment in developing regions can present
user.
significant obstacles [30].
SDS design for smaller languages also introduces challenges
Whilst this pilot study illustrated that telephony services
on other dimensions including prompting and persona. The
could in fact be easily used by semi and low literacy users,
prompts and content of a SDS application will typically be
and that an SDS in local language can be a powerful health
translated from a language such as English to the local
education tool, the decisive factor in widespread uptake is
language. Thus, great care has to taken in the prompt writing
likely to be the cost incurred by the caller for the service. The
phase to ensure that intended meaning of the original prompt
majority of caregivers said that even though the service would
(English) is still preserved in the translated prompt (local
be useful to them they would only be able to make use of it, if
language) and conveyed in the simplest and shortest way
the service is toll-free. An average phone call in Botswana of
possible. Often, a concept described by a single word in
5-10 minutes to the SDS would cost a mobile phone user $1-2
English has no direct translation in another language. For
USD. From our questionnaires, the average cost per month for
instance, whereas a keyword in the English version of our
mobile phone usage was $10.5 USD. Thus, a single phone call
application was “Safe food”, it became the phrase, “Dijo tse di
to the SDS would consume, 10-20% of a caregiver’s monthly
siameng” after translation, in order to adequately describe the
mobile phone budget, making the case for a toll-free number
concept.
all the more imperative.
Moreover, not only should the translated SDS prompts
Notwithstanding our findings that show promise for SDSs
convey the intended meaning but the designer should ensure
for low literacy users, we did encounter challenges in
that the persona of the local language system is in line with
introducing the concept of a SDS to our users, for instance a
cultural and contextual expectations of the intended audience.
few users did not fully realise that the system was automated.
For instance we ensured our prompts not only had the right
For example, at the end of one call, a user proceeded to ask
balance of formality and gravity appropriate for the message
the ‘nurse’ (system persona) a question when prompted by the
(HIV info) but were also understandable and conversational.
system to leave a comment (and waited for the answer).
Another user repeatedly acknowledged what the ‘nurse’ was
Our experience in employing multiple data-gathering
saying by responding with “Yes, yes” or “I agree with you”.
techniques in the needs investigation phase (interviews and
An obvious solution here might be to use a text-to-speech
discussions, observations, field visits, and focus groups) better
(TTS) voice for the prompts. However, we run the risk of a
equipped us in trying to comprehend the needs of our users.
mismatch between the target audience and the context (e.g.
The interviews and discussions helped us establish a rapport
emotionally sensitive in terms of health care) or culturally-
with our users and stakeholders (Baylor staff), whilst also
based communication norms of the community. This in turn
providing the flexibility of follow-up questioning and
may affect the willingness of users to interact with the system.
recalibration of interviewer terminology when needed.
Observations on the other hand helped to further reveal the
A general design challenge for SDSs is to ensure a minimal
issues that users are unable or unwilling to articulate or
cognitive load on the user. This magnifies more so in the case
express. Field work allowed us to pick up on the cultural
of information access applications where lengthy pieces of
nuances and social-economic context of the users and enabled
information need to be provided (e.g. in a health care context).
us to gain a deeper understanding of the sensitive environment
In our case, all the users indicated in the subjective
(HIV/health) we were working in. Finally, through focus
questionnaire that the length of the content was not too long;
groups we were able to observe the interaction amongst
however, during the experiment some users did mention the
caregivers and most importantly it enabled us to obtain
need to concentrate in order “not to miss all the things being
specific targeted design information in a group setting and
said by the ‘nurse’ ”. There is a need to address the above-
correct our earlier assumptions on the ranking of SDS health
mentioned issues by developing the dialog to be more
topics.
106
REFERENCES
[1] R. Tucker and K. Shalonova. "The Local Language Speech Technology
V. CONCLUSION Initiative", SCALLA Conference, Nepal, 2004.
[2] SIL International. (2001). Facts about illiteracy. [Online] Available at::
http://www.sil.org/literacy/litfacts.htm (last accessed 15 Sept 2008).
This paper has addressed the frequently debated question of [3] ITU: International Telecommunications Union. (2003). Mobile
input modalities of touchtone vs. speech in a completely new overtakes fixed: Implications for policy and regulation. [Online].
Available:
context; low literacy users and a domain (health) different
http://www.itu.int/osg/spu/ni/mobileovertakes/Resources/Mobileovertak
from the usual business centric applications (call routing, es_Paper.pdf, (last accessed 5 September 2008).
voice mail systems or banking). Our pilot study also served to [4] Statistics South Africa. (2007). Community Survey 2007. [Online]
confirm the feasibility of SDS applications for semi and low Available at: http://www.statssa.gov.za/publications/P0301/P0301.pdf
literate populations. (last accessed 15 Sept 2008).
[5] E. Barnard, L. Cloete, and H. Patel. “Language and Technology Literacy
Barriers to Accessing Government Services,” Lecture Notes in Computer
It is interesting to speculate on the applicability of our Science, vol. 2739, pp. 37-42, 2003.
findings to other resource-poor countries. Our intuitive [6] M. Plauche, U. Nallasamy, J. Pal, C. Wooters, and D. Ramachandran.
judgment is that the only characteristic of our user population “Speech Recognition for Illiterate Access to Information and
Technology,” in Proc. IEEE International Conference on Information
that strongly influenced our results was their relative and Communications Technologies and Development ‘06, pp. 83-92,
familiarity with mobile telephones; however, the matter May 2006.
requires detailed investigation. In future work we would [7] P. Nasfors. "Efficient Voice Information Services for Developing
therefore like to further explore the space of non-literate users Countries", Master Thesis, Department of Information technology,
Uppsala University, Sweden, 2007.
and the suitability of SDS applications for them, as well as [8] J. Sherwani, N Ali, S. Mirza, A. Fatma, Y. Memon, M. Karim, R.
investigate the interaction of task types (linear vs. non-linear) TRongia and R. Rosenfeld, “Healthline: Speech-based Access to Health
and application domain – informational (Openphone) vs. Information by low-literate users”, in Proc. IEEE International
transactional (tracking a social services payment) with the Conference on Information and Communications Technologies and
Development ‘07, Bangalore, India, Dec. 2007.
input modality. Once the system is free of charge and has been [9] S. Agarwal, A Kumar, AA Nanavati and N. Rajput, "VoiKiosk:
in use for several months, we would like to study whether Increasing Reachability of Kiosks in Developing Regions", in Proc. of
OpenPhone leads to a change in health habits and improves the 17th international conference on World Wide Web, pp. 1123-1124,
the ability of a caregiver in providing care to children, like 2008.
[10] K. M. Lee and J. Lai “Speech Versus Touch: A Comparative Study of
change of hygiene habits, better nutrition, and fewer the Use of Speech and DTMF Keypad for Navigation” International
misconceptions about HIV. Journal of Human–Computer Interaction, vol. 19, no. 3, pp.343–360,
2005.
Throughout the design and development process, we [11] B. Suhm, J. Bers, D. McCarthy, B. Freeman, D Getty, K Godfrey and P
Peterson, "A comparative study of speech in the call center: natural
experienced that beyond usability and creating simple,
language call routing vs. touch-tone menus". In: Terveen, Loren (ed.)
accessible user interfaces for low literacy users, factors such Proc. of the ACM CHI 2002 Conference on Human Factors in
as cultural and social context, establishing relationships with Computing Systems Conference, Minneapolis, Minnesota. pp. 283-290,
stakeholders and user communities, and localizing content, April 2002.
play a vital role in the success of an ICT intervention in the [12] C. Delogu, A. Di Carlo, P Rotundi and D. Sartori, "Usability evaluation
of IVR systems with DTMF and ASR", in Proc. of the 5th International
developing world. We intend to draw on these valuable Conference on Spoken Language Processing (ICSLP 98), paper 0320,
experiences and carry our work forward in serving the Australia,1998.
information needs of citizens of developing regions with [13] T. Pearce and M Bergelson. (2008). Alignment index for speech self-
accessible and usable telephony based services. service. Dimension Data Technical Report [online]. Available at
http://www.dimensiondata.com/NR/rdonlyres/9191A848-5F35-459F-
8239-
8D9D2248414E/8791/mainstreamspeechalignmentindexreport2.pdf
ACKNOWLEDGEMENTS (Last accessed 15 Sept 2008).
[14] J.P. Migneault, R. Farzanfar, J.A. Wright, and R.H. Friedman. “How to
The authors wish to thank the staff of Botswana Baylor write Health Dialog for a Talking Computer,” Journal of Biomedical
Children’s Clinical Centre of Excellence for the generous Informatics, vol. 39, no. 5, pp. 468 – 481, Oct. 2006.
manner in which we were received, in particular Dr Paul [15] Tellme Networks Inc. Who uses TellMe? [Online]. Available at:
http://www.tellme.com/about (Last accessed 15 Sept 2008).
Mullan, for the amount of help and time given to us. We [16] UNAIDS (August, 2008). Report on the Global AIDS Epidemic. Joint
would also like to thank members of the HLT group at United Nations Programme on HIV/AIDS. [Online] Available at:
Meraka Institute who provided valuable contributions http://www.unaids.org/en/KnowledgeCentre/HIVData/GlobalReport/200
throughout the project; Victor Zimu, Jama Ndwe, Mpho 8/2008_Global_report.asp (last accessed 7 Sep 2008).
[17] K. Seipone, W. Jimbo, F.d.l.H. Gomez, K. Ampomah, J. Othwolo, O.
Kgampe, Louis Joubert, Richard Carlson, Bryan Mcalister,
Kaluwa, M. Busisiwe. “Trends in HIV Prevalence Among
Mark Zsilavecz, Marelie Davel and Alta de Waal. We also
Pregnant Women in Botswana 2001-2005” 16th Intl. Conf
greatly appreciate the time given to us by Connie Ferguson, Aids. Toronto, Canada, Aug. 2006.
our celebrity voice talent. This research would not have been [18] Botswana-Baylor Children’s Clinical Centre of Excellence, Annual
possible without the joint support of OSI and OSISA. The Report 2006, Gaborone, Botswana, 2006.
project was also in part supported by the NRF under the Key [19] A. Van Dyk, HIV Aids Care & Counselling, a Multidisciplinary
Approach. 2005, Third Edition: Pearson Education, South Africa.
International Research Capacity (KISC) programme, UID no.
63676.
107
[20] D. Werner, C. Thuman and J. Maxwell. Where there is no Doctor, a
Village Care Handbook.. 2007, New revised edition: Hesperian, CA,
USA.
[21] R. Granich and J. Mermin. HIV, Health & your Community, A Guide for
Action. 2001, Hesperian, CA, USA.
[22] G.A. Miller. "The Magical Number Seven, Plus or Minus Two: Some
Limits on Our Capacity for Processing Information," The Psychological
Review. vol. 63, no. 2 pp. 81-97, 1956.
[23] Suhm B. “IVR Usability Engineering using Guidelines and Analyses of
end-to-end calls” in D. Gardener-Bonneau and H.E. Blanchard (Eds).
Human Factors and Voice Interactive Systems. 2008, pp. 1-41, Second
Edition, Springer Science: NY, USA.
[24] N.M. Fraser and G.N. Gilbert, “Simulating speech systems”, Computer
Speech and Language, vol. 5, no. 2, pp. 81-99, 1991.
[25] I. Medhi and K. Toyama, “Full-Context Videos for First-Time, Non-
Literate PC Users” ”, in Proc. IEEE International Conference on
Information and Communications Technologies and Development ‘07,
Bangalore, India, Dec. 2007.
[26] M.A. Walker, D. Litman, C.A. Kamm and A. Abella. “PARADISE: A
general framework for evaluating spoken dialogue agents” in Proc. of
the 35th Annual Meeting of the Association of Computational
Linguistics. ACL/EACL 97, 1997.
[27] Nielsen, J., (1993). Usability Engineering. AP Professional, Boston,
MA, USA.
[28] Foster, J. C., McInnes, F. R., Jack, M. A., Love, S., Dutton, R. T., Nairn,
I. A., et al. (1998). An experimental evaluation of preference for data
entry method in automated telephone services. Behaviour & Information
Technology, 17, 82–92.
[29] Goldstein, M., Bretan, I., Sallnas, E.-L.,&Bjork, H. (1999). Navigational
abilities in voice-controlled dialogue structures. Behaviour &
Information Technology, 18, 83–95.
[30] E. Brewer, M. Demmer, M. Ho, R.J. Honicky, J. Pal, M. Plauche, and
S. Surana. “The Challenges of Technology Research for Developing
Regions,” IEEE Pervasive Computing, vol. 5, no. 2, pp. 15-23, April-
June 2006.
108
2
The concept of empowerment originated in work on gender relations and
community participation (e.g. [33], [34]) and has been increasingly discussed
in development studies (e.g. [35],[36], [37], [38]. There are several competing
1
For a more in-depth discussion of the capability approach, see also [32]. definitions of the term.
110
struggle to quantify “social capital” [48]. As a result, critics A. Outcomes
have argued that “everything social” gets packed into the True to Sen’s statement that choice is both the aim and the
social capital variable [49]. principal means of development [51], the primary
development outcome is choice itself. Secondary development
Individuals own or have access to this portfolio of capitals, outcomes depend on the individual’s choice as to what lives
their “livelihood assets” with which they negotiate “policies, they value. These may include, for example, easier
institutions and processes”. They operate within a communication, increased knowledge, more income or time
“vulnerability context” and develop livelihood strategies saved. Information and communication technologies might
which then result in livelihood outcomes. Livelihood prove useful tools in achieving these outcomes. Just like other
outcomes are defined a priori – In the DFID version, “more attempts to operationalise Sen’s work, here capabilities are not
income” is listed at the top, even before “increased measured directly, though participatory research with
wellbeing”. The SLF offers a broad and systemic view of individuals and groups may reveal them to some degree.
development processes, but its Mainly, the outcome component will map or measure the
STRUCTURE
Key:
ER = Educational Resources SR = Social Resources
• institutions and organisations PsR = Psychological Resources NR = Natural Resources
In = Information MR = Material Resources
• discourses FR = Financial Resources GR = Geographical Resources
CR = Cultural Resources He = Health
• policies and programmes
• formal and informal laws
including:
- norms on usage of space DEVELOPMENT
- norms on usage of time OUTCOMES
• access to ICTs
- availability of ICTs
- affordability of ICTs Principal: Choice
- necessary skills for ICTs
DIMENSIONS OF
CHOICE
Secondary:
• existence of choice • easier communication
• sense of choice • increased knowledge
• access to markets
AGENCY • use of choice
• business ideas
SR
• achievement of • increased income
He CR choice
• more voice
Age
ER NR • time saved
Gender
Ethnicity • higher job satisfaction
PsR MR
... •...
In GR
FR
set of capitals is limited and in it the development goals are achieved functionings resulting from an individual’s choices
predetermined and not up to the individual to choose. In this as a proxy to the capabilities. 3 An analysis based on the
respect, the SLF fails to mirror the thinking behind Sen’s Choice Framework would then work backwards, from the
approach. outcomes, into the systemic relationships between agency,
structure and choice, thus analysing how the outcomes were
arrived at.
III. THE CHOICE FRAMEWORK
Based on Sen’s capability approach, inspired by Alsop and B. Dimensions of Choice
Heinsohn’s work on operationalising Sen’s work, taking
Alsop and Heinsohn’s dimensions of choice, which they
elements from the SLF and informed by an in-depth research
call “degrees of empowerment” include, firstly, the existence
project with microentrepreneurs’ use of ICTs in Chile [50],
the Choice Framework was developed. After presenting it in 3
Two disadvantages of this method are that some of the individual’s
diagrammatic form (Fig 1), the following sections will in turn capabilities are not captured in the achieved functionings and that it is difficult
explain each of the key components of the framework. to trace the choices related to apparently negative outcomes. However, so far,
capturing functionings is methodologically easier and more precise – in
regards to both quantitative and qualitative methods - than capturing
capabilities.
111
of choice – whether the different possibilities exist and are, in face to face contact in the urban economy [55]).
principle, attainable for the individual if the combination of
their resource portfolio and the structural conditions would Human Resources: The term “human resources” has been
allow it. The second dimension, a sense of choice, not used for decades in the economics and industrial relations
originally included by Alsop & Heinsohn, was added as a literature. 4 In the Choice Framework, this term needs to be
result of fieldwork experiences relating to ICT and disaggregated into Health and Education and Skills
development. Individuals were aware of some possibilities the (educational resources). Within Sen’s paradigm of
new technology offered them, like email and online chat, but development, good health is a prerequisite for a person’s
not of others, like Voice over IP. This was precisely because ability to choose the life she/he values. Educational resources
their educational resources (including computer skills) and the represent education and skills acquired through formal and
dominant discourse in the Chilean media stressed some usages informal means.
over others. For any piece of research focused on a technology
which is new to the respondents, the dimension of “sense of Psychological resources: Alsop and Heinsohn [56]
choice” will play a significant role. The “use of choice” recognise the significance of “psychological assets” and give
dimension refers to whether or not an individual actually as an example “capacity to envision”. More broadly,
makes the choice and the “achievement of choice” refers to psychological assets may include self-confidence, tenacity,
whether the outcome matches the choice expressed. optimism, creativity and resilience. Spiritual or religious
beliefs stand in complex interrelation with psychological
resources – they can strengthen or weaken an individual’s
psychological resources.
C. Agency
Instead of using a terminology of capitals and asset or Information: Alsop and Heinsohn list informational assets
capital portfolios, Sen uses the term “resources” within the as a key resource. Heeks [58] calls for putting information at
capability approach [52]. Resources can be interpreted as the centre for analysis of ICTs and Development, and Gigler
individual agency-based capability inputs which, together [59], adds “informational capital” to the capital portfolio.
with structure-based capability inputs, can be converted into Access to information is the first step to knowledge
capabilities [53]. acquisition, the process of filtering and transforming
information into meaningful knowledge.
In the Choice Framework age, gender, ethnicity etc. are
conceptualised as personal characteristics of an individual Cultural resources: “Cultural capital” – which in the Choice
which may in a given social context become related to socially Framework is called cultural resources – exists, according to
constructed axes of exclusion and influence the scope and Bourdieu [60], in three states: an embodied state (the habitus a
scale of the resource portfolio. The resource portfolio consists particular person lives in); an objectified state (objects like
of: paintings, instruments and monuments which only the
initiated can use or appreciate); and an institutionalised state
Material resources: These sum up the material objects (prestige attached to, for example, academic titles).
owned, such as machinery, computer hardware and other
equipment. They are also essential inputs in the production Social resources: “Social capital” – or social resources – is
process. included in both the SLF and Alsop and Heinsohn’s work. It
has been both immensely influential and highly contested in
Financial resources: These stand for financial capital in all development discourse. For the Choice Framework,
its forms (cash, savings, shares etc.). The ability to obtain Bourdieu’s definition of social capital is used:
credit is a combination of the structural character of the
banking rules and individual collateral. “the aggregate of the actual and potential resources which
are linked to possession of a durable network of more or less
Natural resources: This includes issues such as institutionalized relationships of mutual acquaintance and
geomorphological and climatic conditions in a locality and recognition – or in other words, to membership in a group –
related aspects such as soil quality and the availability of or which provides each of its members with the backing of the
access to water as well as the attractiveness of the surrounding collectivity-owned capital, a “credential” which entitles them
nature. to credit, in the various senses of the word.”[61]
Geographical resources: Covers the practical implications Membership of these groups can be defined by kinship,
of location and relative distances, and also includes the friendship, shared ethnicity or class, friendship or informal
intangible qualities of a location alluded to by writers from commonality ties.
Marshall (who refers to the mysteries of the trade “in the air”
[54]) to Storper and Venables (who describe the “buzz” of 4
For examples, see the collection by Fitzgerald and Rowley [57]
112
diagram of the Choice Framework.
Thus, these ten types of resources – material, financial,
natural, geographical, psychological, cultural, social, and The Choice Framework is an attempt to operationalise the
educational (education and skills) resources; health; and capability approach in a holistic and systemic way, thus
information – represent an attempt to holistically map aspects maintaining much of its conceptual richness. While it may
of the agency element of the systemic framework. prove particularly useful in the area of ICT4D, the framework
could also be applied in other areas of development work.
However, it is important to recognise that this resource-
based agency can only be realised within the confines of and
in systemic interaction with a given structure. This aspect of
IV. IMPLICATIONS FOR METHODOLOGY
the Choice Framework will be analysed in the following
section.
The basic challenge that the capability approach offers to
the orthodox methodologies of development research, and
D. Structure
ICT4D in particular is that, on a fundamental level, it
Both the empowerment framework suggested by Alsop and questions the validity of outcomes that are defined a priori and
Heinsohn and the SLF take into account not only individual without consulting the individual in question. Both the
agency, but also structures which aid or constrain this agency. inclusion of a development goal and its position within a set
Alsop and Heinsohn list “formal and informal laws, of development priorities, however, relate to the question
regulations, norms and customs” [62] as elements of this which kind of life people would choose to live and this,
structure, while the DFID SLF includes these as laws and according to Sen, is what development is about. A funding
“culture” – the latter running the risk of being used as a kind institution or government may set, say, economic prosperity as
of black box into which all locally specific aspects can be the top priority. Once basic needs such as food and shelter
subsumed. The SLF includes not only laws, but also policies, have been met, however, an individual may value being close
institutions and processes. Rules, laws, norms and policies are to family members above earning more money, or may value a
embedded in, and often emanate from discourses, and healthy environment for themselves and their children over
hegemonic discourses can define the thinkspace in which economic growth. In the practice of development projects, this
policies, including ICT policies, can be conceived. Thus means that before undertaking an intervention designed to
discourses are included as part of the structure element of the improve people’s lives and later measuring its effectiveness,
Choice Framework. practitioners and researchers would have to ask individuals
about their own development priorities and let these guide the
In particular with respect to ICTs, relevant elements of the planning, implementation, monitoring and evaluation of
structure which influence an individual’s agency include development projects and programmes. Work done in this
dimensions of access, such as availability, affordability and area includes participatory monitoring and evaluation [64]
capabilities needed for using different ICTs [63]. To avoid and, in a broader context, initiatives around participatory
confusion with Sen’s use of the word, the term “skills” is used budget planning [65].
instead of capabilities. These dimensions of access are
nationally and often locally specific, path dependent and Setting development priorities in a participatory way may
embedded with other elements of the structure. make the process of development planning more complicated,
but it comes with major benefits: Firstly, morally it is the right
Structural factors such as these stand in a complex thing to do to engage the people themselves in the decisions
relationship with an individual’s resource portfolio. For that will affect their lives. Secondly, if the outcomes have
example, with the help of social resources an individual might been agreed upon in a participatory way, they are more likely
have access to the internet (at a neighbour’s house) which to be locally and culturally appropriate and may reduce the
might lead to frequent email contact with a distant family rate of failure. Thirdly, a participatory process will lead to
member, thus increasing occasions of, in Bourdieu’s terms, greater local buy-in to measures and therefore higher future
legitimate exchange with both the neighbour and the distant institutional sustainability. Fourthly, such a process harbours
relative, in turn potentially increasing social resources. the chance that the current overly economistic focus of
Similarly, a person with higher educational resources (skills development work can be broadened to include
and education) and information might find it easier to use the environmental, social and cultural aspects and thus better
existing access facilities to enhance their skills and gain mirror the diversity of the kinds of things people value in their
information. The interface between the opportunity structure lives. Last, and perhaps not least, ICT and development
and individual agency thus includes a host of reciprocal and practitioners work with multi-purpose technologies which
cumulative processes. Structural constraints need to be offer far more significant changes to people’s lives than the
recognized as being as important an element as individual economic impact they have been proven to have. Moving
agency. To reflect this, structure is placed above agency in the away from an a priori, top-down and often overly economistic
113
set of development priorities offers the chance to recognise here in the telecentre.
the diversity of the contributions ICTs can make to the social,
cultural, environmental and economic aspirations individuals If one were to apply a typical questionnaire on telecentre
may have for their lives. usage to this case, this woman’s usage experience might be
subsumed in the category “personal usage” or “other”. Yet the
The Choice Framework is one way of conceptualising such following section will offer a careful application of the Choice
choices by the individual. There are some specific Framework to this case:
implications for research design which result from the model:
A focus on the individual’s own development outcomes
A. Outcome
means that the research needs to start from these wished-for
outcomes, measure the degree to which they have been The primary outcome was that the respondent had improved
attained and work systemically backwards through structure, choice, in this case, between “no visit to Kaiserslautern” and
agency and choice to understand how these outcome have “virtual visit to Kaiserslautern”. The secondary outcome
come about. The extensive list of resources covers six less achieved was defined by the individual: “to see more of the
tangible resources (social, cultural, educational, psychological world” – which in this case, translated into “virtual visit to
resources, health and information) which pose challenges to Kaiserslautern” - in Sen’s terms an “achieved functioning”.
measurement but need to be taken into account. On the The aspiration was “visit to Kaiserslautern”, which since it is
structure side, access to ICTs is conceptualised as availability, feasible, could be seen as a “capability” in Sen’s terms. The
affordability and necessary skills. ICTs are embedded in the achieved functioning ”virtual visit” is not equal to the
wider set-up of institutions, policies, programmes, norms and capability “visit in person” but it is an improvement in
discourses. As such they need to be analysed as firmly and outcome over no visit at all.
historically engrained in the societies they affect.
Studies of development outcomes, or more commonly of
impacts, often operate with a set of impacts as defined by the
V. APPLYING THE FRAMEWORK: IMPACTS OF TELECENTRES IN funding body, government, international organisation or
RURAL CHILE commercial sponsor. This set of impacts then acts as a
checklist informing the construction of questionnaires and
The following example is part of an extensive ethnographic
interview guides, possibly with some scope for “other
study of how state ICT policies affected microentrepreneurs in
activities”. Sen’s approach, with the individual’s choice as the
rural Chile [66]. When interviewing microentrepreneurs and
primary outcome, however, would suggest that the analysis
their partners who were using a telecentre located in a public
needs to start from the ground up, asking people about what
library in rural Chile, open-ended questions revealed that apart
lives they value and what outcomes they want to see. For this
from business-related usages such as looking up prices of
individual, one of the greatest impacts the telecentre had made
machinery on the internet (carpenters), looking up photos of
was that it had given her the chance to virtually visit
furniture models (carpenters), communicating via email with a
Kaiserslautern, something few policymakers or researchers
supplier (carpenters, spice vendor) or buyer (spice vendor),
would have predicted. Indeed, some might question whether
looking up recipes (cake vendor), and looking up guidelines
this is a valid “development outcome” or “impact” for a
for government business assistance (carpenters), there were
telecentre. In Sen’s approach, expressed via the Choice
several answers which reflected what in many studies of
Framework, it is.
telecentre usage is described as “personal usage”.
Respondents’ faces lit up when they described how they now
could exchange emails or chat with relatives who were living B. Agency
abroad, children who were studying in a larger town or family The individual in question was a Chilean-mestizo 5 woman
members who were working as temporary labour on fruit in her 50s, married with four children who were all grown up
farms or in the mines in the north of Chile. One respondent, a now and had left the home. Her material resources did not
woman in her 50s whose household income was around 440 include a computer and internet access at home and her
USD per month and who together with her husband ran a financial resources made it difficult for her to spend money on
carpentry business, described how while the Word Cup was using a computer in the local cybercafés. However her social
on in Germany in 2006 she visited the world cup site to find resources (contacts with friends) had helped her gain the
links and take virtual tours of some of the German cities she information that there was free access to the internet available
was not able to visit in person. As a young woman, she had at the telecentre in the local library. Her geographical
had a pen friend from Kaiserslautern, and while he had come resources (the location of her house) and her state of health
to visit her in Chile, her dream of visiting him had never been were such that she could easily reach the telecentre on foot.
possible because, she said, the money she saved had been She had the cultural resources to not feel intimidated when
spent on her children’s education. Eventually, they had lost
touch, but now, she told me with tears in her eyes, over 25 5
i.e. not considered part of the indigenous minority which had historically
years later, she was finally able to “visit” Kaiserslautern, right been discriminated against.
114
entering a space like a library and to know the behavioural information, and psychological resources, knew that the
code there. With the help of her social resources (knowing the telecentre offered a computer and a good connection, had
librarian who was now also the director of the telecentre), her acquired the skills to navigate the Internet and run an
educational resources (literacy, rudimentary English) and her application in the free digital literacy courses offered at the
psychological resources (extrovert, willingness to ask telecentre. She felt that informal, gendered social norms
questions) she quickly learnt how to use the computers. The allowed her to go to the telecentre during the morning before
information she gained online, together with her psychological having to prepare lunch. Thus she developed a sense of
resources (curiosity, tenacity) allowed her to understand the choice, was able to choose (use of choice) and achieved her
choices she had and find the site which offered the virtual tour desired outcome (achievement of choice).
of Kaiserslautern, thus achieving her chosen development
outcome.
VI. CONCLUSION
C. Structure Applying the Choice Framework to this particular case
The agency of the individual is a shaper of, and is shaped allows us to firstly, theorise the use of ICT in a systemic and
by the structure in which it operates. In this case, as part of the procedural way which reflects the systemic and pervasive
national ICT policy, the Agenda Digital, the state of Chile had impact of ICT. The “impact of ICT” is not conceptualised in a
signed an agreement with the Bill and Melinda Gates cause- and effect chain, instead effects are carefully
Foundation for them to provide 9.2 million USD worth of disaggregated and their systemic interrelatedness and co-
hardware to be installed in public telecentres based in libraries causality is demonstrated. Secondly, the Choice Framework
around the country, running Microsoft software. The local offers a way to operationalise Sen’s capability approach in the
library was an existing institution which was able to context of ICT. Sen’s approach is currently the most well
accommodate the hardware, delivered as part of the known heterodox alternative to orthodox, growth-focused and
Biblioredes telecentre programme (availability of ICTs). The often economistic conceptualisations of development. Given
Chilean digital literacy campaign provided free IT courses to the enormous potential of ICTs to give individuals choices,
adults, and public discourse in Chile stressed the importance and indeed a greater sense of choice, Sen’s approach is of
to become “digitally literate”, so the woman had taken the particular interest to the ICT and development research
course (necessary skills). The formal rules for users of the community.
telecentre stipulated that access was free (affordability of
ICTs) but limited to 30 minutes per person at busy times, and There are three obvious limitations to the application of the
the informal rules were that people were left to use the Choice Framework, and this is where more theoretical work
computers on their own unless they asked for help from the needs to be done:
telecentre director. Norms on the usage of space made it easy
for her to go to a library as a mestizo woman with a completed Firstly, the Choice Framework aims to be comprehensive in
school education who was known in town. However, norms its modelling of the complex relationships between agency,
on the usage of time meant that she could only use the structure, degree of empowerment and outcome, and this
telecentre when she was not supposed to be home preparing automatically entails a trade-off with the depth of theorisation
meals (gendered norms on time) or when she was expected to of each element. Behind each of the terms included in the
attend to customers (business norms on time). framework lies a wealth of theoretical literatures which may
need to be synthesised for different research purposes and key
issues brought to the attention of researchers in the
D. Dimensions of choice development field. While for example, social resources can be
In the Choice Framework, an individual’s resource-based theorised by linking to the wider debate on social capital (see
agency can operate within a given structure to achieve degrees [66]), which has been received in the development studies
of empowerment, such as existence of choice, sense of choice, discourses, work on cultural capital (in Bourdieu’s sense) is
use of choice and achievement of choice. In this case, both the hardly ever linked to development discourses in the South.
choice “travel to Kaiserslautern in person” and “take a virtual
tour of Kaiserslautern” existed, the latter only since the links Secondly, the Choice Framework is relatively easily applied
were offered via the World Cup website in 2006. In a in qualitative work on the micro-level of the individual. A
capitalist market system, however, the former choice required further challenge will be how to apply the framework to
an amount of financial resources which the individual felt groups of individuals, communities, or even nations. Within
unable to dedicate to this idea. The choice “take a virtual tour this and related to a theoretical tension evident in Sen’s
of Kaiserslautern” however, required a good internet original approach, there is a complex relationship between
connection, a computer, the knowledge that the tour was individual and collective choice which will have to be
available via the website, the skills to find and run it, and time. conceptualised carefully.
The individual, thanks to among others, her social resources,
115
The example used here to illustrate the potential of the Secondly, there are some macro-methodologies which
Choice Framework is part of a far more extensive reflect the ethos of giving people the power to choose.
ethnographic and longitudinal study of how the state ICT Voucher schemes are a good pragmatic way to monitor, in a
policies affected microentrepreneurs in rural Chile [66]. In heavily supply-driven development field such as ICT4D, what
this study, connections are made between individual and products (hardware, software etc), services (trainings,
collective choice. For example, the local authority’s public computer repair, communication etc) and content (economic,
procurement policy was analysed as an expression of social, political, cultural etc) people would, after considering
collective choice, but related to individual’s views as to how their options, actually choose. From the field of participatory
their tax money should be used to create the community they urban planning come methodologies for participatory budget
wanted to live in. However, further empirical work is design, where communities get to debate and decide which of
necessary to gather experiences in the applicability of the their desired outcomes to prioritise and pursue. This is a
Choice Framework in other cultural and socio-economic practical and democratic way to aggregate individual
settings. The relationship between individual choice and capabilities in order to enable collective decision making, and
collective choice needs to be conceptualised carefully in these could also be used for ICT4D.
local contexts in order to allow for the empirical application of
the Choice Framework at the more aggregate level such as so- Thirdly, practitioners may deduce that if the ideal is for
called target groups and communities. development projects’ intended outcomes to reflect the
individual’s choices, then the more individuals are aggregated
The third limitation is a very practical one: funders prefer to a group, the less probable it is that they can agree on a
predefined and clearly measurable impacts. The Choice similar set of capabilities. From this follows that the further
Framework, however, suggests that impacts of ICTs occur in a down the directional control continuum an ICT4D project is,
systemic, pervasive and transversal way, and that outcomes the more sensitive/locally customised it has to be to the
should be defined, in line with Sen’s approach, by the choices of a smaller number of people. Big, uni-directional
individual, based on their choices as to what kind of life they development programmes with specific, a priori defined
value. There are, however, some funders who are open to desired outcomes designed for a large number of people are
methods such as participatory evaluation and monitoring, and most likely to be in contradiction to a people-centred holistic
this means there might be hope for genuinely people-centred development process as proposed by Sen and expressed in the
development work – and development theory. Choice Framework.
There are some key implications emerging from this Ultimately, researchers working on ICT, particularly the
theoretical work for practitioners of ICT. Firstly, while no internet, need to consider the question: Should we try and fit a
technology is ever completely politically neutral [67][68], groundbreaking, multi-purpose and potentially liberating
ICT4D projects can be placed on a continuum of “directional technology into orthodox notions of development – such as
control”. At one end there are projects and programmes more ICT for higher GDP, more ICT for better school results
which focus on providing people with access to a technology et cetera – impacts which we may struggle to prove? Or can
which is recognised as multi-purpose, like some telecentre the field of ICT and Development serve as a test case and
projects. On the other end of the continuum are projects and breeding ground for thinking about development in a more
programmes which carry a much more narrow set of holistic way, putting the individual and their own choices at
intentions, for example training microentrepreneurs to use a the centre of development? If the latter is the case, then we
specific e-procurement system in order to “train them” to have plenty of work to do, but the Choice Framework may
operate in a more competitive market environment under a serve as one part of the big puzzle we have to begin putting
specific set of rules [67]. The further down the directional together.
control continuum a particular project and programme is
located, the more risk there is that the intended outcomes of
an ICT4D project diverge from the capabilities, or desired ACKNOWLEDGMENT
outcomes individuals in the so-called target group would I would like to thank Diane Perrons, Robin Mansell,
choose. Thus, the more directional control is involved in the Alexandra Norrish, Macarena Vivent, Rodrigo Garrido, Tim
project or programme, the more participation of the set of Unwin, and four anonymous reviewers for their comments on
individuals who are the intended group will be needed to earlier versions of this paper.
reduce this gap. This would include conceptualising the
development process as open-ended and the so-called target
group as individuals empowered to choose the lives they
themselves value. Participatory project design and
participatory monitoring and evaluation techniques would be
most appropriate.
116
Chile, unpublished PhD thesis, Department of Geography and
Environment, London, London School of Economics, 2007
[33] C. Moser, Gender Planning and Development, London, Routledge, 1991
REFERENCES [34] W. Van Eyken, The Concept and Process of Empowerment. The Hague,
[1] R. Heeks, “Information systems and developing countries: Failure, Bernard van Leer Foundation, 1991
success and local improvisations” in The Information Society, vol. 18, [35] J. Friedmann, Empowerment: The Politics of Alternative Development,
no. 2, pp. 101-112, 2002 Oxford, Blackwell, 1992
[2] W.A. Lewis, “Economic development with unlimited supplies of [36] N. Kabeer, “Resources, agency, achievements: reflections on the
labour”, The Manchester School of Economic and Social Studies, vol.22, measurement of women’s empowerment”. Development and Change,
no. 2, pp.139-91, 1954 vol. 30, no.3, pp.435-464, 1999
[3] G. Myrdal, Economic Theory and Underdeveloped Regions, London, [37] P. Oakley (ed.), Evaluating Empowerment, Reviewing the Concept and
Duckworth, 1957 the Practice, Oxford, INTRAC, 2001
[4] A. O. Hirschman, The Strategy of Economic Development, Yale, Yale [38] A. J. Bebbington, M. Woolcock, M. Guggenheim and E.A. Olson (eds),
University Press, 1958 The Search for Empowerment. Social Capital as Idea and Practice at
[5] W. Rostow, The Stages of Economic Growth: A Non-Communist the World Bank, Bloomfield, CT, Kumarian Press, 2006
Manifesto, Cambridge, Cambridge University Press, 1960 [39] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
[6] A.G. Frank, Capitalism and Underdevelopment in Latin America, Structuring Analysis and Framing Indicators, Washington D.C.: World
London, New York, Monthly Review Press, 1967 Bank, p.5, 2005
[7] T. Dos Santos, Dependencia y Cambio Social, Santiago, Universidad de [40] Ibid, p.6
Chile, 1970 [41] Ibid, p.8
[8] M. Blomström and B. Hettne, Development Theory in Transition. The [42] Ibid, p.9
Dependency Debate and Beyond: Third World Responses, London, Zed [43] R. Chambers and G. R. Conway, “Sustainable Rural Livelihoods:
Books, 1984 Practical Concepts for the 21st Century”, Discussion Paper 296, Institute
[9] C. Kay, Latin American Theories of Development and of Development Studies, University of Sussex, UK, 1992
Underdevelopment, London, New York, Routledge, 1989 [44] A. J. Bebbington, “Capitals and Capabilities: A Framework for
[10] M. Nerfin (ed.), Another development: Approaches and Strategies, Analyzing Peasant Viability, Rural Livelihoods and Poverty”, World
Uppsala, Dag Hammarsköld Foundation, 1977 Development, vol. 27, no.12, pp. 2021-2044, 1999
[11] R. Chambers, Rural Development: Putting the Last First, Harlow, [45] D. Carney, Sustainable Livelihood Approaches Compared, London,
Longman, 1983 Department for International Development
[12] A. Escobar, Encountering Development: The Making and Unmaking of [46] DFID, Sustainable Livelihoods Guidance Sheets, Department for
the Third World, London, Longman, 1995 International Development, London, 1999
[13] J. Nederveen Pieterse, “After post-development”, Third World [47] R. Duncombe, “Using the Livelihoods Framework to Analyze ICT
Quarterly, vol. 21, no.2, pp.175-191, 2000 Applications for Poverty Reduction through Microenterprise”,
[14] A. Sen, Development as Freedom, Oxford, Oxford University Press, Information Technologies and International Development, vol. 3, no. 3,
1999, p.3 pp.81-100, 2006
[15] J. Harriss, Depoliticizing Development – The World Bank and Social [48] DFID, Sustainable Livelihoods Guidance Sheets, Department for
Capital, London, Anthem Press, 2001 International Development, London, 1999
[16] A. Sen, “Equality of what?”, in S. McMurrin (ed.) The Tanner Lectures [49] A. Munasib, Social Capital at the Individual Level: A Reduced Form
on Human Values, Salt Lake City, 1980 Analysis. Available:
[17] A. Sen, Resources, Values and Development, Cambridge, MA, Harvard http://munasib.myweb.uga.edu/papers/SocialCapitalDemography.pdf,
University Press, 1984 2004
[18] A. Sen, “Capability and well-being”, in M. Nussbaum and A. Sen (eds.) [50] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs,
The Quality of Life, Oxford, Clarendon Press, 1993 Information and Communication Technologies and State Policies in
[19] A. Sen, Development as Freedom, Oxford, Oxford University Press, Chile, unpublished PhD thesis, Department of Geography and
1999 Environment, London, London School of Economics, 2007
[20] A. Sen, Development as Freedom, Oxford, Oxford University Press, [51] A. Sen, Development as Freedom, Oxford, Oxford University Press,
1999, p.3 1999
[21] M. Nussbaum, Women and Human Development: The Capabilities [52] A. Sen, Resources, Values and Development, Cambridge, MA, Harvard
Approach, Cambridge, Cambridge University Press, 2000 University Press, 1984
[22] S. Alkire, Valuing Freedoms, Oxford, Oxford University Press, 2002 [53] I. Robeyns, “The Capabilities Approach: An Interdisciplinary
[23] D. Clark, Visions of Development, Cheltenham, Edward Elgar, 2002 Introduction”, Department of Political Science and Amsterdam School of
[24] I. Robeyns, “Sen’s capabilities approach and gender inequality: selecting Social Sciences Research Working Paper, University of Amsterdam,
relevant capabilities, Feminist Economics, vol. 9, no 2-3, pp. 61-92, Amsterdam, 2003
2003 [54] A. Marshall, Principles of Economics, London, Macmillan, 1920, reprint
[25] I. Robeyns, “The Capabilities Approach: An Interdisciplinary edition 1961
Introduction”, Department of Political Science and Amsterdam School of [55] M. Storper and A.J. Venables, “Buzz: face-to-face contact and the urban
Social Sciences Research Working Paper, University of Amsterdam, economy”, Journal of Economic Geography, vol. 4, no. 4, pp. 351-370,
Amsterdam, 2003 2004
[26] R. Sugden, “Welfare, resources and capabilities: a review of Inequality [56] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
Re-examined by Amartya Sen”, Journal of Economic Literature, vol. Structuring Analysis and Framing Indicators, Washington D.C., World
XXXVI, pp. 1947-1962, 1993 Bank, 2005
[27] J. Roemer, Theories of Distributive Justice, Cambridge, Cambridge [57] R. Fitzgerald and C. Rowley (eds.), Human Resources and the Firm in
University Press, 1996 International Perspective, Cheltenham, Edward Elgar, 1997
[28] A. Sen, “Capability and well-being”, in M. Nussbaum and A. Sen (eds.) [58] R. Heeks, “Information and communication technologies, poverty and
The Quality of Life, Oxford, Clarendon Press, 1993 development”. Development Informatics Working Paper No.5,
[29] A. Sen, “Maximisation and the act of choice”, Econometrica, vol. 65, Manchester, IDPM, 1999
no.4, pp.745-779, 1997 [59] B.S. Gigler, “Including the excluded – can ICTs empower poor
[30] A. Sen, Inequality Re-Examined, Oxford, Clarendon Press, 1992 communities? Towards an alternative evaluation framework based on
[31] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice – the capability approach,” presented at the Fourth International
Structuring Analysis and Framing Indicators, Washington D.C.: World Conference on the Capability Approach, Pavia, Italy, Sept. 5-7, 2004
Bank, 2005 [60] P. Bourdieu, “The forms of capital”, in: J. Richardson (ed.), Handbook
[32] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs, of Theory and Research for the Sociology of Education, New York,
Information and Communication Technologies and State Policies in Greenwood Press, 1986
117
[61] Ibid, p.249
[62] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
Structuring Analysis and Framing Indicators, Washington D.C., World
Bank, 2005, p.9
[63] R. Gerster and S. Zimmermann, Information and Communication
Technologies (ICTs) for Poverty Reduction?, Bern, Swiss Agency for
Development Cooperation, 2003
[64] I. Guijt and J. Gavena, Participatory Monitoring and Evaluation:
Learning from Change, Brighton, Institute of Development Studies,
1998
[65] A. Shah (ed.) Participatory Budgeting. Washington D.C., The World
Bank, 2007. Available:
http://siteresources.worldbank.org/PSGLP/Resources/ParticipatoryBudg
eting.pdf
[66] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs,
Information and Communication Technologies and State Policies in
Chile, unpublished PhD thesis, Department of Geography and
Environment, London, London School of Economics, 2007
[67] D. Kleine, “The ideology behind the technology – Chilean
microentrepreneurs and public ICT policies”, Geoforum, vol. 40, no.2.,
2009
[68] D. Kleine, “Striking a Balance”, Engineering and Technology, February
2007, p. 30-33
118
Figure 7. Urban Hospital (DSL), Wed-Tue Figure 8. Teaching Hospital (VSAT), Thu-Wed Figure 9. Rural Hospital (VSAT), Mon-Thu
Packet Loss Rates (%) and Round-Trip Average (ms.) for three hospital servers in Ghana. Left y axis shows median, 5th and 95th percentile of RTA.
environment with his experience in a leading hospital in the GCN anticipated power outages and network
U.S. Thus, while GCN framed its findings as “most doctors do disconnections occurring several times a day, but the
indeed have some experience with computers” and would use regularity with which (nominally acceptable) bandwidth was
them more often given better access, ML concluded that unusably low was a surprise. For example, certain pieces of
“computers are not part of the daily routine of a Ghanaian systems software on the local servers needed to be configured
physician” and so are not a useful option. with estimates of the worst-case time needed to transfer an
8KB file chunk; the initial estimate of 20 seconds (~3.3Kb/s)
C. Discussion
was eventually increased to 5 minutes (~0.2Kb/s).
Part of the difficulty in producing consistent assessments of Connectivity data obtained after the deployment of GCN show
conditions on the ground is that while metrics such as the that the difference between best- and worst-case performance
frequency of power outages and the availability of DSL can be can be extreme. Figures 7, 8 and 9 illustrate the volatility of
measured and mapped in great detail, it is up to each network performance at three sample hospitals. Figure 7
investigator to determine exactly what metrics need to be shows a DSL network connection with moderately variable
measured, how rigorous the measurement needs to be, and packet loss rates and round trip times. However, Figure 8
what quality of service constitutes something usable by the illustrates the case of a satellite connection so overloaded that,
target community for the specific application. GCN had no on weekdays between 10:00 AM and 4:00 PM, packet round-
resources or expertise with which to improve the mobile trip times to the public GCN server in Ghana consistently
phone infrastructure, so it focused more on network exceeded 10 seconds and loading google.com took 15
infrastructure. ML had no expertise with network minutes. While the GCN software was successfully able to
infrastructure, but instead saw an opportunity to address a very transmit doctors’ messages when congestion decreased in the
significant, non-technical barrier in mobile phone usage. evening, a doctor who requested a consultation in the morning
A closer look at the details of connectivity suggests that would wait until the following day to receive a response (even
both projects were somewhat optimistic in downplaying the if the consultant responded immediately upon receipt of a
infrastructural limitations. The qualitative assessments verified message). (The irony here is that this hospital, with the largest
expected connectivity barriers, whereas additional issues number of dedicated IT staff - 6 people - thus experienced the
remained hidden. most apparent ‘outages’.) In contrast, Figure 9 shows the
characteristics of a satellite connection that worked reliably
and consistently.
As ML is not based on Internet connectivity, an analogous
examination of ML’s assessment of mobile phone coverage
would require assessing the quality of service of OneTouch
voice calls and text messages. While we do not have this
information, we can look instead at geographical coverage
(Figure 10). While OneTouch has more coverage than any
other provider in Ghana, the coverage map shows that vast
regions of the country remain out of coverage area, leaving
doctors working in those regions at a significant disadvantage.
Clearly, a trip to a single urban center in the North does not
systematically gauge the limitations of coverage throughout
the country, let alone quality of service within current
coverage areas.
To be clear, the issues illustrated above have not been a
cause of project failure in either case. Further, exhaustive
Figure 10. OneTouch GSM Coverage Areas. (Source: GSM Association)
125
quantitative assessment of all such issues in advance of was surprising in light of how ‘medicalized’ the investigators
deployment is not always cost-effective or even possible (for perceived the interface. Yet perhaps because their means of
example, one cannot measure DSL links that are not installed). communicating with remote doctors are so limited, it appears
What we suggest here is that care must be taken to prevent the barriers between profession and person are much more
“blind spots” in assessing infrastructure and that such blind fluid than anticipated. Second, many of the doctors discussed
spots can arise from the investigators’ backgrounds. and requested functionality in the system for sharing literature
and PowerPoint presentations, to the extent that these features
VI. SOLUTION USAGE AND APPROPRIATION were included in the upgrade from the pilot to the ongoing
In this section, we discuss the usage of the technological deployment. This is consistent with the findings mentioned
solutions that arose from each project as well as the ways in earlier that a larger screen size was important in dealing with
which the solutions were appropriated for uses other than more information-intensive tasks, and also provides insight
those they were intended to enable. We then discuss the role into the kind of tasks matched to a computer’s affordances.
of what we call technological specificity in these processes. B. ML
A. GCN As previously noted, ML experienced an incredible rate of
As described previously, GCN adoption has been relatively adoption, with 1700 of 2000 members of the GMA signing up
slow; what is notable is that GCN experienced a better rate of within the first four months. While there are no statistics
adoption in the smaller hospitals than in the large ones. At the available concerning the fraction of usage that is related to
smallest deployment hospital, with only 2 doctors on staff, one consultation, there is little doubt (judging from the interviews
of the doctors continues to log into the system and post cases and from multiple instances of observing doctors as they
on a roughly bi-weekly basis. In the mid-sized hospitals, the received calls) that ML is used frequently for consultation.
response was mixed, with a handful of doctors using the There are multiple reasons for this popularity, some of
system every week or two but the majority losing interest. At which are hard to distinguish from the affordances and
the largest deployment hospital, however, after the initial popularity of mobile phones themselves. (One doctor claimed
flurry of activity surrounding the presentations and training, his phone bill dropped from 150 USD per month to 8 USD per
none of the doctors continued to use it on a consistent basis. month after joining the program.) First, mobile phones are a
Follow-up interviews shed some light on this disparity. popular medium for medical consultation because the real-
One issue was computer access. In the smallest district time nature of voice calls is often critical to treating an
hospital, the doctors shared one computer, but since there were emergency case – the three doctors who volunteered
only two doctors, having access to the computer was never a information on the breakdown of emergency/non-emergency
problem. In contrast, access to computers was an issue at the cases reported that around 80% of cases that require further
larger hospitals. One doctor said: consultation are indeed emergency cases. One said:
Sometimes you go to the library, you see someone at the I prefer phone calls to SMS, because I prefer an immediate
computer for 15 minutes; you don’t have that time to waste answer, and also so I can make sure the phone is on. If I’m
[waiting for the computer]. dealing with a case right now, I want to know what to do
when moving ahead as soon as possible.
Another issue was the match between the system’s use case
and the needs of the pilot participants. Through its design and Second, phone skills are more widespread than computer
pool of GPSF consultants, GCN had targeted general skills. Because mobile phones are practically ubiquitous in
practitioners (GPs) in the North who wished to consult with Ghana, there is a lower learning curve as opposed to
urban and overseas medical specialists. This matched the computers, which are owned by only one in three doctors. A
needs of district hospital doctors in the South who had no few doctors claimed that texting on a mobile phone was easier
ready access to specialist consultation and saw great value in than typing on a keyboard. While many of the junior doctors
the ability to connect with specialists (either in Ghana or demonstrated great proficiency with both typing and texting –
overseas). The popularity of this system in the district the vast majority sent multiple text messages a day – a few of
hospitals is a very promising precedent for future deployments the senior doctors demonstrated great difficulty typing during
in the more rural North. In contrast, physicians at the larger training sessions. Third, it promotes more tightly-integrated
Southern hospitals could speak and consult with other doctors workflows. One doctor said:
and specialists more easily in-person than over the computer. I use [ML] a lot and I think it is wonderful. If you want to
GCN did not address the needs of urban specialists looking to talk to anybody concerning a case… concerning anything
tap into the global community of sub-specialists. (In relating to your practice... it gives you a chance to relax
evaluation interviews, many of the urban doctors requested a and really talk. It’s so good. It’s a wonderful idea.
greater number of sub-specialists within the system.) Another said:
GCN Appropriation There has been a move to ban mobile phones in certain
Beyond the expected use for consultation, there were hospitals. It is a very very big mistake, because all they are
unanticipated uses of the system. First, 6 of the 39 going to realize is that this is actually going to decrease
‘conversations’ observed were purely social in nature. This
126
efficiency rather. Consultation is not going to be working While it is tempting to conclude that providing solutions
as well as it used to. with lower specificity is strictly more desirable – and in many
ML Appropriation cases it may be more desirable – it must also be remembered
that utility comes in many forms. GCN’s higher degree of
ML was originally framed as a system to ensure continuity specificity is due to multiple factors. First, the social
of care in long-term and referral patients, but after the system networking application is required to meet the GCN goal of
had been established for three months the ML investigator enabling isolated doctors to build social capital in an extended
noticed its emergent effect in fostering camaraderie in the geographic network of colleagues. While ML assumed that
medical community. He said: the doctors that needed to work together already had each
I was talking to doctors and they were telling me. “Yeah, other’s mobile phone numbers, GCN had determined that a
I’m reconnecting with classmates.” large portion of rural, immigrant, and junior doctors did not
Indeed, many doctors were up-front about the fact that ML have a strong network of contacts [20]. Second, GCN’s
had been a boon, not just for consultation, but for facilitating emphasis on overseas consultants implies a need for low-
social interaction within their community. bandwidth, asynchronous communication as opposed to voice
The mobile phone has also integrated itself into calls. Third, GCN operated as a development project,
administrative and management processes within large maintaining statistics of usage metrics in order to facilitate
hospitals. One doctor said: evaluation; for now, ML relies entirely on the built-in
reporting mechanisms in the OneTouch network. GCN can
You don’t have walk down somewhere or you send a
map consultations made to specific case outcomes, while
patient down... it reduces the whole bureaucratic...
pushing around of patients. OneTouch – which does receive very high praise from
enrolled doctors who offered their feedback on the system –
Personal mobile phones, both on the ML program and not, are cannot. Moreover, as previously mentioned, GCN’s goals
regularly used to set up diagnostic tests at the laboratory, to include technological innovation as well as social impact.
confirm insurance claims forms, and also for the doctor on call GCN assumes connectivity in rural areas will remain an
during the night to discuss a change in a patient’s care with the ongoing challenge, while ML relies on the assumption that,
doctor who admitted that patient. In short, where a U.S. with time, OneTouch coverage will be able to reach even the
hospital might use infrastructure such as pagers and site-wide most remote doctors. This question echoes one of the
internal communications systems, large Ghanaian hospitals fundamental tensions in ICTD research: as pragmatists, we
improvised solutions using personal mobile phones. aspire towards demonstrating the greatest social impact in real
C. Discussion communities today, while as researchers, we try to identify
In reflecting upon differences between GCN and ML in what fundamental limitations exist and how these can be
usage and appropriation, we will focus on an important tackled in years to come. It is to be expected that untested
distinction that we will call ‘specificity’. Two solutions with technology would experience more hurdles in the short run,
different ‘specificity’ can be targeted at exactly the same task while its long-term contribution is yet to be seen. In short,
and be based on an equally nuanced understanding of then, a project can easily have a number of goals that might be
workflows and use cases; the difference lies in the types and frustrated by a lower-specificity solution.
number of layers of technology which make up the solution A final note about lower-specificity solutions concerns their
and the degree to which they are specific to the solution. For potential to be too widely appropriated. In the smallest district
example, the GCN system presents not only a robust hospital (where GCN experienced greater adoption), both
asynchronous communication medium, but also an in-hospital doctors tended to switch off or mute their phones while at the
server as well as an end-user Web application. ML, in its first hospital in order to minimize distraction. Even in the larger
phase, focuses exclusively on tackling the cost barrier of urban hospitals, half of the 20 doctors who discussed their
existing phones using existing networks. mobile phone usage claimed to keep their phones off while
As we have seen, these two projects with similar goals at working because of distraction. Hospital administrators have
the start ultimately resulted in two very different usage related concerns; at one teaching hospital, a memorandum was
outcomes – particular in the area of adoption. Investing time circulated to all the doctors banning the use of mobile phones
and resources in needs assessments and design process, the in many locations
GCN project produced a highly ‘specific’ system to address …to forestall the negative impact of mobile phone
not only failures in internet infrastructure but also social frequencies on medical equipments and improve the work
network gaps. However, adoption has been slowed by the ethics of staff. It is also announced… that it is a serious
need to introduce the system incrementally into hospitals. The offence to disconnect life-supporting equipment in order to
ML project achieved broader adoption over a much shorter use their power sources (socket) to charge mobile phones.
period of time, in part by relying on the existing availability In view of concerns such as these, we suggest that researchers
and high adoption levels of mobile phones. Similarly, the keep in mind both the (immediate) benefits and (eventual)
lower specificity of the ML solution seems to contribute to a costs of rapid adoption.
greater range of user appropriation behaviors.
127
VII. RELATED WORK specific technologies such as connectivity and a post-1999
trend towards specific solutions such as software and VOIP.
ICTDs for Healthcare in the Developing World
The research on ICTDs to promote healthcare has a long VIII. CONCLUSION
and rich history [15],[34]. In the context of developing
regions, remote medical consultation has been a popular and In this paper, we have presented parallel case studies of two
relatively successful approach [32]. Computers [3], mobile ICT projects addressing the need for improved medical
phones [25] and a combination of both [27] have demonstrated consultation among doctors in Ghana. We have examined
utility in a variety of settings. how the two projects have been shaped by the institutional
context and the identity of their researchers. We looked at how
Framing the Problem the partnerships formed affect the solution outcome, delivery,
[21] provides a useful overview and categorization of how and adoption; we argued that even the task of assessing
ICTs are conceptualized from a variety of different fields. technological infrastructure is far from objective; and we
Theory on ‘framing the problem’ can be drawn from social- noted the implications of the ‘specificity’ of each solution.
constructivism, although we are not advocating here for a We conclude with an open question raised by these
change from ICTD’s traditional focus on pragmatism and discussions, summarized in Table 3. In ICTD research, much
advocacy/participatory research. Much of the work on attention has been given to the socio-cultural, political, and
qualitative methodology emphasizes the importance of economic contexts of target communities – yet ‘difference’ is
reflexivity and self-awareness in order to minimize such bias a measure between communities; it is only by critically
[8], but the fact remains there are institutional accountabilities examining our own research community that we can
which no amount of methodology can shake. We can learn understand the influence and impact of communities on each
from the example of anthropologists who have challenged other. ICTD is linked with economic, social, political, and
their own role in the colonial apparatus [11],[30]. For human development agendas [6]. Regardless of whether we
example, [17] provides an interesting comparison of how take an integrated approach [10] or focus instead on specific
different sectors tackle the management of kiosks in India. local needs [14], there are still institutional, societal, and
individual layers at play in our interventions. Researchers are
Partnerships and Deployment
embedded in contexts of existing friendships, collaborations,
ICTD researchers have an important role to play at the expertise and agendas, and we need to be conscious of what
crossroads of business administration and government policy, kind of consequences our decisions to draw on them have on
identifying how both contribute to the goal of development project outcomes. In this paper we have seen impacts on
and how their contributions interplay. Many examples in everything from research agendas to infrastructure
ICTD research and development literature study the role of assumptions, yet the literature on these avenues of choice
institutional players, although these papers are typically remain fragmented across a variety of other communities
evaluative rather than action-oriented [18],[23],[24]. The including CSCW and organizational behavior. ICTD
literature on partnership selection and cultivation in ICTD is researchers need to increase the exchange of ideas between
small but growing rapidly [26],[28],[31]. these communities. For example, in the course of establishing
Specificity GCN, the task of generating a viable ‘business model’ for the
technology providers was not addressed, but a social
In reaction to the variable success of many deployments of
entrepreneurship community would never allow such an
general ICT, such as a number of telecenter and ‘hole-in-the-
omission. ML has had great adoption success, but technical
wall’ computing initiatives, much of the focus of the HCI
infrastructure for evaluating its impact was neglected. We
ICTD research community has turned towards purpose-built
need a theoretical framework under which to unite the
technology [4],[33]. Many of those doing systems research
different fields of research for ICTD practitioners, one
focus instead on providing basic computing or connectivity
surrounding not just technology design, but also technology
functionality [13],[29]. [6] presents a useful survey of ICTD
framing, partnership, assumptions, and deployment.
literature which mentions a pre-1999 trend towards non-
during the two halves of the study, we analyzed the results as 160
effect of the tutor while the children did not attend school (and
20
0
so did not attend English class). This modified study design is S1 S2 S3
Tutor usage: Another significant difference between the two B. Methodology and Implementation
experiments is that tutor usage in experiment 2 was much The selection of the school for the pilot test was done after
lower than in experiment 1, due to a higher level of meeting with the headmasters and teachers at both schools and
absenteeism from the study during the school vacation. As assessing the state of the computer labs and the potential for a
Table VIII shows, this is particularly true for participation of successful study. Subsequently, the pilot study consisted of
the S3 students whose total usage per student in the second conducting interviews with teachers at the selected school,
half of the study dropped to under a third of what it was in the providing training for the teachers, and making qualitative
first half. This reduced usage may partly explain why there is observations of teachers and students as they used the Reading
no statistically significant effect of the Reading Tutor for S3 Tutor. We also explored with the teachers possible details of
children in experiment 2. Nineteen children who were absent a longer controlled study.
for testing either at the beginning or end of each experiment
C. Results and Observations
are not included in this summary.
Computing infrastructure: Both schools had a computer
Table VIII - Tutor usage in each experiment room with about 20 computers, most of which were 266 Hz
School Experiment 1 Experiment 2 Pentium II machines with 64-128MB of RAM and 10GB hard
Number of days of tutor S1 29.6 (12.0) 22.2 (9.4) drives. Some teachers from each school had taken a computer
use per student S2 36.9 (2.4) 24.0 (11.7)
S3 30.7 (2.8) 12.6 (3.9)
skills training course when the computers were initially
Total time spent reading S1 12.4 (5.4) 11.9 (5.2) donated, but had not had the opportunity to reinforce their
with tutor per student S2 18.6 (3.1) 11.5 (6.7) skills through additional training or guided practice. The
(hours) S3 17.3 (2.7) 6.8 (2.5) schools did not have internet access and the computers were
Average daily time S1 22.9 (7.1) 32.6 (5.3) being used primarily for tasks such as typing exam questions,
spent reading with tutor S2 30.1 (3.7) 26.6 (7.0)
per student (hours)
rather than as educational tools for students or for accessing
S3 33.9 (4.6) 31.9 (5.3)
online teaching resources. Given the large class sizes—which
ranged from 50 to over 100 students—and strict timetable, the
V. ZAMBIA PILOT STUDY
teachers also faced challenges in feasibly using the one lab to
The work in Zambia complements the prior studies in Ghana teach their students from grades one to nine about computers.
by investigating an in-school usage model and testing the tutor Some computers at both schools were not functioning due to
in a different English-speaking African country. broken keyboards, mice, and power strips; some computers
ProjectEDUCATE, a non-profit organization supporting some were not being protected from dirt during the dusty winter
schools in Zambia, introduced us to two under-resourced months; power outages were a daily occurrence in Mongu;
public schools in Mongu, the capital of Zambia’s Western and for one school, even maintaining electricity for the
Province. The schools had received donated computers and computer lab was a challenge due to limited financial
were enthusiastic about the possibility of testing the Reading resources. For the pilot study, we selected the school with
Tutor. However, the Mongu District Education Board better maintained equipment and a higher likelihood of
Secretary’s office directed us to select only one school for the maintaining communication by telephone and email since the
pilot test. Three of the authors conducted the study in school had a telephone and the headmaster had a working
Zambia. email address that he accessed weekly.
A. Goal Training Teachers: We trained three teachers to use the
Reading Tutor. The hour-long session covered the educational
The goals of the Zambian pilot study were to answer the
features of the tutor as well as administrative tasks such as
following questions:
managing users. After basic instruction and some time to
1. Computing infrastructure: What is the state of the
practice on their own, the teachers were able to guide students
school’s computing lab, how is it currently used,
in using the tutor. They would often give feedback to the
and can it feasibly be used for sessions with a
children as a complement to the tutor when the students had
computer-based reading tutor?
136
difficulty reading stories. For example, they would instruct the
children to click for help when they needed it. The teachers Test instruments and Reading Tutor content: We discussed
would sometimes correct a mispronounced word that the tutor with the teachers other reading assessment options for a
did not detect. We think this involvement of teachers in the longer term study, in addition to the fluency test and Test of
early stages as the child gets used to the tutor could be an Written Spelling. We learned that through the Zambia Primary
important part of longer-term use of the Reading Tutor in this Reading Programme (PRP), the students’ literacy levels are
school setting. A similar role was played by project staff in the measured by reading standardized story books aloud to their
Ghana studies. The teachers especially appreciated the ability teachers. Each color-coded book is associated with a given
to track their students’ performance using the tutor, including reading level. Once the teacher determines that the child can
the number of new words the children had read and the successfully read at a given level, he/she moves the child up to
number of words they read per minute. a higher level reading group with a different set of books. The
Children’s response: We observed one group of eleven teachers were interested in engaging the students with the
children in grades two through four as they read one or two computers in a way that would support curricular
stories from the tutor. The students were selected by the requirements from the Ministry of Education, and this was
teachers and had varying levels of reading ability, English also emphasized to us by the Ministry of Education officials.
comprehension, and speaking fluency. As in Ghana, all but Accordingly, we discussed incorporating the standardized
one student were completely new to computers. We PRP reading material into the tutor, which will also have the
introduced them as a group to the basic components of the advantage of providing additional content choices for
computer and then gave them the opportunity to demonstrate students. We saw that many of the existing Reading Tutor
use of the mouse and keyboard to each other. We provided stories, such as those related to baseball, zoos, or recycling,
verbal instruction to students individually as they began had little relevance to children’s lives in Mongu.
reading stories from the Reading Tutor. Just as in Ghana, the Feasibility of long term study: The teachers suggested that a
speech recognition capability appeared to work adequately random subset of the children in grades 2 through 4 could be
with the Zambian students’ and teachers’ accents. Some Grade selected to participate in a controlled study. They explained
2 students seemed to have difficulty understanding English. that students in these grades have 60 minutes of reading class
This is because English is introduced as a language of per day and suggested that the intervention group of students
instruction only in Grade 2, as a part of the Zambian Ministry could spend 30 minutes of that time working with the Reading
of Education’s initiative to encourage basic literacy by Tutor, with teacher supervision, while the control group of
teaching in a familiar language in the first year of school [21]. students would remain in class. We anticipate several
The students enjoyed using the Reading Tutor; when we challenges in conducting this project as a remote field study.
Although we have email and telephone contact information
returned to the school on another day to meet with the
for the key collaborators at the school, we expect
teachers, the group of students was using the word building
communication to be difficult given their limited internet
exercises in the Reading Tutor on their own time.
access and the frequent disruptions in telephone
communications due to daily power cuts. Success would
depend on the teachers taking ownership of the project as a
result of their enthusiasm. We hope that communication with
local contacts will help address some of the expected
challenges. For example, we plan to involve the Mongu
District Education Board Secretary’s office in evaluating the
project’s progress at the school. We also engaged a local
ProjectEDUCATE technical contact who joined us at our
meetings with the schools, received training from a member of
our team, and assisted us with installation and use of the
Reading Tutor software. We hope that he will stay in touch
with us remotely, help incorporate new stories into the tutor,
Figure 3 –Teachers and a child test the Reading Tutor in Mongu check in periodically with the school, and be available to the
teachers to assist with any technical challenges.
L ITERACY is one of the great challenges in developing developing world, and an increasing fraction of these phones
regions. Despite huge improvements in recent decades, feature multimedia capabilities for gaming and photos. These
literacy levels in many poor countries remain shockingly low. devices are a promising vehicle for out-of-school learning to
Even more challenging is the tension between regional and complement formal schooling. In particular, we believe that
global “power” languages, that economic opportunities are ESL learning games on cellphones present an opportunity to
often closed to those literate only in a regional language. For dramatically expand the reach of English learning, by making
instance, India is a country with 22 regional and 2 national it possible to acquire ESL in out-of-school settings that can be
languages, i.e. Hindi and English. But English, together with more convenient than school. Games can make learning more
computer skills, are the two most requested skills in surveys of engaging while incorporating good educational principles [6].
poor parents [1]. English is a great economic enabler. It is the More important, a large-scale evaluation with urban slums
children in India has shown significant learning benefits from
games that target mathematics [7]. We believe that similar
Manuscript received September 22, 2008. This work was supported in part outcomes can be replicated with e-learning games that target
by the U.S. National Science Foundation under Grant 0326582, a Qualcomm
BREW Wireless Reach award, and sponsorship-in-kind from Sony Creative
literacy.
Software. The challenge in evaluating any language learning project,
Matthew Kam is an Assistant Professor in Carnegie Mellon University’s however, is that language acquisition is a long-term process
Human-Computer Interaction Institute (phone: +1 412-268-9805; fax: +1 412-
268-1266; email: mattkam@cs.cmu.edu). He was with the Berkeley Institute
on the learner’s part. Worse, with a novel technology solution
of Design and the Department of Electrical Engineering and Computer that has yet to be institutionalized, there were tremendous
Sciences at the University of California, Berkeley when this work was carried logistical obstacles in running a pilot study over a non-trivial
out. John Canny is the founding director of BID and holds the Paul and Stacy duration. After 3 years, in which we commenced with needs
Jacobs Distinguished Professorship of Engineering at UC Berkeley.
Shirley Jain participated in the pilot study as a curriculum developer and assessments and feasibility studies, followed by subsequent
local supervisor. rounds of field testing interleaved with numerous iterations on
Anuj Kumar and Akhil Mathur are with the Dhirubhai Ambani Institute of our technology designs, we have established the necessary
Information and Communication Technology, Gujarat, India.
relationships with local partners for such an evaluation. This
140
paper describes the results from a semester-long pilot study – conducting a learning assessment for a novel educational
the longest so far in this project – which took place during the technology over a substantial timeframe when it is not yet
project’s 4th year. The study involved 27 rural children who integrated into the everyday operations of a formal entity.
participated in an after-school program that we implemented
in their village.
III. OVERVIEW OF PILOT STUDY
The pilot study was carried out in collaboration with a non-
II. RELATED WORK government organization in North India under the terms of a
Most work on technology-assisted language learning in the Memorandum of Understanding. The study took the form of
developing world does not explore the convenience that the an after-school program, which we held during the afternoons
cellphone’s mobility offers. Banerjee et al. [7] report a large- at a private village school affiliated with this NGO. However,
scale evaluation with mathematics learning games on desktop our goal was to investigate learning impacts that ESL learning
computers, carried out over 2 years with urban slums children games on cellphones have on lower-income rural children. As
in India. Mitra et al. [8] describe a study in India with slums such, students who were already enrolled in this school were
children over 5 months, which involved a “hole-in-the-wall” ineligible to participate in the study. Instead, we invited those
public computer installed with speech-to-text software. Dias et parents who could not afford the fees for this private school –
al. describe a computer-based tutor software for improving and hence sent their children to less expensive schools in the
reading in Ghana [9]. Kothari’s karaoke-like approach [10] same area – to give consent for their children to participate.
targets native language – not second language – literacy using In the after-school program, we ran three sessions per week,
television. on average. Each session lasted two hours in the afternoon.
With the cellphone’s increasing ubiquity in Africa, Brown Children from neighboring villages attended the after-school
[11] argues that it is timely to envision a future where the sessions after finishing their regular classes in the morning. In
cellphone plays a pivotal role in education in Africa. Kam et the after-school sessions, we loaned cellphones preloaded with
al. [12] describe how a set of ESL-learning cellphone games ESL learning games to participants. The after-school program
that targets children in rural India have undergone numerous took place from late December 2007 to early April 2008, and
iterations, based on successive, short-term formative studies. spanned sessions on 38 days in total.
With the exception of Kam et al., the only education-related
projects we know of in the developing world that leverage the
cellphone are Islam et al. [13] in Bangladesh, and Librero et IV. DATA COLLECTION
al. [14] in Mongolia and the Philippines. Both projects rely on As our preparation for this pilot study, we made two trips to
Short Messaging Service and target university students, unlike India, i.e. once in the summer of 2007 to familiarize ourselves
our approach. with the pilot location and end-user community, and a second
To date, Horowitz et al. [15] is the only study we know of time in December 2007 to kick-off the actual pilot. 4 local
that examines the cellphone for promoting literacy. However, staff members were hired to run the after-school sessions on
this study took place in an industrialized country (USA), even an everyday basis. 3 of them were engineering undergraduates
though participants included households below the poverty in their last semester, while the last member had graduated a
line. In the study, Sesame Street videos that target the English few years ago. We spent two weeks training them to run the
alphabet were streamed to preschool children over cellphones after-school sessions and perform data collection, and
throughout an 8-week period. Our paper therefore contributes continued to coordinate with them regularly via conference
to the literature as the first learning assessment on cellphone- calls and emails after we left India.
based language instruction in developing regions. We interviewed participants on their demographics such as
He et al. [16] describe a 2-year randomized evaluation of a their ages and the grades they were currently enrolled in in
LeapPad-like device that supports custom software modules school. During the interviews, we also asked other questions,
for English learning. This interactive system involves a paper such as the number of cellphones that their households owned,
book attached to a stylus and supports audio output. It lacks a what they currently and/or had previously used cellphones for,
visual display, unlike a cellphone, but overlaps with our goal their television watching habits and frequency, as well as their
of making literacy learning more accessible in the developing parents’ occupations. The questions on media and technology
world using portable devices. exposure were included because these variables were expected
Among the learning technologies for developing countries, to impact participant ability to learn using cellphone games.
one of the most novel devices is the multiple-mice computer To ensure that each participant has the basic numeracy and
described in Pawar et al. [17]. It was intended for collocated ESL literacy to benefit from cellphone-based learning in the
learning by a group of children around each computer, with a program, participants were required to pass a qualifying test,
mouse input device for each child. It has since been extended i.e. obtain at least 50% of the total score. The test required
to distance learning in Moraveji et al. [18]. The evaluations in them to complete one-word blanks using English words about
both papers were short-term, and underscored the difficulty in themselves, e.g. name, age, school, grade, etc. They were also
141
asked to fill in the missing letters in the alphabetic sequence, V. PARTICIPANTS
write numbers in the Arabic notation, match words with their Owing to the strong relationships that our NGO partner had
pictures, spell the words for everyday objects, and describe a built with the local community over more than a decade, we
picture of a market scene with short sentences. The qualifying were able to generate a high level of support among parents in
test was designed such that an average child in India with no this community. In total, we obtained consent for 47 children
learning disabilities who has finished 1st grade in a reputable to participate in the study. However, we needed to turn 16 of
urban school should obtain a perfect score on it. them away; 15 children did not pass the qualifying test while
By using the qualifying test as a screener, we ensured that the 16th was attending private tuition for English. The latter
participants were numerate. This was important because we represented a confounding variable. Of the 31 children whom
had previously found it difficult to teach children to use the we started the pilot with, 4 of them left the program mid-way.
cellphone keypad’s to play e-learning games when they were Reasons for attrition include time conflicts with private tuition
not familiar with the numbers from 0 to 9 in Arabic notation. (2 children) and disinterest in attending the sessions (another 2
Similarly, by ensuring that participants were familiar with the children). From post-deployment interviews, we understand
English alphabet, we could target a more advanced syllabus the latter was due to caste tensions between those 2 children,
that went beyond the alphabet. We made this decision since who belonged to the lower castes, and some upper-caste
Horowitz et al. [15] had already investigated the efficacy of children in the program.
cellphone-based learning for the English letters in the context
of preschool children in the USA. A. Demographics
Since success in acquiring a second language is correlated The 27 children who participated in the study until it ended
with literacy in one’s native language, we administered a test were aged 7 to 14 (mean = 11½ years) and belonged to grades
which evaluated the ability of the participants to read in Hindi. 2 to 9 (mean = 6th grade). There were 11 boys and 16 girls. 5
Every child was given a short passage that described a diet for children came from the upper castes while others belonged to
promoting dental health. Each child was then asked to read the the lower castes. The gender and caste breakdown seemed to
passage aloud so that we could observe his or her fluency and mirror the demographics in the community. Every participant
accuracy. These sessions were videotaped. After that, every attended between 8 and 29 sessions (mean = 20) in the after-
participant was asked to write answers to written questions school program, broken down according to the following three
that tested his or her comprehension of the passage, in Hindi. functions:
We had designed this test such that an average child who has • Cellphone training: 0 to 5 sessions (mean = 4) where we
finished 3rd grade in a reputable urban school in India should taught participants how to use the cellphones, perform
obtain a perfect score on it. alphanumeric input and play mobile games,
Our primary method of assessment was to administer pre- • ESL learning: 4 to 17 sessions (mean = 10) in which
and post-tests which evaluated participants on their ability to participants played ESL learning games on the cellphones,
spell the common nouns that the curriculum for the pilot study and
targeted. Although the curriculum targeted other competencies • Assessment: 4 to 7 sessions (mean = 6) for administrative
such as listening comprehension and the recognition of written tasks and data collection, e.g. demographics interviews and
words, our assessment emphasized spelling, which as a recall various tests.
task was cognitively more difficult than recognition tasks.
We maintained attendance records for the participants for In India, traditionally, only the upper castes owned land. As
every session. We also videotaped each session so as to have such, the upper castes earn their livelihood on the land or run
contextual data that could potentially account for their test small businesses, while lower castes graze their goats, work as
performances. The video recordings captured the classroom daily-wage laborers or perform menial jobs in the homes of
proceedings, and individual participants’ interactions with the the upper castes. Land-owning and non-land-owning families
games. The latter recordings captured participants’ levels of told us that they earned up to 100,000 (US$2,500) and 50,000
engagement with the games as shown in their facial and body rupees (US$1,250) respectively per year.
expressions. The pilot staff member who was responsible for B. Hindi and English Baseline
videotaping the sessions tried to ensure that every participant 26 of the 27 participants were enrolled in the same school,
was videotaped playing at least one level in the curriculum per where Hindi is the medium of instruction. The last participant
day. The recordings were later transcribed and translated from was a school dropout. Assuming regular school attendance,
Hindi to English. the typical participant would have taken classes on Hindi and
Finally, for every session, we asked pilot personnel to write English for 5½ and 3½ years respectively prior to the study.
a report which summarized what happened in that session, as We devised a grading rubric to evaluate each participant on
well as how well each participant interacted with the games. the Hindi literacy test and qualifying test. On the former test,
The latter not only covered usability and learning obstacles, participants scored 7.9 out of 18 on average (σ = 4.5, n = 19).
but also included pilot staff’s observations on the attitude and 2 participants turned in blank answer sheets. We observed the
persistence that each child demonstrated towards learning. following problems in the submissions:
142
• Wrong answers due to poor comprehension of the questions participants came from families that owned cellphones with a
(5% of the test-takers) or passage (53%), or responses that color screen, as opposed to monochrome display. Two of the
simply repeated the questions (32%) above cellphones – both of which belonged to 2 of the above
• Spelling errors (16%) 3 boys – contained built-in cameras.
• Grammatical errors, i.e. using the incorrect form of the verb Nonetheless, cellphone ownership and access were separate
for the subject’s gender (21%), or the incorrect form of the issues. Among the 25 children whose families owned at least
noun for the subject’s singularity vs. plurality (5%) one phone, 6 of them – 5 girls aged 7-11 and a boy aged 12 –
• Inability to phrase responses in complete sentences (5%) were prohibited by their parents from using the phone, either
entirely or most of the time. In general, in poorer families, it
TABLE I
BREAKDOWN OF PARTICIPANT PERFORMANCE ON QUALIFYING TEST
seemed that children were allowed to receive (free) incoming
Poor Fair Good calls, but not play mobile games lest they drop the devices. On
About 13% left blanks 32% filled in blanks 55% filled in blanks the other hand, in wealthier families, children were allowed to
myself empty or filled them with at least 1 with correct
in Hindi misspelling spellings for most
play on the phones. As such, although most participants were
questions familiar with cellphones, it appeared that rural parents were
Alphabet 5% filled in less than 13% filled in ~11 82% filled in at least
7 blanks in blanks in alphabetic 12 out of 13 blanks
more willing to entrust these relatively costly devices to their
alphabetic sequence sequence correctly in alphabetic sons (vs. daughters). In total, 15 of the 27 participants (56%)
correctly sequence correctly
14% matched up to 2 14% matched 4 out 72% matched at least
reported that they had played cellphone games before.
Word
recognition out of 6 words with of 6 words with 5 out of 6 words
correct pictures correct pictures with correct pictures
Spelling 41% spelt up to 2 18% spelt about 3 41% spelt at least 4
out of 6 words out of 6 words out of 6 words VI. CURRICULUM AND GAME DESIGN
correctly correctly correctly
Picture 18% made no 50% wrote answers 32% wrote One of the major challenges with carrying out a pilot study
description attempt to answer as individual words, intelligible answers over a non-trivial timeframe was that we needed to develop
this section not full sentences in sentences
sufficient digital content that could last throughout its entire
duration. We ensured that our syllabus was aligned with local
On the qualifying test, on average, participants scored 44.0
ESL learning needs in India by recruiting a local ESL teacher
out of 50 (σ = 5.5, n = 22). Our grading rubric indicated how
as our curriculum developer. She had a decade’s experience as
participants should be classified as “poor,” “fair” and “good”
an ESL teacher at a prestigious urban school, located in the
on every section of the test. Table I gives the breakdown of
same geographic region as the after-school program.
how test-takers were distributed across categories for selected
sections, and descriptions of the categories. In summary, the A. Curriculum Design
average participant had a good knowledge of the alphabet and Given the above attendance rate, the ESL curriculum for the
a fair vocabulary of written words that she could read. On the pilot was designed to be comparable to the amount of material
other hand, she was weak in recalling and spelling everyday that a qualified teacher could reasonably cover in 18 hours
nouns, and even weaker in constructing complete sentences with rural children in a classroom. The syllabus was situated
with these words. Despite the wide range in the ages of the within the classroom theme, which participants could readily
participants, it appeared from their performance on the above relate to. Concretely, the syllabus included:
tests that the variation in their English proficiency was much • Common nouns that are found in the typical classroom, e.g.
narrower. More specifically, we estimated that the average chair, table, door.
participant was comparable to an urban child in India who had • Verbs that can be performed with the above nouns, e.g. sit,
taken between 1 and 2 years of English classes. write, open, close.
Notably, only 10 children (45%) could spell their names • Sentence structures for constructing sentences out of the
correctly in English on the qualifying test papers. We had a above nouns and verbs, e.g. “This is a __.”
chance later to interview the teacher who taught them English • Sentence structures for phrasing question-and-answer
in their school. She revealed that her pedagogical approach sequences with the above nouns and verbs, e.g. “What is
revolved around having students copy sentences from English this?”, “Where is the __?”
textbooks into their notebooks. She felt that it was not worth
The curriculum design took participant performance on the
putting in more effort to teach English since she believed she
qualifying test into account. The curriculum was also based on
was underpaid.
our attempts to converse informally with participants, during
C. Technology Baseline which we learned they did not comprehend simple questions
Among the 27 participants, 25 of them came from families about themselves, did not know the English words for objects
who owned at least one cellphone; 5 participants belonged to around them (e.g. in the classroom), and made grammatical
families that owned 2 phones each while 2 participants came errors. The curriculum therefore targeted the above syllabus in
from families which owned 3 phones each. The cellphone was terms of listening comprehension, word recognition (of the
usually used by the eldest male member in every family, and written word), sentence construction and spelling.
in fact, 3 boys aged 13-14 possessed their own cellphones. 8
143
B. Game Design
We designed a set of ESL learning games for the cellphone
platform that targeted the above curriculum, and piloted them
in the after-school program. Our designs drew on 3 resources,
namely:
1) recurring patterns in state-of-the-art commercial software
applications for language learning, which represented best
practices that we reused to avoid reinventing the wheel,
2) traditional village games, which more closely matched the
expectations and understandings that rural children have
about games, in comparison to contemporary videogames Fig. 3 and 4. Figures 3 and 4 introduce additional phrases that the nouns and
that were largely Westernized, and verbs in the syllabus can be used in. Figure 3 teaches a phrase that associates
3) lessons from several previous rounds of field-testing and the verb “sit” with the noun “chair.” Figure 4 shows how to ask questions
using the “Where” keyword. Abstract phrases and function words such as
iterations with rural children elsewhere in India [12].
“where” are difficult to convey graphically. Hence, when they are taught for
In this subsection, we walk the reader through a subset of the the first time, the software explains their meanings orally in Hindi.
screen designs.
The games tested players on their comprehension and recall
of the words and phrases. For example, the game shown in
Figure 5 says the word aloud for one of the objects displayed
on screen. The player needs to identify the correct object and
push it onto the area that is blinking blue. At the same time, he
needs to avoid the balls thrown by the computer-controlled
opponent. This game was an adaptation of Giti Phod, which
was one of the traditional games that children play in Indian
villages. In Giti Phod, players in a team have to arrange some
objects (e.g. rocks) into a given configuration (e.g. a heap),
while avoiding being hit by a ball thrown by members in the
opposing team. In our experience, we have observed that rural
Fig. 1 and 2. Figure 1 introduces the English vocabulary for common nouns
in the classroom. Word-picture association is a technique employed by many children found it more intuitive to understand videogame rules
successful commercial language learning software. As the boy moves to each when the designs of these videogames drew on the rules found
object, the software highlights the corresponding word in a different (green) in the traditional village games that they play everyday.
color and says the word aloud. Figure 2 situates these objects in a classroom
scene and builds on the earlier screen by demonstrating how to use the nouns
in complete sentences. As the boy moves to each object, the software says the
“This is a __” phase aloud for the corresponding object.
Fig. 9. The second level of hints in the spelling activity. For every blank, a set
of possible letters are displayed to provide the player with some assistance, if
Fig. 6 and 7. Localized characters from the Indian production of Sesame he was unable to spell the word correctly after two attempts.
Street tested the player on his ability to engage in question-and-answer style
dialogues, and performed a victory dance for the player upon successful
completion of each level in the game. The curriculum is broken up into a total of 6 levels in the
games. On every screen, the player can access a menu through
The activity that targets spelling skills is shown in Figure 8. a shortcut button. Among various options, this menu permits
The player is given an image (e.g. blackboard) and is required him to move to an earlier level in the curriculum to repeat the
to spell the word corresponding to it. Some of the letters in the material, as well as to move to higher levels in the curriculum.
word are displayed, while blanks are shown for the remaining The software was designed so as not to require airtime, which
letters. The player moves between blanks with the arrow keys. was expensive for most rural families. We implemented the
Once he has filled in all the blanks and submitted his answer, games on Adobe’s Flash Lite and Qualcomm’s BREW
the correct and wrong letters are displayed in green and red (Binary Runtime Environment for Wireless) platforms. We
respectively. This feedback constitutes the first level of hints piloted the games on Motorola’s Razr V3m cellphone model,
that we have designed to help the learner arrive at the correct which has a fairly large screen.
spelling. If the player spells the word correctly, he proceeds to
the next game. Otherwise, all blanks are cleared after a short
pause and he is required to spell the same word again. VII. PILOT SESSIONS
In those sessions where participants were taught how to use
the cellphone, they were shown how to move their sprites with
the arrow buttons. They were also taught how to perform
alphanumeric text entry, since most of them did not know this.
Sprite movement and text input were essential skills for the
games we designed. Pilot staff were therefore asked to write
some simple, short sentences on the blackboard, and ensure
that each participant demonstrated his ability to enter those
sentences via text input.
Some other sessions focused on administrative tasks, such
as the above tests and demographics interviews. We learned
that a few participants had difficulty reading a small subset of
the English alphabet despite having passed the qualifying test.
Fig. 8. The first level of hints in the spelling activity. After the player has We spent two sessions coaching them on those less-frequently
tried to spell the word by filling in the blanks with letters, the correct and encountered letters, so that they would be better prepared for
incorrect letters are shown in green and red respectively. the syllabus targeted in the pilot. Next, at least 8 children had
seen the localized Sesame Street characters on television, but
If the player is unsuccessful in spelling a word correctly
did not know their names. To help participants better relate to
after two attempts, the second level of hints (Figure 9) appears
the characters, so that our games would appeal to them even
to provide him with additional learning support. Based on the
more, we introduced the characters at the start of the semester.
blank that the cursor is currently located at, the game displays
We also screened 3 localized episodes on separate occasions.
a set of possible letters for him to narrow down the choice of
These episodes were chosen such that they were educational
candidate letters.
but did not target English learning. Each episode lasted ½
hour, and we observed that participants enjoyed the humorous
acts performed by the characters.
145
Fig. 10. In the after-school sessions, each participant was loaned a cellphone Fig. 11. Frequency histogram of participant scores on the pre- and post-tests.
preloaded with English language learning games. Participants were taught
how to start the games, and were asked to focus on learning English when
playing the games on their own.
The average post-test gains was 3.4 out of 18 (σ = 3.3, n =
24). The gains exhibited a fairly large variation, and ranged
The remaining, and majority of, sessions focused on ESL from -2 (two participants exhibited negative gains) to 9 out of
learning. A 2-hour session was typically structured as follow: 18. We present the frequency histogram for post-test gains in
after an exchange of greetings, pilot staff took attendance and Figure 12.
briefed participants on the learning objectives for that day. If
new games were deployed that day, pilot staff explained and
demonstrated how to play them to the participants in small
groups. Each participant was then handed a cellphone to play
the games on her own (Figure 10), and were told to focus on
learning the English syllabus that the games covered. Children
who were absent on previous sessions received help from pilot
personnel in learning how to play those games that they were
unfamiliar with. Pilot staff were limited to providing technical
support; and were explicitly instructed not to teach English or
communicate with participants in English. There was a short
break of 10 to 15 minutes in the middle of each session. At the Fig. 12. Frequency histogram of participant post-test gains.
end of each session, pilot staff took back the phones so that
they could charge their batteries overnight and download new B. High-Gains vs. Low-Gains Learners
games onto them. Each participants received a small packet of
We sought to understand how participants’ post-test gains
biscuits after every session.
were correlated with their demographics and performance on
other tests. We also categorized participants into two groups,
namely, “high-gains learners” and “low-gains learners,” based
VIII. QUANTITATIVE RESULTS
on their post-test gains. A learner whose post-test gains
On the pre- and post-tests, test-takers were awarded 1 point exceeded the mean of 3.4 was categorized as a “high-gains
for each common noun in the syllabus that was spelt correctly. learner,” else he was classified as a “low-gains learner.” In all,
A. Post-Test Gains 9 participants were classified as high-gains learners while 15
The mean pre-test score was 5.2 out of 18 (σ = 3.3, n = 27) participants were categorized as low-gains learners. 3 of the
while the mean post-test score was 8.4 out of 18 (σ = 5.5, n = 27 participants could not be classified since they were absent
24). Participants exhibited significant post-test gains on a one- on the day when the post-test was administered.
tailed t-test (p = 0.007). We present the frequency histograms On a normalized scale, when the 27 participants were taken
for both scores in Figure 11. They illustrate that the score as one group, the average pre-test score was 29% whereas the
distribution had shifted toward the higher end of the spectrum average post-test score was 47%. The latter score did not seem
after the deployment. high in absolute terms, i.e. on average, a participant could not
spell over half of the common nouns targeted in the syllabus
by the end of the intervention. However, once the participants
had been classified, on a normalized scale, high-gains learners
scored 41% (80%) on the pre-test (post-test) whereas low-
gains learners scored 19% (27%) on the pre-test (post-test), on
average. In other words, high-gains learners not only showed
larger post-test gains but also appeared to have a higher mean
146
pre-test score at the start of the intervention. More important, The proportions in Table III were presented for the sake of
post-test gains for both high-gains (p < 0.001) and low-gains completeness. We were unable to perform any statistical tests
learners (p = 0.076) were significant, with effect sizes of 0.54 on these proportions due to the small sample size, which for
and 2.24 respectively. That is, both categories of participants example did not satisfy the standard binomial requirement.
exhibited learning gains. (But average post-test gains for low- We thus caution the reader against drawing firm conclusions
gains learners were only marginally significant, because of the from these statistics. However, when examining individual
2 participants who obtained lower scores on the post-test learners to identify surprising cases, we took the demographic
compared to their pre-test.) variables in Table III into consideration. The analysis is
deferred to the following section.
TABLE II
HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF DEMOGRAPHICS TABLE IV
Age Grade Days Days HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF TEST SCORES
(Years) Enrolled Spent Spent Qualify- Qualify- Hindi Pre- Post-
in School Learning Playing ing Test ing Test, Literacy Test Test
ESL ESL (out of Spelling Test (out (out
Games Games 50) Section (out of of 18) of 18)
Low- Mean 10.7 5th 3.9 10.5 (out of 6 18)
gains σ 1.8 1.9 1.4 4.2 words)*
learners Min. 7 2nd 0 5 Low- Mean 42.9 1.2 6.3 3.5 4.8
(n=15) gains 2.9 1.1 4.2 2.3 2.4
Max. 14 9th 5 16 σ
High- Mean 12.8 8th 3.8 10.2 learners Min. 37 0 0 0 2
gains σ 1.2 0.9 0.8 5.2 (n=15) Max. 46.5 4 14 10 12
learners Min. 11 7th 2 4 High- Mean 47.1 3.4 12.0 7.4 14.4
(n=9) gains 1.8 1.6 1.7 3.1 3.6
Max. 14 9th 5 17 σ
Is difference Yes Yes No No learners Min. 43.5 2 10.5 2 6
between means (p = 0.002) (p < 0.001) (p = 0.4) (p = 0.4) (n=9) Max. 49 6 14 13 18
significant? Is difference Yes Yes Yes Yes Yes
Correlation with 0.45 0.61 0.11 0.10 significant? (p < 0.001) (p = 0.001) (p < 0.001) (p = (p <
post-test gains (r) 0.003) 0.001)
Correlation with 0.57 0.70 0.45 0.46 0.86
post-test gains (r)
In Tables II and III, we examined how high-gains learners *In this column, we present the number of words that participants spelt
may differ from low-gains learners in terms of demographics. correctly on the spelling section of the qualifying test, out of a total of 6
On the whole, the high-gains learners did not appear to differ words.
significantly from low-gains learners in terms of the number
of days that they spent on learning how to play the cellphone- Table IV compares the high-gains and low-gains learners in
based games (p = 0.4) or actually playing the games to learn terms of test scores. The former outperformed the latter on the
ESL (p = 0.4). Instead, high-gains learners belonged to higher Hindi test (p < 0.001). Next, we analyzed the qualifying test
ages (p = 0.002) and were enrolled in more advanced grades results at two levels, namely, the score for the entire test as
in school (p < 0.001). In fact, post-test gains exhibited high well as the score on the spelling section. We found that high-
correlation with grade levels that participants were enrolled in gains learners outperformed low-gains learners on the entire
school (r = 0.61) and medium correlation with age (r = 0.45). test (p < 0.001) as well as on the spelling section (p = 0.001).
High-gains learners also obtained higher scores on the pre-test
(p = 0.003) and post-test (p < 0.001), vis-à-vis low-gains
TABLE III learners. In fact, participants’ post-test gains exhibited a high
HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF DEMOGRAPHICS degree of correlation with their qualifying test scores, for both
Sex Caste Media Exposure Attitude* the entire test (r = 0.57) and spelling section (r = 0.70). On the
Low- 67% 77% 73% have played 24%, 38% & 38%
gains (33%) (23%) games on cellphones were described as other hand, post-test gains had a lower correlation with Hindi
learners were belonged prior to pilot; below average, literacy levels (r = 0.45) and pre-test scores (r = 0.46).
females to lower 60% (40%) watched average and
(n=15) (males) (upper) less (equal to or more) above average
castes than 1 hour of TV per learners IX. QUALITATIVE RESULTS
day respectively
High- 44% 88% 56% have played 29% and 71% The above quantitative results suggested that current levels
gains (56%) (12%) games on cellphones were described as
were belonged prior to pilot; below average
of spelling proficiency and grades enrolled in school were the
learners
(n=9)
females to lower 40% (60%) watched and above strongest predictors of success in learning how to spell new
(males) (upper) less (equal to or more) average learners
castes than 1 hour of TV per respectively
words through the cellphone-based games which we designed.
day Higher levels of Hindi literacy and academic preparation were
*The Attitude column is based on the observations that pilot personnel also associated with higher post-test gains.
have on the seriousness and aptitude that participants exhibited as learners
On the other hand, the number of sessions that participants
throughout the pilot. These qualitative comments were subsequently coded
into the “below average”, “average” and “above average” learner categories. had with the cellphone games – both for learning how to play
the ESL learning games and learning ESL through the games
147
– were not associated with post-test achievements. Among the on their first attempt in the spelling activity), or with only the
24 children whom we have post-test gains data on, 5 of them first level of hints. It seemed that they did not require much
were classified as high-gains learners despite having played scaffolding support from the software. In fact, from the video
the games on only 4 to 7 days (mean = 6 days). Conversely, 7 recordings of 9 high-gains learners, we saw that 5 (56%) and
participants were classified as low-gains learners in spite of 1 (11%) of them depended on the first and second levels of
having played the e-learning games for 13 to 16 days (mean = hints respectively. In contrast, 12 (80%) and 8 (53%) out of
14.7 days). More important, pilot personnel described 6 of the 15 low-gains learners who were videotaped relied on the
these 7 low-gains learners as “hardworking” or “serious” first and second levels of hints respectively. It seemed that the
about learning ESL. Similarly, we were curious about how the low-gains learners, as compared to the high-gains learners,
two students who exhibited negative post-test gains, as well as were less able to rectify their errors in filling in the blanks for
the school dropout, had interacted with the cellphone-based the missing letters through the first level of hints, and required
games. the second level of hints to attain the correct spellings.
The above quantitative trends raise the following questions: Worse, the inability on the part of the low-gains learners to
How did some of the high-gains learners played the games spell correctly with help from only the first level of hints made
such that they benefited despite lower attendance? In the case some of them visibly unhappy or bored when the second level
of some low-gains learners, why did they improve little on the of hints appeared. The reason for this distress was unclear.
post-test despite spending numerous days with the games and The learner could be frustrated that he was spending too much
being perceived as diligent? In order to address such questions time to learn how to spell the word. Alternatively, on seeing
at the interaction design level, we turned to our video records the second level of hints show up, he could be demoralized
and daily reports. Our hope was to recommend improvements that he had just been relegated to the ranks of the most inferior
to the technology designs and/or after-school setting. learners and needed the second level as a “crutch” in order to
succeed.
A. Interaction Patterns with the Technology
Furthermore, some learners struggled despite both levels of
At first glance, it seemed that participants needed to attend hints. In the videos, two of them turned to their neighbors and
the after-school program for more days. Our video recordings asked for the correct letters, and/or to chat. In some cases,
showed that only 3 of the participants reached the last level in participants were embarrassed to ask their neighbors for help
the curriculum by the last session in the program. This was a again after so soon, and hence used the menu to skip to other
surprise. Given that the curriculum was designed for 18 hours levels in the games.
of instruction, we expected an average attendance rate of 10 More important, we observed that participants – especially
gameplay sessions to constitute enough time with the games. among the low-gains learners – may be able to spell the words
On examining the video recordings more closely, we saw that in the spelling activity, but were not able to spell the same
at least 8 participants were using the game menu to skip ahead words on the post-test. We offer two plausible explanations.
to other levels whenever they were unable to spell the words Firstly, some children may have learned to spell the words by
in the current level correctly after a few attempts. (We note their last session in the program, but had forgotten their
that the menu was not necessarily a negative feature. Among spellings between the last session and the post-test. Secondly,
those 8 participants, at least 2 of them used the menu to skip some participants never learned to spell the words in their
those words that they already knew how to spell.) entirety, since the spelling activity only involved filling a few
We needed to understand why learners gave up on retrying blanks and did not require the learner to spell the entire word.
the spelling activity for difficult words despite the hints in the Nowhere in the video recordings did we observe any child
spelling activity. On the whole, we observed 4 different levels struggling with usability problems.
of behavior associated with the spelling activity in the videos:
1) When students encountered a word that they could spell,
they pressed the keypad buttons quickly and with ease to X. CONCLUSION
fill in the blanks for the missing letters.
Our reactions to the results of the learning assessment were
2) When students saw a word that they did not know how to
mixed. In an underdeveloped region where rural children did
spell, some of them learned to spell it correctly with the
not have access to quality English instruction in their regular
help of the first level of hints.
school or elsewhere, we were excited to see the participants --
3) Some of those students who failed to learn how to spell a
both high-gains and low-gains learners – in the after-school
word with the first level of hint eventually learned how to
program exhibit statistically significant post-test gains that
spell it correctly with the help of the second level of hints.
could be reasonably attributed to our cellphone-based English
4) Other students never succeeded in learning how to spell
learning games. On the other hand, the learning benefits were
certain words despite both levels of hints.
uneven among participants. This could be a cause for concern.
In general, we observed that high-gains learners succeeded
To begin with, high-gains learners outperformed low-gains
in learning how to spell words after having seen their written
participants on the pre-test, qualifying test and Hindi literacy
forms displayed on earlier screens (i.e. such learners were able
test. In fact, participants’ post-test gains appeared to be highly
to spell those words correctly – without requiring any hints –
correlated with their existing levels of spelling proficiency (as
148
measured by their performance on the spelling section of the the after-school program is a model that is readily replicable,
qualifying test) and the grades in school that they are currently we encourage the reader to adopt and experiment with the
enrolled in. This observation suggests that those rural children lessons from this paper. Our results suggest that the cellphone
with a stronger academic foundation are the same children – which remains a relatively scarce resource in the developing
who are most well positioned to take advantage of the benefits world – is most effectively utilized in an after-school program
that cellphone-based learning confers. that targets more advanced children. This restriction may be a
Our results are consistent with the outcomes of a study with necessarily evil until we gain a deeper understanding of how
rural and urban low-income children in India described in He to design instructional scaffolds for less well-prepared rural
et al. [16]. This study showed that weaker students benefited children.
more from a teacher-directed pedagogical intervention, while Next, children’s tendency to seek help from their neighbors
stronger students benefited more from a self-paced, machine- can be channeled productively if the latter are taught to offer
based approach to English learning. These results should not, help appropriately (e.g. instead of only telling their neighbors
however, be interpreted to mean that we rule out technology- the correct spelling, help them to associate and remember the
augmented learning completely in the context of low-income correct spelling). Such peer coaching strategies are especially
children. Horowitz et al [15] reported a study on videos for crucial since cooperative group learning is unfamiliar to many
learning the English alphabet streamed over cellphones. In rural children, whose schools (if they attend one) are more
this study, a greater proportion of lower-income parents, vis-à- likely to implement rote learning. Alternatively, such an after-
vis their higher-income counterparts, perceived the videos to school program can hire facilitators to provide academically
have improved their children’s knowledge of the alphabet. less prepared learners with similar coaching.
In the face of the above overwhelming odds, what can we
do to promote more equitable educational opportunities in the ACKNOWLEDGMENT
developing world? One possible – and perhaps cautiously We thank Suraksha (Urvashi Sahni and Shalini Mathur) and
optimistic – interpretation of the above results is that future Sesame Workshop India for collaborating with us on this pilot
research needs to be directed at understanding how e-learning study. Lauren Bailey and Anuj Tewari provided much-needed
software can provide more scaffolding support for those rural administrative assistance, while Aman Anand and Siddhartha
children who have less academic preparation. As an example, Lal provided technical assistance. We are especially indebted
the spelling activity needs to be redesigned such that the to Mehnaaz Abidi, Jatin Chaudhary, Neelima Purwar, Gautam
learner is guided to spell the entire word eventually. With this Singh and Kavish Sinha for providing ground support. Lastly,
redesign, however, gameplay becomes prolonged and can we thank the parents and their children for their participation.
potentially increase player frustration, as we have witnessed
above. One remedy is to have e-learning games track learner REFERENCES
performance, so that the software can be adaptive in skipping [1] J. Pal, M.Lakshmanan, and K. Toyama, “’My child will be respected:’
stages that are similar to those that the player has previously Parental perspectives on computers in rural India,” in Proc. of 2nd
performed well in. Another implication for instructional IEEE/ACM International Conference on Information and
Communication Technologies and Development, Bangalore, India, Dec.
design, which calls for additional investigation, is scaffolds
2007.
such as hints that are less conspicuous, so that their appearing
[2] S. Shukla, “From pre-colonial to post-colonial: Educational transitions in
on screen does not diminish the learner’s sense of self-esteem southern Asia.” in Economic and Political Weekly, 31(22), 1996, pp.
or achievement. 1344-49.
Unfortunately, adaptive educational applications require the [3] D. Faust, and R. Nagar, “Politics of development in postcolonial India:
application state to be stored and retrieved on the same mobile English-medium education and social fracturing,” in Economic and
device. From the logistics standpoint, this requirement is more Political Weekly, India, July 28, 2001.
difficult to implement in developing regions since it is harder [4] M. P. Kishwar, “Deprivations’s real language,” in The Indian Express,
September 15, 2005. Available:
to ensure that the same learner uses the same phone – which http://www.indianexpress.com/printerFriendly/12662.html.
stores his performance from prior session(s) – across sessions. [5] Azim Premji Foundation, “The social context of elementary education in
For instance, in an after-school program where attendance rural India,” 2004. Available: http://www.azimpremjifoundation.org/
fluctuates from session to session, it would be prudent to keep downloads/TheSocialContextofElementaryEductaioninRuralIndia.pdf
a shared pool of cellphones, such that children who show up [6] J. P. Gee, What Video Games Have to Teach Us About Learning and
for the day’s session can draw from. In these circumstances in Literacy, Palgrave Macmillan, 2004.
which it is not possible to reserve a cellphone for each child, a [7] A. Banerjee, S. Cole, E. Duflo, and L. Lindon, “Remedying education:
Evidence from two randomized experiments in India,” NBER Working
wireless networked mechanism for synchronizing application
Paper No. 11904, December 2005.
state across all cellphones may be necessary. This, and other
[8] S. Mitra, J. Tooley, P. Inamdar, and P. Dixon, “Improving English
issues that we have raised above, require further investigation pronunciation: an automated instructional approach,” in Journal of
for cellphone-based literacy learning to be more effective in Information Technologies and International Development, Vol. 1, No. 1,
targeting less academically advanced rural children. Fall 2003, pp. 75-84.
As we think more widely beyond the cellphone to consider [9] M. B. Dias, G. A. Mills-Tettey, and J. Mertz, “The TechBridgeWorld
initiative: Broadening perspectives in computing technology education
it as a component in the broader learning environment, since
149
and research,” in Proc. of the international symposium on Women and
ICT: Creating Global Transformation, ACM Press, 2005.
[10] B. Kothari, A. Pandey, and A. R. Chudgar, “Reading out of the ‘idiot
box’: Same-language subtitling on television in India,” in Journal of
Information Technologies and International Development, Vol. 2, No. 1,
Fall 2004, pp. 23-44.
[11] T. H. Brown, “The role of m-learning in the future of e-learning in
Africa?,” in Proc. of 21st ICDE world conference, Jun. 2003.
[12] M. Kam, A. Agarwal, A. Kumar, S. Lal, A. Mathur, A. Tewari, and J.
Canny, “Designing e-learning games for rural children in India: A
format for balancing learning with fun,” in Proc. of ACM international
conference on Designing Interactive Systems, Cape Town, South Africa,
Feb. 2008.
[13] Y. M. Islam, Z. Rahman, S. S.Razzaq, M. A. Sayed, and S. Zaman,
“Effect of feedback during lecture style delivery both in a face-to-face
classroom and during a distance education television session in a
developing country like Bangladesh without the use of Internet,” in
Proc. of 6th IEEE international conference on Advanced Learning
Technologies, pp. 469-471, 2006.
[14] F. Librero, A. J. Ramos, A. I. Ranga, J. Triñona, and D. Lambert, “Uses
of the cell phone for education in the Philippines and Mongolia,” in
Distance Education, vol. 28, no. 2, pp. 231-244, Aug. 2007.
[15] J. E. Horowitz, L. D. Sosenko, J. L. S. Hoffman, J. Ziobrowski, A.
Tafoya, A. Haagenson, and S. Hahn, “Evaluation of the PBS Ready to
Learn cell phone study: Learning letters with Elmo,” reported prepared
by WestEd, Sep. 2006.
[16] F. He, L. L. Linden, and M. MacLeod, “How to teach English in India:
Testing the relative productivity of instruction methods within the
Pratham English language education program,” working paper, Jul. 1,
2008.
[17] U.S. Pawar, J. Pal, R. Gupta, and K. Toyama, “Multiple mice for
retention tasks in disadvantaged schools,” in Proc. of ACM conference
on Human Factors in Computing Systems, San Jose, CA, Apr. 2007.
[18] N. Moraveji, T. Kim, J. Ge, U. S. Pawar, K. Inkpen, and K. Mulcahy,
“Mischief: Supporting remote teaching in developing regions,” Systems,
pp. 753-762, Florence, Italy, Apr. 2008.
150
Management interviews
We conducted structured interviews with five facility Figure 1: Hours of active usage of the Kelsa+ PC per day
managers at the pilot location (across transport, security, (average over each fortnight)
housekeeping, etc.) and recorded their observations around the
Using a different measurement instrument, the responses
Kelsa+ intervention including: the usage patterns of workers,
to the questions on Kelsa+ usage or non-usage included in the
any effects on worker productivity (positive or negative), any
second English proficiency test revealed that all but two
other changes in service staff behavior or work since PC workers in the sample (0.88 share) were active users of the
access was given, etc.
155
Kelsa+ PC. They reported using it on an average of 4 days in Group usage of the Kelsa+ PC was very popular; in fact,
the week preceding the test, and for an average of 60 minutes some users actively sought out colleagues with whom to use
per sitting. the PC. Peer learning sometimes took the form of an informal
teacher-student relationship for a particular task or
application. For instance, as [19] describe, “many workers had
Usage dynamics gone through the email account creation process with a peer
The applications and content accessed on the Kelsa+ PC who was an existing email user. During the registration
changed over the study period. Figures 6 and 7 (on pp. 13) process, in the space where a second email address was
showcase the changing distribution of application and internet required (usually of the person creating the account), the
usage over 31 fortnights. experienced email user friend would enter his email address,
Initial application usage was spread evenly across the since for most workers, this was their first email account.” For
gamut of available software on the PC, indicating heavy other workers, group usage involved a symbiotic learning
exploratory usage. However, starting in the third fortnight, relationship: “I use it with a friend generally, so that we can
Internet Explorer becomes the dominant application launched, learn from one another. What he doesn’t know I tell him, and
followed closely by multimedia applications. There is a shift vice versa.”
away from basic interactive applications that are very popular
initially (Microsoft Paint and offline games, whose usage fell
from 15% to 0), to increased usage of the Internet with its
dynamic content and more sophisticated applications.
As Figure 7 shows, workers spent many months primarily
using the internet for entertainment – viewing music videos or
films on YouTube or Stage6. However, we see a rise in using
the Internet for communication, through email and social
networking sites like Orkut, eight months after the project was
started (correlated with 6 workers simultaneously creating
email accounts at this time).
There is a strong motivation to use the computers for self-
expression, as Figure 3 shows. The ability to create a personal
digital presence, both through images and later through email,
was a source of great pride among workers, and seemingly
altered the way they perceived themselves. Figure 3: Workers creating personalized desktop
backgrounds using the Kelsa+ PC’s webcam
Learning
Workers followed individual and group learning paths. Usability preferences
Several workers simply observed their peers using the PC for Given their educational and linguistic background, workers’
weeks before attempting to touch the PC themselves. As one revealed a number of adaptations in input techniques to
worker commented, “For the first one and half months, I just achieve functional use of the PC. Navigation using a mouse
watched how other people used the computer.” We asked why dominates since typing is problematic. Browser history is
he did this without trying to use the PC himself, to which he repeatedly used as an easy way to access content of choice,
responded that “what if something happened when I used it?” indicating implicit peer learning [Figure 4]. Search queries
[19] Observation and individual exploration established (with the “.com” tag) are repeatedly used to access online
‘learning by doing’ routines, which were instrumental to content within and across sites, given that it involves only
meaningful PC literacy gains. memorizing and typing in one keyword or a short phrase, as
opposed to an entire url. Email forwarding is the dominant
method of online communication, with one user forwarding
up to 15 emails per day to 10 or so people (mostly images).
search result and high motivation levels, and their experiences are anecdotal. At
go back to results best, their outcomes indicate the extent of welfare impact that
list is possible when worker initiative and circumstance interact
Sign-in to E-mail X X X 9 with access to the Kelsa+ facility to produce welfare-
client improving outcomes.
Start MS-Word X 9 X ?
application Upward mobility within the same office
Type in MS-Word X X X X A housekeeping worker with a Class 7, local-language,
Save document X X X 9 rural government school education, and no prior exposure to a
Print document X X X 9 PC, began using the Kelsa+ PC for 60-90 minutes each day
Close Word X 9 X 9 after his work shift. He used it for games, internet video,
application music, etc. A few months later, he created an email ID with
Shut down PC X 9 X 9 the help of a driver colleague, began using the Learn-English
9 : Successfully completed the task CDs available at the Kelsa+ PC, and various Microsoft Office
? : Partially fulfilled the task/ completed the task with applications. He was given an initial typing job on MS Excel
extensive prompting by the IT staff in the office, which he completed and
X : Does not know/ did not attempt submitted satisfactorily. He was then given additional
Source: [19] inventory data entry jobs (on Excel), before being hired as a
dedicated worker for the IT staff.
VII DISCUSSION
In proposing Kelsa+ as a potentially effective ICT4D
intervention for low-income office service staff, we return to
the related work in this field and discuss this intervention’s
comparative strengths and limitations.
Figure 7: Distribution of internet usage at the Kelsa+ PC per fortnight over 18 months.
Increase in Internet video usage is clear. Steady and high usage of email is seen after fortnight 15. Social networking
websites see increased usage starting in fortnight 21.
163
other social enterprises supported these disadvantaged people Khmer Great jobs for young people and students Buying raw
in Cambodia. Mith Samlanh works with street children Mekong agriculture from farmer to push them grow more fruit and
Food vegetable Provide nutrition and quality Safe food to
through their restaurants Friends and Romdeng, and Khana is consumers
an association working with NGOs in support of people living
with HIV/AIDS. Both recognize the potential that income Rajana Find Markets for small producers, give them the
employment, send them to school
Association
generation activities can have in creating sustainable
livelihoods for their target beneficiaries. Senteurs Good working conditions, share benefits with staff through
D’Angkor a bonus system. Create employment for local villagers
Now we turn to social enterprises themselves to measure
their impact on their target community and see how ICT does SotheaKhmer Social welfare for people with disability and vulnerable
Silver Crafts poor people. Work and wage for poverty
or does not play a role in achieving their blended mission. The
analysis will address the blended value proposition in
conjunction with the social enterprise typology to determine What is of interest is the intertwining of both business and
the outcomes and impact of ICT on social enterprises. social goals in many of the respondents answers. This
suggests that the mission of these organisations is a blended
V. KEY FINDINGS proposition providing evidence of this new ‘business’ segment
operating with specific commercial and community
Outcome and impact of ICT – achieving social enterprises’
development goals.
mission
The paper now turns to assess if ICT does deliver benefit to 1) Reframing to reflect the blended mission
social enterprises by considering how ICT helps them to As a result of this investigation, there has been a slight
achieve their blended mission which combines commercial departure from the originally perceived dual mission to the
and community goals. It is hoped that the evidence presented more blended version which is reflected in Figure 2. the
here will contribute much needed insight into the impact of Social Enterprise Blended Mission, which highlights the
ICTs on development more generally. potential outcomes of ICT in social enterprises. As this is a
A. Blending the dual mission qualitative study, following naturalistic inquiry, it is usual for
the research questions to evolve over the period of
In addressing how the dual mission of social enterprises is investigation. What was the original understanding, that the
enacted in practice this paper asks in turn, how ICT interacts dual mission was two separate yet complementary missions,
in this context. The new organisational model of social was due to the perspective as an outsider. Through examining
enterprises embodies a shift in the pursuit of sustainable the data it became evident that this distinction is more blurred
development on all levels, financial, social and environmental. than originally expected.
This move to a ‘blended value proposition’ [14] reflects the
integration of the dual mission, which provides the context as
167
TABLE 6
INDICATORS OF IMPROVED LIVELIHOODS
Indicator % agreement
Earning a livable wage 73%
Provided new career opportunities 63%
Improved food security 71%
Fig. 2. Social Enterprise Blended Mission Improved family stability 71%
Improved health 57%
As Figure 2. demonstrates, it is difficult to differentiate Improved access to education 57%
between the business and social drivers within the blended Increased purchasing power 71%
value proposition. Improved communication can be both with
donors regarding monitoring of a social program such as DDD The nature of social development lends itself to monitoring
and UNDP, or to customers purchasing silk handbags for and evaluation [11], and often the indicators for these
resale in Australia, in the case of Hagar Design through Craft outcomes are clearly set during the start up of the venture. In
Network. the case of donor initiated enterprises there is often a logical
The findings indicate that there is not a clear distinction framework or ‘logframe’ that sets out indicators for
between an NGO that earns income from selling craft, such as monitoring and evaluation. In the case of social enterprises the
Tabitha Cambodia, to an IT company which has a policy of need to measure the triple bottom line means a range of
employing disadvantaged young people, such as Yejj Info. In indicators from both social and business areas are required. In
fact, Yejj Info provides one of the best examples of the evaluating the impact of the current sample, a number of
blended value proposition with their comprehensive mission indicators were presented to social enterprises in the survey.
statements which combines commercially, socially, ethically The 44% incidence of evaluating performance against
and environmentally sustainable internally structured social indicators is promising as it
2) Mapping the blended mission in action suggests that these organisations are very clear of the need to
Social enterprise impact is predicated on the organisation's meet their dual mission. This potentially could mean they will
mission, the social objectives it intends to achieve, and what be more successful than an organisation that is not as clear on
impacts can be measured as a result of the business initiatives. their strategic mission.
A defining ambition of social enterprises is that they support Another category for measuring the success of achieving
those disadvantaged in the community. Leading by example, the social mission is through recognition from stakeholders.
social enterprises employ disabled computer operators at This included acknowledgement from peers, awards, and
DDD, market landmine survivor’s art work at Colours of notes of appreciation from the most important people,
Cambodia, and encourage local communities to invest for beneficiaries and customers which accounted for the highest
their future through a community savings program at Tabitha response rate. Although the survey data had a low incidence
Cambodia. Exactly how this complex interweaving of social of recognition from development agencies/donors through
and financial sustainability is achieved is what this section awards, the other data collection discovered a number of
aims to address. This manifestation of dual sustainability awards and pride in showcasing social enterprises success in
illustrates the true outcomes of the blended mission. What this Cambodia is further explored below.
study acknowledges is that these are inextricably linked, so to C. Identified benefits of ICT
measure one in isolation of the other denies the new paradigm It is through the case of the Cambodian social enterprise
that social enterprises present. sector that this paper demonstrates the opportunity to address
B. Measuring the social impact social issues through business initiatives. Benefits of using
Social impact indicates a positive effect on the target ICT, either directly as in the case of IT services with skilled
population as a result of an intervention and can be measured. labour, or indirectly by selling via an ICT-enabled
Social enterprises, like all development programs, have direct intermediary. If one of the promised benefits of ICT is
as well as indirect impacts. Alter [1] provides some examples increased efficiency of both internal and external operations
of social enterprise impact measurements and corresponding then it is important to measure how ICT impacts on social
indicators. These were adapted to identify the outcomes enterprise operations. In measuring the extent of ICT
(benefits) of ICT to social enterprises and in particular to utilisation in the organisations surveyed, the perception on
assess the impact (attributable to ICT) of social enterprises in impact on operations was deemed an appropriate measure.
the target community. Survey respondents were asked ‘In The most common way ICT helped improve operations was
your opinion how has employment with your organisation through enabling external communications with clients and
improved the livelihoods of your staff?’ and their response is customers. This was further supported by the interview
shown as percentages of those who either strongly agreed or results, with most social enterprises seeing ICT as a core tool
agreed with the indicators listed in Table 6 below. for customer relationship management. Simple ICTs such as
phone and email were the dominant forms.
168
a) Access to Markets when direct sales are for goods produced by the target
Does ICT enable social enterprises to link into the global beneficiaries, customers like to know their purchase actually
trade of goods and services? Through the application of ICT goes to help those who created them. This is where ICT can
are social enterprises able to access new markets that they assist with clear communication and monitoring. An example
previously would have been excluded from? The fair trade of how this is enacted in practice can be seen on the Khmer
movement is a response by civil society to the potentially Silk Processing Association (KSPA) website (See
negative impact of globalisation [33]. It is important to http://www.phaly-craft.org) where they state clearly the
acknowledge that the networked economy can potentially promise that the income goes to those who need it. All income
further marginalise producers in developing countries. Those from the sale of these products are partly used to support
wishing to redress this imbalance see the benefits that ICT can Future Light Orphanage (FLO) activities and children whose
bring as a tool for trade [5]. The majority of enterprises numbers increase everyday.
surveyed felt that ICT provided benefit in key areas of
communication with customers (80%), and sales and c) Employment Opportunities
marketing (64%), suggesting that social enterprises are using Choosing an appropriate target group requires prior
ICT to connect to markets and facilitate trade in ways knowledge and theoretical understanding of the social
previously not imagined. environment of the specific country. DDD identified their
target group as people who fit one of these criteria: disabled,
b) Income Generation orphans (due to war or HIV/AIDS), rural poor and trafficking
One of the defining characteristics of a social enterprise is survivors. They also focus on youth providing entry-level
its income generation activities. This is in fact what positions and training in order to develop their career
differentiates them from others in the non-profit sector as the opportunities.
goal is clearly to earn income to support the social and DDD found that it is important when dealing with disabled
business development of the organisation. What is of interest people to give them the opportunity to prove their skills. On
to this study is the role that ICTs have played in achieving this first appearance a young woman with no hands below her
goal. Respondents to the survey were asked to respond to a wrist may seem to have limited potential as a data entry
number of statements which variously reflected their operator; however once given the chance she has become a
perception of the impact of ICTs on achieving increased sales valued employee who upholds the required productivity levels
or income. These statements were developed from the pilot without having any fingers. This is a prime example of how
assessment of DDD and some preliminary interviews with DDD has taken on the challenge of providing employment
other social enterprises to identify the various ways impact is opportunities for disabled people and met both individual staff
measured. The level of agreement is a combination of agreed goals and general business needs.
and strongly agreed and suggests how these social enterprises
see the relationship between income generation, employment d) Capacity Building
and ICTs. Some social enterprises’ social mission encompasses
TABLE 7 capacity building, where target beneficiaries are provided
ICT ROLE IN ACHIEVING GOALS
training and work experience in using ICTs directly. In this
How do these statements apply to your organisation? % agreement case the social mission drives the adoption of ICTs. In the
Email has increased our sales 50% hospitality industry students are trained on hotel booking
ICT has increased our profits 33% systems such as at Shinta Mani and are able to be prepared for
We have employed more staff as ICT makes it easier for
20% real life experience not just the theory that class based
us to do more work for our clients learning alone provides. By embracing the experiential
Due to increased sales from using ICTs (website/ email)
we have employed more staff
40% learning approach, ICTs can be used to simulate real life
Due to increased income from ICT related we have been experiences and provide an advantage for skill development.
40%
able to offer more social programs to staff One of the major concerns of the digital divide is not only
physical connectivity the ‘second term level digital divide’
As the blended value proposition of the social enterprise is [18]. Skills to effectively use ICTs highlight that connectivity
an interweaving of social and business indicators, it is not alone is not enough to bridge the gap. It is with this in mind
surprising that ICT helps to achieve both these that some social enterprises, particularly in the IT services
simultaneously. Hence separating the business versus social is sector, are providing direct benefit to their target communities.
not that useful, for it is this unique blended characteristic that DDD, Yejj Info and CIST all have training and skills
perhaps makes the impact of ICT more successful than if it development as a core program of assistance. Another social
was separated into a silo approach for development. enterprise in the IT sub-sector, Yejj Technology, operates a
One concern with income generation is that the proceeds go Cisco Network Academy which provides formal training
to support the beneficiaries and are not inappropriately used programs with industry recognised qualifications so that their
for administration or unrelated costs. Of course this is entirely students from disadvantaged backgrounds can engage in the
dependent on the structure of the social enterprise; however IT services sector.
169
e) Strengthening Sponsor Relations competitors. They have actively sought media coverage and
One unique benefit of sponsorship in the highly competitive see it as a way to promote the ideals of their social enterprise
non-profit sector is that a personal relationship is built to their potential market. With the backlash about outsourcing
between a sponsor and a community. It is this connection that jobs overseas a public awareness campaign in the US is very
keeps them returning or undertaking fundraising activities in important for an enterprise such as DDD. On their website
their home country continues sponsorship of either an they have a News section which lists articles and stories which
individual student as in the case of Shinta Mani, or new highlight the experience of the organisation from an external
facilities for the Future Light Orphanage, which is run from perspective. As the article on their website illustrates, DDD
the proceeds from the KSPA. Communication via email is a knows the benefit their business brings young Cambodians
key tool for this, enabling all phases of the relationship, prior like Nut Pove, the opportunity they would otherwise have
to visit and post visit, to continue seamlessly. missed. But it is sadly not enough to provide jobs alone, this is
KSPA has also branched out to provide some guest where ICT can help spread the word of the work being done
accommodation at the orphanage on the outskirts of Phnom in Cambodia and it is examples such as DDD which provide
Penh. Enabling those who wish to make their contribution to evidence of how this is being done.
the NGO through not only donations, or purchasing goods or Although social outsourcing is more commonly connected
services, guests can volunteer to work with the children with the IT services sector, the model is also used in the
teaching English or computers providing a true link to the manufacturing sector. In the case of handicraft sector in
community. The opportunity to personally interact with the Cambodia, many social enterprises work with designers from
children is actively marketed on the KSPA website and also international markets to produce their products. Small
has an online booking form to ease the process. boutique labels make connection with suppliers in Cambodia
The form is then submitted and emailed to KSPA and then through agents such as AAC and Craft Network to arrange
they communicate with the customer via email and finalise the production. For example the designer in the US or Australia
booking and program for their visit. It is the seamless ability send design, colour specifications via email and the
to engage and interact with KSPA even via a simple online production is made to order. As Cambodian enterprises can
form that shows how ICT facilitates this relationship. handle small quantities and offer individual attention to design
Cambodia unfortunately attracts paedophiles who visit houses they are increasingly popular. Bronwyn Blue a Small
Cambodia to take advantage of poor and disadvantaged Business Development Consultant, specialising in the craft
children and women. The management of FLO take the care sector, works with social enterprises to improve their quality
of their children very seriously and screen visitors to ensure control and designs to meet the needs of international market:
they are not placed at risk but benefit from interacting with The core finding presented in this paper is how the blended
only genuinely interested volunteers who match KSPA's mission is enacted in practice and in turn, how ICT interacts in
concern for the welfare of the children as a priority. this context. This study argues that it is the integration of the
dual mission, which is the very nature of social enterprises,
f) Social Outsourcing that ICT brings benefit directly as it assists the overall blended
Focusing on the philanthropic benefit of outsourcing to mission and the ‘total value creation’ that social enterprises
disadvantaged communities has been a unique strategy to bring their beneficiaries.
counteract the backlash that US firms face as they outsource
jobs overseas. Maximising this market niche, DDD shows that
business can benefit communities in developing countries in a
viable business model that does not exploit workers. DDD in
fact targets beneficiaries from marginalised backgrounds and
provides employment where others reject them. Through their
social mission they provide educational scholarships that
support their workers career development to leave DDD and
take up new job opportunities. This then provides an opening
for other disadvantaged Cambodians to join the workforce.
Focusing on the staff’s professional development and
achieving the business goal provides much more of an impact Fig. 4. Map Total Value Creation
for local communities than charity. Businesses who value
social contribution are happy to partner with social enterprises VI. DISCUSSION
such as DDD as they can satisfy their needs to reduce costs
This study found that the major characteristics of social
through accessing an ethical supply chain. All of which is
enterprises are that their development objectives are to
enabled by ICT as digitising content is the mainstay of DDD’s
achieve a blended mission meeting both social and economic
services.
goals. From this basis the assumption would be that ICTs
DDD also understand the benefit of social marketing and
should be utilised to realize these goals [19], [7]. It is with this
the role it can play in bringing a comparative advantage to
170
in mind that this study aimed to find a realistic view of the establishing the social enterprise and of those surveyed three
role ICTs play in supporting social enterprises to meet their were still registered as an NGO whilst eight others were set up
dual mission. What this study acknowledges is that these are as social enterprises from the outset. The summary of results
inextricably linked, so to measure one in isolation of the other in Figure 8.e shows a number of social enterprises as
denies the new paradigm that social enterprises present. classified according to the SE Typology. The level of ICT
Social enterprise impact is predicated on the organisation's uptake is not directly linked to the level of integration
mission, the social objectives it intends to achieve, and what however it could be claimed that there is some correlation
impacts can be measured. Social enterprises, like all social with full integration and management vision with more
programs, have direct as well as indirect impacts. Alter [1] successful ICT assimilation.
provided some examples of social enterprise impact This study argues that it is the integration of the dual
measurements and corresponding indicators. These were mission, which is the very nature of social enterprises, where
adapted to identify the outcomes (benefits) of ICT to social ICT brings benefit directly as it assists the overall mission and
enterprises and in particular to assess the impact of social the ‘total value creation’ that social enterprises bring their
enterprises in the target community. In evaluating the impact a beneficiaries. This is promising as it suggests that these
number of indicators were presented to social enterprises in organisations are very clear of the need to meet their dual
the survey and the results indicated that evaluation of mission. This potentially could mean social enterprises will be
performance was against internally structured social indicators more successful than an organisation that is not as clear on
such as improved livelihoods, increased income, access to their strategic mission. This bears direct significance on the
affordable health care and access to education. This is framing and uptake of ICTs.
promising as it suggests that these organisations are very clear The institutional form that the social enterprise takes itself
of the need to meet their dual mission. This potentially could is another potential influence as to the uptake of ICT.
mean social enterprises will be more successful than an Management awareness, capacity of staff and commitment to
organisation that is not as clear on their strategic mission. use ICT to achieve the blended mission requires both sides to
The goal for development of disadvantaged communities be successful where meeting economic needs is equally as
can take many forms and in the case of social enterprises there important and social goals.
appears to be a number of models as outlined in Alters Social
8
Statistical power, of course, is a function not just of the sample size but
also the effect size.
177
Table II Table III
Mean Number of Services by Corruption and Ruling Multivariate Analysis of Candidate Explanatory Variables
Government Cohesion Variable Model 1 Model 2 Model 3 Model 4
Ruling Difference Constant 24.7 25.8 24.5 24.9
Government in Means,
Cohesion Single (7.19) (6.85) (2.29) (2.30)
Single Coalition Party - Above Average -14.0** -13.7** -13.3* -12.4
Party Coalition Corruption
(t-statistic) (-3.22) (-3.12) (-2.39) (-2.17)
Ruling Party -3.5 -3.4 -4.2
Corruption Low 27.4 11.3 16.1 Cohesion
(2.87) (-0.80) (-0.75) (-0.88)
High 10.2 11.5 -1.3 Months Since .02 .1
(-0.43) Initiation
Difference in 17.2 -0.2 (0.14) (0.38)
Means, (3.04) (-0.06) State Domestic -9.0
Product per Cap
Low - High
(-0.86)
Corruption
N 16 16 16 16
(t-statistic)
r2adj .38 .37 .32 .30
Entries are unstandardized regression coefficients with t-ratios
Alternative Explanations in parentheses. *p < .05 **p <.01
The length of time elapsed since a service center initiative
was implemented might reasonably be expected to influence
the number of available services. As governments in states B. Selection of Services
that implemented centers early will have had more time to The benefits of computerized service centers to citizens
introduce additional services. The level of economic depend not just on the quantity of services, but also on the
development might also be associated with provision of character of the specific services made available. In this
services, as states with higher incomes might have more funds section I consider first whether high demand and high
available to invest in government reforms. corruption services are made available in services centers
In order to test these alternative explanations, I used across all states. I then consider the specific types of services
multivariate regression models to evaluate the relationship of made available in coalition-government led states.
level of corruption, ruling government cohesion, time elapsed The government of India, and in particular state
since implementation, and state economic development with governments in India, provides hundreds of services to
the quantity of services available in the states. As shown in citizens on a regular basis. Table IV lists the fifteen most
Table III, the level of corruption shows a clear and robust commonly provided services in computerized centers across
relationship with the quantity of services, even when taking the states. Out of the 73 government services offered by at
into consideration these alternative explanations. However the least one service center initiative, only these services are
length of time since centers were initiated and the level of offered by more than 25% of the projects (shown in Table
economic development in a state show no relationship with the IV).9 I also include one service, ration cards, that is offered in
quantity of services. Ruling government cohesion also does 24% of initiatives, but is interesting to include in the analysis
not show a statistically significant relationship with quantity because it is provided by a different department, Food and
of services, but as noted above this may in part be due to the Civil Supplies, than any of the other services.
small number of coalition states in the dataset. Are these services in high demand by citizens? Based on
Transparency International India’s survey of corruption in
public service delivery [16], only a small number of those
services most needed by citizens are provided in computerized
service centers. Of the top twenty-five services required by
citizens, thirteen are offered in at least one state. However
only six of these are among the services offered in at least
25% of states (highlighted in bold in Table IV). Thus, in many
cases there is a disconnect between the services that are
provided by states and those that are in high demand by
citizens. This disconnect is most obvious in states with above
average levels of corruption. High corruption states provide,
9
Birth and death certificates, which are offered by 62% of initiatives, is the
only service provided in more than 50% of states.
178
on average, 2.9 of these thirteen services, while below average Table V
corruption states provide 5.2, a difference of means that is Availability of High Corruption Services in Indian Computer-
statistically significant at the .05 level. When the number of Enabled Service Centers
high demand services is regressed on the level of corruption, Reason for Est. % of < Average > Average
the relationship is in the predicted direction and significant at Bribe Annual States Corruption Corruption
the .05 level (t-statistic of -2.33). Bribes Offering States States
($ ‘000)
Table IV Police First 458,000 13% 2 0
Most Common Services offered in Indian Computer-Enabled Info. Report
Service Centers School Fees 340,000 0% 0 0
Department Service (exemption)
Municipal Corp/Rural Dev Grievances Rural 313,000 0% 0 0
Birth & Death certificates Financial
Caste/tribe certificates Loan
Property tax payment Register 307,000 50% 5 3
Food & Civil Supplies Ration cards Property
Revenue Land records School 288,000 0% 0 0
Income Certificates Certificate
Residency certificates Electricity 166,000 13% 2 0
Electricity Electricity bill payment Connection
Estimated annual bribe value based on Transparency
Telephone Company Telephone bill payment International India data [16] and author’s calculations.
Water Water bill payment
Transport Driving licenses In the case of coalition-led states, I also expect to see
Vehicle registration effects of electoral considerations on the specific services that
Home/Police Arms license are chosen for inclusion in centers. Services controlled by
supporting members of a coalition government should be less
Bold indicates that this service is one of the top 25 most
demanded services by citizens, according to [16]. likely to be included than those services controlled by the
majority party.
An alternative measure of citizen demand for service Figure 2 provides an overview of service provision in
computerization is the estimated level of corruption in a coalition-led states. As shown in this graph, politicians from
particular service. Because bureaucrats are more likely to the smaller, supporting members of a coalition are more likely,
demand a bribe for some services than others, service reforms on average, to acquire ministerial posts that involve control
that have the potential to reduce corruption should provide over services commonly offered in service centers. This is in
greater benefits to citizens when introduced for services with a line with expectations that supporting coalition members will
high likelihood of corruption. However reform of these demand ministerial appointments in departments with high
services will also provide the highest costs to bureaucrats and levels of government-citizen interaction. However, when
politicians in terms of lost income, and so computerization in supporting coalition members control these services, it is less
high-corruption potential services should be least likely in likely that they will be included in service centers, as seen
high corruption states. from the right side of the graph. In the case studies below I
When we consider high corruption services, separate from consider the specific ways in which allocation of ministerial
their overall demand by citizens relative to all services, states posts is linked to these lower levels of service provision.
with higher levels of corruption are less likely to provide these
services through computerized service centers. As seen in
Table V, below average corruption states are much more
likely to provide services associated with bribe payments than
states with above average corruption levels. Out of the six
most corrupt services, only one, land records, is available in
any of the above average corruption states.
179
only one, neither of the two ministers from these parties had
the power to bring down the coalition government on its own:
even defection by all five RSP ministers would leave the
coalition with 71 seats, sufficient to maintain its majority. The
remaining services offered through FRIENDS centers were
delivered by departments overseen by CPI(M) ministers.
Among the common services not offered through these
centers, two key departments, Revenue and Food & Civil
Supplies, were overseen by the CPI, the only party with
enough assembly seats to threaten the stability of the ruling
coalition. The other relevant departments, Local
Administration and Rural Development, were both led by
representatives of the CPI(M). The lack of services offered by
departments under the CPI(M)’s most important coalition
partner aligns clearly with predicted outcomes.
Figure 2 – Average Service Provision in Coalition States –
The Akshaya project, a second computer-enabled center
Lead vs. Supporting Parties
initiative in Kerala, offers a similar story of coalition politics,
with an interesting twist. The Akshaya initiative was launched
Kerala
under the United Democratic Front, a Congress Party-led
I begin the case discussion with Kerala. This state is
coalition that came to power in 2001. In this coalition the
viewed by many as an innovator in both development in
Congress held the largest number of seats, at 62, but needed
general [20] and ICT-enabled development in particular [21],
the support of one or the other of its two main coalition
[22]. Despite below average income per capita, Kerala has the
partners, the Muslim League of Kerala or the Kerala Congress
highest literacy rate in the country and the highest overall
(M), to reach the magic number of 71 for a majority.
human development index [23], [24]. Kerala also applied its
Akshaya was launched in 2003 through a pilot project in
development strategies to the use of technology and was one
the state's Malappurum district. The stated goal of the
of the first states to use IT to reform service delivery (Kurian,
initiative was to increase access to technology for the citizens
2003).
of Kerala, and a key part of the initiative was computer
Yet the services offered in Kerala’s two main service
literacy, or “e-literacy,” for the head of every household, a
center initiatives, FRIENDS and Akshaya, pale in comparison
goal that mimicked Kerala’s well-publicized success in
to many of their counterparts in other states. Through
boosting traditional literacy rates (www.akshaya.net).10 But
FRIENDS centers, only ten government services are offered,
these centers provided an additional opportunity for the
with just three offered in Akshaya centers. Could coalition
delivery of government services, as noted by those affiliated
politics be at least partly to blame for the low quantity of
with the project [25]
services in Kerala?
Overall, however, the Akshaya project has implemented
As noted above, analysts have pointed to the importance
next to no services through its network of what in 2008 is
of coalition dynamics in shaping Kerala’s service center
more than 1,000 rural centers. Only three common
initiatives [19]. Once the first service center initiative was
government services are on offer: electricity bill payment,
approved, the FRIENDS initiative, the ruling Left Democratic
telephone bill payment, and grievance redressal. The
Front (LDF) continued to wield their influence over the shape
Electricity Department is overseen by a Congress minister,
of the centers. This coalition was led by the Communist Party
and the Department of Rural Development, which would
of India (Marxist) (CPI(M)), with its coalition partner the
oversee grievances in the rural areas where Akshaya centers
Communist Party of India (CPI) holding the next largest
are located, is headed by a representative of the Kerala
number of seats in the state assembly during the period from
Congress (M).
1996 to 2001. The coalition in total held 76 seats, five more
While it is surprising that the KC(M) would have one of
than the necessary majority of 71. Ministerial posts were
their services offered through the Akshaya centers, it is
allocated across coalition members, with the largest number of
perhaps more telling to note the number of services overseen
ministries predictably allocated to the CPI(M).
by the Kerala Congress (M) that are not offered through these
Six of the fourteen most common services were made
centers. During the UDF government, the KC(M) was in
available through the FRIENDS centers: electricity bill,
control of both the Revenue Department and Rural
telephone bill, water bill, property tax payment, driving
Development, which together account for six of the remaining
licenses, and vehicle registration. Of these services, two of the
commonly offered services, none of which have been provided
relevant departments were overseen by supporting coalition
through Akshaya. Other coalition parties controlled the
members, the Water Department and Transport, led by the
departments overseeing nearly all other common services,
Revolutionary Socialist Party (RSP) and an Independent
member of the legislature, respectively. Because the RSP held
only five assembly seats and the independent, by default, held 10
At nearly 91%, as measured in the 2001 Census, Kerala has the highest
literacy rate of any Indian state or Union Territory.
180
with the Kerala Congress (Jacob) and the Kerala Congress (B) states in the country, with the IT hub of Bangalore as its
overseeing Water and Transport, respectively. Congress capital.
ministers oversaw only two common services that were not On the face of things, Karnataka has also been at the
offered via Akshaya, arms licenses and ration cards. forefront of using ICTs in service delivery. The Bhoomi land
For a portion of these services, the lack of provision records initiative received more media attention than most
through Akshaya is even more surprising, as these services initiatives in the country, and the former Secretary for
had already been implemented at the district level in eGovernance, Rajiv Chawla, is widely recognized as a father
FRIENDS centers. Thus even though the basic infrastructure of the Indian movement to incorporate information
was in place to extend water bill payments, property tax technologies into Indian government processes. Yet, as we
payments, driving licenses, and vehicle registrations through will see, even the presence of early initiatives and an
Akshaya outlets, this was not done under the Congress important evangelist do not guarantee de-politicized service
administration. This means that citizens must either go to the delivery in the state.
district FRIENDS office or avail of these services through A major service center initiative, Bangalore One, was
traditional department-based offices, which are still be prone initiated under the coalition government of the Congress Party
to any corruption in the system. and Janata Dal (Secular) (JD(S)) in 2005. This particular
The twist in the Akshaya case is the way in which the government was formed despite the fact that the BJP held the
policy itself was used as a tool to target the core voters of a largest number of seats in the assembly, because the BJP was
key coalition member party. The Malappurum district where short of a majority and could not agree to a coalition with
the pilot project was initiated has a largely Muslim population, other parties to acquire a majority [27]. So while the Congress
and the Muslim League is the dominant party in the region has the greater number of seats in the coalition, it is highly
[26]. When the UDF coalition regained power in 2001, a dependent on the support of its JD(S) partner.11 Bangalore One
prominent member of the Muslim League, Shri. P. K. was launched in partnership with the municipal government in
Kunhalikutty, became Minister for the Information Bangalore and was intended to provide a similar one-stop
Technology Department. government services environment as offered by Andhra
According to analyst reports, the Minister subsequently Pradesh’s eSeva initiative [28].
consulted with representatives of Malappurum and determined In the case of Bangalore One, however, despite multiple
that information technology could be a “real enabler of the years of experience with technology-enabled service delivery,
local economy” [26]. In order to facilitate IT use, a plan was many of the most commonly offered services are not
developed by the local government to provide for computer available. Of the three services that fall under the domain of
literacy in the district. This plan was subsequently modified in the Revenue department—residency certificates, income
coordination with the Kerala State IT Mission, under the certificates, and land records—none are offered in the
oversight of the IT Minister, and realized as a “telecenter” Bangalore centers. The exclusion of land records is especially
initiative for the district [26]. If successful, this pilot would surprising, given that the Bhoomi land records initiative was
then be extended to other districts in the state. So while the viewed as such a ‘success’ in the state.
stated intention was to develop an initiative that could Why are these common and highly valued services not
eventually benefit the entire state, the initial benefits would offered in Karnataka? The most plausible answer again lies in
clearly be centered on the core constituency of the standing IT coalition politics. The minister in charge of Revenue was a
Minister, representing a key ruling government coalition politician from the supporting JD(S), who also held the post of
member. Deputy Chief Minister, and so was clearly an important player
Akshaya, then, provides an important example of the in maintaining the coalition. Ration cards are also unavailable
combined effects of coalition rule. First, a second computer in the centers, and it was a minister from the JD(S) who
center initiative in the state was launched for the apparent oversaw the responsible Food & Civil Supplies department. Of
purpose of rewarding key Muslim League constituents, a clear those common services that were offered, the JD(S) was
boon for the holder of the IT Ministry post. Second, the lack responsible for only one, electricity bill payment, while the
of services offered through Akshaya reflects the prevalent use Congress was responsible for the remaining services, via
of ministerial posts by the UDF government to reward Municipal Administration, the department that was also
coalition members. Even those services that should be involved in the implementation of the initiative itself. So even
relatively easy to implement because of their inclusion in the in a state with a strong emphasis on technology and history of
earlier FRIENDS initiative have, for the most part, not been government IT initiatives, the dynamics of corruption and
made available to Akshaya patrons. coalition politics seem closely linked to the selection of
specific services made available in the state’s one-stop centers.
Karnataka
The next state, Karnataka, has a much higher level of Rajasthan
corruption than Kerala, which would, on its own, imply a
stronger emphasis among coalition members on the economic 11
The coalition eventually collapsed in early 2006, when a rebel JD(S)
side-benefits of ministerial posts. But Karnataka is also a leader, H.D. Kumaraswamy, pulled out of the government with a group of
other JD(S) MLAs in order to form an alternative coalition with the BJP
“hard” case, because it is one of the most technically advanced (Rediff, 2006).
181
In Rajasthan, the LokMitra (urban) and JanMitra (rural) service delivery. The state of Chhattisgarh was formed in the
centers were opened in early 2002 under the leadership of the year 2000 when it was carved out of Madhya Pradesh. Despite
Indian National Congress party. During this period the below average levels of economic development and a
Congress party held a majority of the seats in the state persistent and violent separatist movement in part of the state,
assembly and so did not require the support of any other the Chhattisgarh government has implemented a substantial
parties in a coalition. The initiation of service centers was eGovernment program through its Choice service centers.
largely an enterprise of the Chief Minister Ashok Gehlot, who These centers, which to date are largely located in urban
sent representatives of the state to Andhra Pradesh in 2001 to centers, provide one of the largest sets of services of any state
evaluate the eSeva initiative there and determine how in India.
something similar could be used to improve service offerings Why is it that Chhattisgarh has implemented such a robust
to citizens in Rajasthan [29]. According to one of the service center initiative? Here again we see the important role
bureaucrats who participated in that mission, a decision was played by Chief Ministers in the character of public service
made to improve on the Andhra model by offering services in reforms. Choice service centers were initially implemented
both urban and rural areas, through two separate initiatives, at under the first Chief Minister of the state, Ajit Jogi of the
least at the beginning.12 Indian National Congress. However the growth of the project
The choice of services in each set of centers was then from a pilot stage has occurred under BJP Chief Minister
determined based more on urban versus rural needs, with the Raman Singh.
LokMitra centers focused on bill payments and JanMitra According to government officials associated with the
centers emphasizing grievances and non-government services Choice initiative, Chief Minister Singh encouraged an
such as agricultural prices. However, in rural areas only four inclusive decision-making process in which the bureaucrats in
government services were offered, while in urban areas just charge of major citizen-facing departments would participate
seven services were made available. in an “empowered committee” that was tasked with making all
The availability of specific services, particularly in urban key decisions about the project moving forward. The goal of
areas, seems directly linked to the interests of the Chief this model was to ensure consensus on decisions so that no
Minister. When bureaucrats in charge of implementing participants would later attempt to block implementations of
LokMitra services attempted to convince various departments the centers or any specific services [30]. In addition, the Chief
to provide their services in the centers, they faced significant Minister took over leadership a “governing council” made up
resistance. The state telephone company did not want to allow of government ministers, which is responsible for overseeing
outsourcing of its bill payments through the computerized the implementation of the project. This meant that the
centers, and the bureaucratic officers found it necessary to implementation team could go directly to the Chief Minister in
request the intervention of the Chief Minister. In this case the case of any problems with implementation. In no other state,
Chief Minister did intervene and the Telephone company was other than perhaps Andhra Pradesh, has the Chief Minister
forced to provide services for bill payment through the taken on this type of direct oversight role in the development
LokMitra centers. On other occasions, however, such as in the and implementation of service centers.
case of income tax payments, the Chief Minister was What factors might have influenced the Chief Minister to
unwilling to take similar initiative and income tax payments take such a strong position on eGovernment in the state?
were not included in the services offered by the centers [29]. Corruption does exist in Chhattisgarh, but government
The mixed response of the Chief Minister in the case of representatives argued that the expectation of the Chief
Rajasthan may be closely tied to the high level of corruption in Minister was that improvements in service delivery would
public service delivery in the state. As one bureaucrat provide a greater potential electoral boost to the ruling
involved in the initiatives noted, politicians “want to provide government that any threat from reduced access to rents. “The
the maximum services to their vote bank,…[but] they also Chief Minister sees the benefits as greater than the costs. He
want money for the next election” [29]. Thus from the has been the brainchild behind all of these frameworks. Indian
perspective of the Chief Minister, by providing some services,
politicians…have very sharp political minds. The Chief
especially those that do not typically involve high levels of
Minister in this term is focused on good governance, on
corruption, such as bill payment, there is an opportunity to
accountability, transparency, and responsiveness of the
reap some political benefits from the centers. At the same
time, those services that may more often involve the payment government” [31]. Given the difficulties that the government
of bribes, such as income tax, can be restricted from inclusion otherwise faces in areas of development and social stability, it
in the centers in order to maintain the rents from those seems that the Chief Minister has adopted service reforms
services. In this way we can see how high levels of corruption through the use of technology as an important platform for
may affect service delivery in single party states. delivering valued goods to citizens. This is possible, at least in
part, because he does not likely feel the same threat to
Chhattisgarh campaign resources from increased government transparency
The final case to consider is Chhattisgarh, a single-party-led that exists in higher corruption states.
state that has below average levels of corruption in public
IV. CONCLUSION
12
JanMitra and LokMitra were merged into a single initiative, eMitra, in The evidence presented here provides strong support for an
2004. Both Jan Mitra and Lok Mitra are Hindi variations on the term argument about the role of corruption and coalition dynamics
“people’s friend”.
182
in affecting which citizens benefit from technology-enabled [18] Government of Haryana, IT Department official, February 4, 2008.
[19] G.R. Kiran, “Front-End First: Citizen Payment at FRIENDS Centres in
service delivery. Citizens in more corrupt states are less likely Kerala,” UNPAN eGovernment for Development Success/Failure Case
to benefit from new technologies. This is because even if Studies, Number 17, 2002.
service centers are implemented in their state, these centers are Available: http://www.egov4dev.org/friends.htm
[20] P. Heller, The Labor of Development. Ithaca: Cornell University Press,
likely to provide fewer and lower demand services than those
1999.
centers in other states. [21] S. Madon, “Governance lessons from the experience of telecentres in
Citizens may also be at a disadvantage depending on the Kerala,” European Journal of Information Systems, Vol. 14, pp. 401-
characteristics of their ruling government. In coalition-led 416, 2005.
[22] S. Madon, and G.R. Kiran. Information technology for citizen-
states, state governments will chose services that ensure the government interface: a study of FRIENDS project in Kerala. World
stability of the government before considering what services Bank Global Knowledge Sharing Program (GKSP), 2002.
might benefit voters. When MLAs from supporting parties [23] Reserve Bank of India; GOI, 2001
[24] Reserve Bank of India; GOI, 2005.
serve as ministers for departments, the services of those [25] Kerala Government, IT Department official, June 22, 2004.
departments are less likely to be implemented, particularly if [26] G.R. Kiran, “Akshaya, Malappuram, Kerala” in Information and
that party holds enough seats to disrupt the majority of the Communication Technologies for Development: A Comparative Analysis
of Impacts and Costs from India,” Bangalore: International Institute of
coalition. Information Technology, 2005.
Who benefits from technology-enabled service centers is [27] Rediff.com, “JD-S, Congress Team up in Karnataka,” May 16, 2004.
thus highly dependent on the political characteristics of a state Available: http://ia.rediff.com/election/2004/may/16karna.htm
[28] Karnataka Government, eGovernance Department official, February 22,
and in particular the extent of corruption and cohesion in the 2006.
ruling government. While there may be great potential to [29] Government of Rajasthan, former IT Department official, May 7, 2007.
improve service delivery through one-stop service centers it is [30] Government of Chhattisgarh, IT agency official, January 23, 2009.
[31] Government of Chhattisgarh, IT Department official, January 23, 2009.
clear that, at least to date, the actual benefits to citizens are
politically driven and thus, in the case of the Indian states,
highly varied.
REFERENCES
[1] World Bank, World Development Report: Making Services Work for the
Poor. Washington, D.C.: The World Bank and Oxford University Press,
2004.
[2] World Bank, ICTs and MDGs: A World Bank Group Perspective.
Washington, DC: The World Bank Group, 2003.
[3] Accenture, Markle Foundation, UNDP, Creating a Development
Dynamic: Final Report of the Digital Opportunity Initiative. 2001.
[4] R. M. Davison, C. Wagner, and L. C. K. Ma, “From government to e-
government: a transition model,” Information Technology and People,
Vol. 18(3), pp. 280-299, 2005.
[5] United Nations, World Public Sector Report: E-Government at the
Crossroads. New York: United Nations Publications, 2003.
[6] D. West, Global E-Government 2005. 2005. Available:
http://www.insidepolitics.org/egovt05int.pdf
[7] J. Bussell, “Electoral Competition and Digital Development in India and
South Africa,” Proceedings of the IEEE/ACM International Conference
on Information and Communication Technologies for Development 2007
(ICTD2007), 2007.
[8] K. De Tolly, B. Maumbe, and H. Alexander. “Rethinking E-Government
for Development: Issues, Lessons and Future Prospects for the Cape
Gateway Portal in South Africa,” presented at the IST-Africa
Conference, Pretoria, South Africa, May, 2006.
[9] S. Madon, “Evaluating the Developmental Impact of E-Governance
Initiatives: An Exploratory Framework,” The Electronic Journal on
Information Systems in Developing Countries, 20(5), pp. 1-13, 2004.
[10] J. Satyanarayana, EGovernment…the science of the possible. New Delhi:
Prentice-Hall, 2004.
[11] R. Heeks, “Causes of eGovernment Success and Failure: Factor Model,”
2003. Available: http://www.egov4dev.org/causefactor.htm
[12] S. Bhatnagar, “Lessons from eGovernment in Developing Countries”
Regional Development Dialogue, No. 24, 2002.
[13] Government of West Bengal, IT Department official, January 18, 2008.
[14] R. Wade, “The Market for Public Office: Why the Indian State is not
Better at Development,” World Development, Vol. 13(4), pp. 467-497,
1985.
[15] F. De Zwart, The Bureaucratic Merry-go-round: Manipulating the
Transfer of Indian Civil Servants. Amsterdam: Amsterdam University
Press, 1994.
[16] Transparency International India, India Corruption Study 2005. New
Delhi: Transparency International India, 2005.
[17] Government of Orissa, IT department official, January 29, 2008.
183
Abstract—The paper presents the results from an assessment are making ICT investments in the public sector. On the other
study of eight e-government projects from India. The assessment hand, evidence of failed projects has drawn attention to the
framework measured the total value delivered by a project to level of risk involved in implementation. A failure rate of
various stakeholders on three dimensions: (a) cost to clients for
accessing services; (b) perception of quality of service and
more than 50 percent is widely cited in this context [1].
governance; and (c) agency cost and revenue. Data was collected A study commissioned by DFID [2] researched firm-level
from randomly selected users encapsulating their experience of impact of ICT in developing countries, 2 but similar studies
using the computerized and manual systems. Impact was have not been undertaken for the public sector. A report by
estimated as the difference between the rating of the the United Nations (2003) laments the fact that documented
computerized and the manual systems. research on the social or economic impact of e-government
development is virtually non-existent [3]. Multilateral
Citizens indicated an overwhelming preference for organizations such as the World Bank fund nearly seven
computerized service delivery. The number of trips required to
be made to the concerned offices reduced significantly and
billion dollars of ICT investment as part of their lending
waiting time came down by nearly fifty percent. Overall impact programs and grant assistance to various client countries. A
showed wide variations across projects. Of the seven agencies World Bank report noted that the largest yet the least
where corruption was reported in the manual system, five monitored investments are IT components of projects in
services saw significant reduction through computerization but different sectors, highlighting the relevance of systematic
the impact was marginal in three projects. Government of India assessment of the impact of these applications [4].
has adopted the framework used in this study to assess the
impact of 40 projects implemented at the national, state and
This paper presents an early effort at evolving a systematic
local levels. framework and methodology for assessing the impact of e-
government projects based on a review of past efforts at
Paper discusses the implications of the results for further assessment of e-government projects. Sections II and III
development of e-government projects in India and identifies the discuss the review and present the key features of the
limitations of the study. framework. The paper goes on to report the findings from a
study that used the framework to ascertain the impact of a
Index Terms—E-government, governance, impact assessment, selection of eight e-government projects from India.
India
II. EFFORTS TO ASSESS PUBLIC SECTOR IT PROJECTS
I. BACKGROUND While significant amount of academic and policy research
has focused on evaluation of public sector ICT projects, a
I N recent years a large number of countries have launched
“e-government” programs, and several development
agencies and governments have identified e-government
systematic framework for evaluation has not emerged. One
group of studies provides macro-level estimations of e-
government activity using appraisal indices focusing on
implementation as a key policy priority. 1 Driven by the supply-side, quantifiable measures such as web presence of
success of a few projects in improving delivery of services to government, network coverage, institutional and regulatory
citizens and businesses, an increasing number of governments support and human capital provision. 3 Such factor-based
assessments of e-readiness do not have immediately obvious
Manuscript received February 5, 2009. and tractable policy implications. They tend to focus almost
S. C. Bhatnagar is with the Indian Institute of Management, Ahmedabad,
India (phone: 91-79-6632-4802; fax: 91-79-6632-6896; e-mail:
2
subhash@iimahd.ernet.in). The study sponsored by DFID at the London Business Schools collected
N. Singh is with the Centre for E-Governance, Indian Institute of data from firms in India and Brazil to establish a relationship between ICT
Management, Ahmedabad, India (e-mail: nupurs@iimahd.ernet.in). investments, profitability, growth and productivity at the firm level. The
1
Examples of e-government initiatives launched by international agencies degree of collateral organizational change was studied as a mediating
can be found on the following links: World Bank variable.
3
(http://www.worldbank.org/egov); ITU (http://itu.int/itudoc/itu-t/workshop/e- Examples of such frameworks are (1) UNPAN (2004) E-Government
gov); and DFID (http://www.publictechnology.net/modules.php? Readiness Report www.unpan.org and (2) Brown University (2004). Global
op=modload&name=News&file=article&sid=1533) E-Government Report, available on www.insidepolitics.org/egovt05int.pdf
184
exclusively on measurement of physical access to certain creation in e-government projects have been documented in
types of ICT without incorporating issues such as reports by governments and external consultants [6] – [8]. An
affordability, appropriateness, ICT capacity and training, and example is the study conducted jointly by the Danish Ministry
the regulatory and macroeconomic environment [5]. The of Finance and Accenture which identifies two primary
second group of studies, largely anecdotal, done in a dimensions of value: (i) value to investors in terms of tangible
piecemeal fashion, provide project-level evaluations with little financial benefits, cost savings, cost avoidance, and increased
prospect for synthesis from past approaches. Evaluative revenue; and (ii) value to users in terms of improved services,
studies have been done to serve a variety of purposes. Some reduced cost and/or time savings to citizens, and reduced
studies looked at implementation success in terms of whether administrative burden to businesses. Investment in
the systems were functioning as they were designed to, or the government employees and technological infrastructure are
degree to which the intended outcomes were achieved. Some identified as key enablers for creating value through e-
studies looked at long term sustainability and scope for government. The focus has been on using business case
replication, 4 while some measured the benefits that were methodologies from the private sector to demonstrate the
delivered to agencies. 5 A few focused on benefits to the economic value of e-government projects. The calculation of
clients. 6 There was hardly any comprehensive study that time and money spent in finding and using public information
assessed the impact on all the stakeholders and covered both are the most direct and measurable benefit of e-government
short-term and long-term direct and indirect impacts. A few applications to users, as shown in recent OECD and EU
studies had carried out a cost-benefit analysis. Often, studies [9],[10]. These reports identify improved revenue
evaluation studies had been done by agencies that were likely collection, lower costs due to efficient processing of
to have an interest in showing a positive outcome. transactions, and a reduction of administrative burdens due to
A variety of approaches have been used for evaluation. simplification or elimination of procedures as some direct
These included surveys, expert opinion, ethnographic studies impacts that can be measured. Financial savings to users in
and internal assessments produced by lending agencies. The terms of time and money spent in finding and using public
utility of such evaluations has been limited because: information are the most direct and measurable benefits of e-
government applications for clients. A significant amount of
▪ Different studies of the same project showed very
work has also been done on studying users’ perceptions of
different outcomes, thus indicating a lack of credibility of
results. 7 Part of the reason for different outcomes was the quality in terms of attributes such as accessibility, attitude of
use of very small samples and a lack of rigor in sampling staff, cost of service, provision of information, procedural
and collecting data from clients of the systems. The fairness and convenience.
results could therefore not be easily generalized over the Calculation of net economic benefit of an IT project has
entire population of clients. proven to be challenging as outcomes are multi-dimensional
and composed of both quantitative and qualitative indicators.
▪ The studies evaluated the functioning of the Further attention is required to be given to the linkages
computerized system but were not able to assess the amongst issues of quality of service, governance, wider
difference made by ICT use, as the need for
impacts on society, and ICT investment. Newer frameworks
counterfactuals (evaluation of systems as they worked
are evolving around a notion that an exclusive focus on
before computerization) was ignored. Often, the impact
financial costs and benefits for the government ignores many
of ICT use was not separated from other interventions
that were made simultaneously with the computerization important non-economic benefits [9]. One such benefit relates
effort. to improvements in various aspects of governance activity.
Another non-economic benefit relates to the addressing of
▪ Finally, since different studies did not use a standard wider policy priorities, which, in the context of developing
methodology, it was difficult to compare the outcome of countries could constitute the priorities as articulated in the
one project with that of others. UNDP Millennium Development Goals.
Work done in EU countries has focused on understanding III. MEASUREMENT FRAMEWORK AND METHODOLOGY USED
the ‘processes’ that deliver a return on investment and IN THE STUDY
generate value [6]. Common best practice factors for value
Two approaches - MAREVA developed by the ADAE
4
(Agence pour le Development de l’Administration
Government of India, Ministry of IT has commissioned quick assessment of
29 projects. Electronique - Electronic Administration Development
5
Korea’s eProcurement agency has evaluated the impact on different Agency) in France with the help of Bearing Point and the
government agencies using the system. WiBe Economic Efficiency Assessment methodology being
6
Global Knowledge Sharing Program got 4 Indian projects evaluated where
clients were surveyed.
used by the German federal administration [11] were useful in
7
For example the Bhoomi project of issuing copies of land title has been developing the framework proposed in this chapter. These
evaluated by Public Affairs council reporting a significant positive outcomes methodologies developed by two Governments in EU
including reduction in bribes. Recent studies by a team from MIT and IIIT
Bangalore found that corruption had not declined and major benefits were
countries focus broadly on the same dimensions. They offer
derived by land sharks. two levels of impact assessment: first, in terms of how the
project provides a business case justification of expenditure
185
and second, in terms of how the project meets the goals of the of existing frameworks has been used in this study. An e-
agency concerned and, in turn, how this helps in meeting service delivery project impacts three groups of stakeholders:
wider government strategies. Guiding these assessment (i) clients receiving the service; (ii) agency (including
frameworks are strategic outcomes pursued by respective implementation partners) that delivers the service; and (iii) the
governments representing broad policy priorities that drive larger society consisting of citizens, businesses, government
the direction of government. These considerations have been as a whole and civil society. The impact can be assessed in
incorporated in the assessment framework proposed for e- terms of a variety of outcomes experienced by each type of
government projects in this section. 8 stakeholder. Table I lists key dimensions of outcomes for each
Borrowing from the above approaches and the work done type of stakeholder.
in EU countries, the framework used for the study reported in TABLE I
KEY OUTCOME DIMENSIONS
the paper focused on the idea of measuring the total value
Stakeholders Key Dimension of Impact
delivered by a project to different stakeholders and takes a
Client ▪ Economic (Direct and indirect)
balanced approach between case study and quantitative ▪ Governance (Corruption, accountability,
analysis. It recognizes that some part of the value for each transparency, participation)
stakeholder can be monetized and other part needs to be ▪ Quality of service (Decency, fairness,
assessed qualitatively. convenience, etc.)
Most of the assessment models discussed in the literature Agency (Including partners ▪ Economic (Direct and indirect)
in implementation) ▪ Governance (Corruption, accountability,
have not been used in the context of a developing country and transparency, participation)
many have not yet been applied in practice even in developed ▪ Performance on key non-economic
countries. Nor do they account for a variety of delivery objectives
▪ Process improvements
models used in developing countries such as common service
Society ▪ Long term impact on Millennium
centers and franchised outlets that can retail e-services offered Government as a whole Development Goals
by the government. Practical issues of paucity of data have ▪ Image of the government
not been taken into account, particularly in case of a
developing country where baseline surveys are not done and The primary objective of the study was to measure the
monitoring and evaluation systems are weak. Recognizing impact of computerization on clients (users) of selected
that adequate data for quantitative assessment is not available service delivery projects and to test the applicability of the
for most of the projects, collection of such data using a framework across a variety of projects. For the purpose of this
standard measurement framework for sufficiently large study a sample of eight mature projects shown in Table II was
number of projects is the first task that needs to be selected. These projects covered services to rural as well as
undertaken. Since impact assessment is the key objective, urban citizens (G2C), services to businesses (G2B) and
establishing counterfactuals is an important element of the services for internal government users (G2G). Most of the
proposed measurement framework. services are offered by state-level agencies except in eSeva
A number of empirical studies suggest that ICT has had an where services from the federal government are also offered.
impact in improving the performance of private sector In eProcurement services are accessed through a portal. In the
organizations particularly in developed countries. However, remaining projects service delivery is through assisted
regarding ICT investments by the public sector in developing computerized counters set up by the agencies at various
countries, many researchers have noted that past evaluation locations.
studies have not used a common framework or methodology For each project, the measurement framework was
and that rates of success/failure have been declared based on converted into a set of data collection instruments including:
purposive samples [12]. Since ICT is introduced mostly in the i) a profile of the project identifying services, clients and other
context of governance reform to improve efficiency, stakeholders; ii) agency level data on activity levels,
effectiveness, and transparency of governments, a crucial first investments and operating costs; iii) a client survey
stage is to ascertain to what extent these intended outcomes questionnaire covering direct cost of access, quality of service
from e-government applications have been achieved. and governance, and a few measures of overall satisfaction;
A common measurement framework evolved on the basis and iv) an employee survey for perceived impact on work,
efficiency and effectiveness.
8
The indicative items are based on a review of the following documents: The survey assessed both the manual system and the
i) Performance Reference Model of the US Federal enterprise Architecture computerized system that replaced it on all the above
framework used by the office of Management and Budgets in US Federal
Government; ii) European Commission, eGovernment Economics Project dimensions. An analysis of the differences between the old
(eGEP), Measurement Framework Interim Version Deliverable (D.2.2), 2005; and the new system provided a measure of impact. Random
iii) TP Rama Rao, V Venkata Rao; SC Bhatnagar; J Satyanarayana, ‘e- samples of about 30 users were chosen from 8 different
Governance Assessment Frameworks’, 2004, http://egov.mit.gov.in; iv)
Edwin Lau, ‘Electronic Government and the drive for growth and equity’, locations in a state. Locations were chosen to represent
OECD, 2005; and v) Lanvin Bruno, ‘METER E-strategies Monitoring and different levels of activity at the service centers, and different
Evaluation Toolkit’, in Robert Schware (ed), E-Development: From levels of development of the region. Thus nearly 250
Excitement to Effectiveness, Prepared for the World Summit on the
Information Society Tunis, 2005, World Bank.
respondents were surveyed for each project. The
questionnaires (translated into local languages) were
186
administered at the home of the sampled users by request in all applications. 9 This has resulted in increased
investigators who were trained to understand the nature of predictability in service delivery and reduced the number of
projects and to interpret individual items in the instrument. trips.
The survey was combined with secondary documentation and 4.00
primary data gathered through field visits, interviews, surveys Manual Computerised
3.37
and opinions solicited through e-mail. 3.50
TABLE II
3.00 2.81
LIST OF EIGHT PROJECTS FOR DETAILED STUDY 2.71 2.74
80.00 3.00
68.06 2.75
63.16 60.06
60.00 2.50
S.E. 8.45 4.85 67.68 8.82 32.90 15.35 23.94 8 8.26 8.02
Wage loss 98.25 94.66 262.38 36.25 161.75 65.00 88.33
6
(Rs)
Computerized 4.74
4 3.88
3.31 3.49 3.56
N 81 59 122 8 63 22 24
2 1.98 1.67
S.E. 10.69 13.50 24.38 6.73 32.79 22.55 20.81 1.52 1.14 1.26
0.74
N: Sample size 0
Bhoomi-RTC Bhoomi- KAVERI Khajane- Khajane- CARD eSeva AMC
While the difference between manual and computerized Mutation DDO Payee
systems on all the cost indicators are statistically significant Manual Computerised
(in most cases at a confidence level of 99 percent), the
accuracy of the estimate of the difference in direct costs can Fig. 4. Error Rate (in percentage)
be improved by using larger samples.
C. User Perception of Governance and Corruption
B. User Perception of Service Quality
The proportion of users paying bribes has generally
For all eight projects, users rated the computerized services
declined. The decline is higher in some projects than in others
higher in overall quality (Fig. 3). The differences were
(Fig. 5). For example, in the Bhoomi project, while about 30
statistically significant. eSeva has shown a significant
percent of users were paying bribes in the manual system, less
improvement in service quality whereas eProcurement has
than one percent is continuing with the practice in the
shown a marginal improvement. Bhoomi also shows very
computerized system. The proportion of transactions in which
significant improvement in service quality, amongst the
a bribe was paid in KAVERI continues to be high at 21.61
highest in the projects assessed in the study.
percent in the computerized system. Out of the seven agencies
The rate of error in documents (as reported by the users),
(and eight projects) where significant corruption was reported
which is an important measure of quality, was also lower for
in the manual system, four were able to eliminate or
all the projects except the issue of RTCs (see Fig. 4). The
significantly reduce corruption through computerization. The
number of RTCs issued has gone up many-fold after
impact was marginal in the remaining three agencies as
computerization and even minor mistakes in the name are
corruption continued at a significant level even in the
now being noticed and reported. Reduction in errors saves
computerized systems. Although, in reducing corruption the
additional trips for users and improves productivity by
188
outcome is mixed, e-government seems to have the potential D. Comparison of Projects on Overall Client Impact
for significant reduction in corruption in service delivery, as Respondents were asked to rate each project on 18 common
shown by one project. attributes encompassing the three dimensions discussed
40.00
earlier. They were also asked to select the three attributes that
Manual Computeris ed they considered most important. A weighted composite score
35.00
33.82 34.32 that factors in the importance and rating of the attributes was
calculated for each project. Table IV compares the composite
30.00 29.71
28.02 scores across projects on a 5-point scale. It indicates that
25.00 Bhoomi shows the maximum improvement as its manual
23.71
21.61 version had the lowest score. eSeva also shows a very
20.42
20.00
significant improvement as the computerized counters are
15.00 14.48
rated close to very good in the composite score. KAVERI and
14.17
CARD indicate only marginal improvement over the manual
10.00 system. Ranking of projects using the composite rating, and
5.00 5.68 particularly its improvement, can represent the degree of
2.71 3.35 success of the project from the point of view of clients.
0.00 0.84 0.74 0.00 0.00 0.40
0.00 0.84
Bhoomi and eSeva can be rated as very successful projects,
Bhoomi-RTC Bhoomi- Kaveri Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost AMC as moderately successful and KAVERI and CARD as
Mutation Payee
projects where there is considerable scope for improvement.
Fig. 5. Proportion paying bribes (percentage). TABLE IV
DESCENDING ORDER OF IMPROVEMENT IN COMPOSITE SCORES
An interesting phenomenon is that the amount of bribe Manual S.E. Computerized S.E. Difference S.E.
users pay is generally higher in computerized systems in all Bhoomi 2.86 0.07 4.46 0.04 1.60 0.09
the projects. This can perhaps partly be attributed to general eSeva 3.39 0.04 4.66 0.03 1.27 0.05
inflation in the price levels. It may also suggest that the Khajane DDO 3.24 0.08 4.43 0.05 1.19 0.10
complexity of the types of favor requested through bribes has Khajane
3.08 0.07 4.19 0.05 1.10 0.10
increased. It must be noted that the estimates of bribe amounts Payee
are less accurate than other results as the sample of eProcurement 3.22 0.04 4.26 0.04 1.04 0.05
respondents paying bribes was small resulting in high Checkpost 3.48 0.05 4.32 0.04 0.84 0.05
standard error of the estimates. AMC 3.37 0.04 4.12 0.06 0.75 0.06
The overall governance rating is higher for computerized KAVERI 3.35 0.06 3.89 0.05 0.55 0.05
systems than manual systems (Fig. 6). In the case of Bhoomi, CARD 3.78 0.03 3.93 0.03 0.15 0.03
there was a marked improvement in transparency and fairness
whereas KAVERI showed hardly any improvement in the As shown in Table V, three kinds of attributes have been
quality of governance. selected more often than others. They include attributes
5.00 pertaining to: i) transactional efficiency; ii) improved
Manual Computerised
4.48
governance; and iii) quality as measured by error rate and
4.50
4.20 4.19
4.31
4.21 convenience. An important conclusion from this analysis is
4.03
3.94 3.99 that important attributes are different for different projects. If
4.00
3.58 3.61
3.74 such an exercise was to be done before designing an
3.50 3.51
3.50 3.25
3.34 3.33 3.33 application it would provide useful insights into benefits that
3.15
need to be targeted and the kind of process reform required. It
3.00 is rarely done in practice either before or after the project
implementation.
2.50
2.00
1.50
1.00
Bhoomi Kaveri Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost
Payee
Diff. 0.76 0.19 0.70 0.61 0.61 0.38 0.79 0.75 0.88
S.E. 0.07 0.04 0.06 0.06 0.04 0.04 0.04 0.04 0.06
Fig. 6. Overall governance score (5-point scale). A paired t-test was applied
to assess whether the differences between the overall governance scores for
the computerized and manual systems were significantly different.
189
TABLE V with manpower, as was seen in the case of KAVERI. The
TOP FOUR DESIRED ATTRIBUTIONS OF SERVICES
human resource and other costs incurred by the government
Project Desired Attribution of Services
have remained within a narrow range for the manual and
Bhoomi Error free No delay in Less waiting Fewer visits
transaction
computerized systems. In terms of the impact on agencies, the
transaction time
KAVERI Less Greater Error free
ability to cope with growth in transactions was enhanced in all
Less waiting
corruption transparency transaction time cases. In some cases, computerization helped in the growth of
Khajane DDO Simplicity of Convenient Friendly Error free transactions.
procedures time schedule attitude of transaction Data from employees on perceptions about impact on work
officers load and managerial processes was analyzed. The staff in
Khajane Payee No delay in Convenient Good Error free these agencies did not perceive that cost had been reduced.
transaction time schedule location transaction
However, they felt that there was a significant positive impact
CARD Less time Less waiting Less Fair
and effort time corruption treatment on the quality of governance. Computerization was also seen
required to be partially responsible for improved tax collections by
eProcurement No Easy access Equal No need to some agencies.
corruption opportunity visit
to all Government
office VI. KEY FINDINGS AND IMPLICATIONS
eSeva Less time Less waiting Convenient Fair Respondents (citizens and businesses) who have used both
and effort time time schedule treatment the manual and computerized systems indicated an
required
overwhelming preference for the computerized service
AMC Less time Less Greater Good
and effort corruption transparency complaint delivery in most projects (see Table VI). However, in the case
required handling of Bhoomi, 79 percent of respondents preferred the
system computerized system whereas in most other projects the
Checkpost No delay in Error free Error free Proper figure is 95 to 99 percent.
transaction receipt transaction queue system TABLE VI
Legend: Underline - Improved Governance; Bold - Transactional PREFERENCE FOR COMPUTERIZATION (PERCENTAGE)
Efficiency; Italics - Quality Project Preference for Computerization (%)
Bhoomi 79.34
eProcurement 83.71
V. RESULTS OF IMPACT ON AGENCIES
Checkpost 91.25
Data on transaction volumes, operating costs, investments, eSeva 96.84
tax collection (if applicable), and revenues from transaction CARD 96.98
fees was to be collected for three years prior to the AMC 97.49
introduction of computerized system and for the entire period KAVERI 98.31
since computerization. However, it was very difficult to
collect accurate data for pre-computerization as often agencies A profiling 11 of those who preferred the manual system
did not have an MIS that was reporting such data. Post- revealed that higher-income farmers with large holdings in
computerization data on transactions and revenues was easier urban areas preferred it to the computerized system. Perhaps
to collect as it had at least been recorded in individual such farmers were able to manipulate the flexibility in the
agencies. system to their advantage - something that needs to be further
Often, the prevalent costing systems do not permit investigated.
identification of operating costs for the computerization Preference for the computerized system is backed with
project alone. Many costs are joint costs being incurred to specific areas where concrete benefits have accrued to them.
support multiple activities. For different agencies, In most cases the cost of accessing service was reduced
computerized delivery of services touches different portions because the number of trips that needed to be made to the
of the overall activity portfolio. For example, AMC is an concerned offices saw a significant reduction and the waiting
organization which delivers large scale physical services such time came down by nearly fifty percent. Corruption was
as cleaning of roads and public areas, public lighting and significantly reduced or eliminated in five projects. Quality of
maintenance of urban infrastructure. Computerized delivery service delivery and quality of governance were also
of services impacts less than one percent of all employees. perceived to have improved significantly with
For agencies such as offices of the sub-registrar the impact is computerization in most cases.
more widespread as the core service is computerized. There is considerable variability in the composite scores
By and large, the major cost component was manpower. across the eight projects. The scores range from a 0.15 point
The study revealed that the staff was not reduced in any
project after computerization. In some projects work load was 11
An interaction tree was constructed by considering preference for the
reduced and therefore, additional time could be devoted to computerized/manual system as a dependent variable. The independent
other tasks. There could be cost implications in the future as variables comprised those that are indicative of the user’s profile, viz.
the transactions are growing at a much faster rate as compared occupation, income and urban/rural status.
190
difference, which represents virtually no improvement, to 1.6 of disaggregated data at the level of each service location or
(on a 5-point scale), which is quite significant. Other studies for different types of clients.
have shown a similar variation for the same project The conclusions on impact on agency are not as robust as
implemented across different ststes [13]. This suggests that those for impact on citizens because time series data on
projects are not being designed using learning from other different types of costs and revenue streams related to the
states. service being investigated could not be collected for sufficient
The distinct preference for computerized systems by the time periods for most of the projects.
clients of poorly rated projects seems to suggest that even The overall positive assessment of projects should be seen
small gains for the users can trigger major positive change in in the context of the eight projects that were selected for
perception about service delivery systems. assessment. These were mature projects, with wide reach and
Incremental operating costs for agencies (including scope of activity. These projects have been operational for at
amortized investment) per transaction could be compared with least three years, implying successful implementation. Nearly
the reduction in direct cost of accessing the service reported all the projects serve urban clients. Projects that serve rural
by the clients. This would indicate the level of user fee that clients could have a very different cost structure and demand
can be charged (should be less than the monetized gain to the pattern. Rural projects also face a greater challenge in the
user) and the degree to which the project could be made maintenance of infrastructure. For all the projects
economically viable. The fact that many projects can become improvements were measured in comparison with the manual
self-sustaining through revenues from user fee indicates that systems that were replaced. When countries move from
private sector investment can be tapped. Five of the eight computerized systems in departments to a 24x7 service
projects have private partners and the expansion of the sixth delivery through portals, the degree of perceived
project in rural areas is being planned through private improvement could be very different.
partnership. For understanding the costs and benefit of ICT investments
Further analysis of data indicated significant variability of at a macro level, more projects from different contexts need to
impact across different delivery centers in a given project. be evaluated and failed projects (that could not be
This variability should be a cause for concern for delivery successfully implemented) would have to be included in the
models in which physical service centers are created. Part of analysis. It would therefore be hasty to generalize the overall
this variability is explained by poor infrastructure (power and conclusions of positive impact and economic viability of
connectivity), particularly in rural areas. Also, given the large electronic service delivery projects from this study.
variation in activity levels at different centers, it is often Further studies need to be undertaken to explain the
difficult to match the capacity to the demand at each of these variation in impact on various dimensions; differences across
centers. Portal-based delivery accessed via the Internet can be locations for a project; and differences across projects. Studies
a solution. However, unequal access to Internet can put some could also be undertaken to understand the effectiveness of
users at a disadvantage in such systems. different delivery models and implementation modalities such
A detailed study of one of the centers in KAVERI as the use of public private partnerships. More data
indicated that any type of system breakdown leads to (quantitative as well as qualitative) will need to be collected
corruption. The breakdown can be on account of an overload from the agencies to undertake such studies.
of demand in comparison to the capacity of the system to The establishment of acceptable counter-factual was a
process registrations. Systematizing queues by appointments challenge for the study team. All the projects had
helps prevent break-down. Agents play a key role in discontinued manual delivery of service and had mandated the
promoting corruption. Private operators also exhibit rent- use of the computerized systems for the citizens. For the
seeking behavior if given an opportunity. assessment of manual system, respondents needed to rely on
memory. In case of systems that have been operational for a
large number of years, such recall can be prone to error. There
VII. LIMITATIONS OF THE STUDY AND AREAS FOR FURTHER has been no benchmarking of the service delivery in a manual
RESEARCH system prior to implementing a new computerized system -
The study was exploratory in nature undertaken with something that needs to be done for new projects that are
limited resources. More countries could have been covered taken up in the future. Other forms of counterfactuals such as
and a greater variety of applications could have been assessed a user group from an adjoining state could be used to avoid
for generalizing the conclusions. The study used a sample size the problem of recall. However, use of a different context
that was considered relatively small (and barely adequate) at could result in other types of biases.
the stage of planning the field work. While the difference The study assessed direct economic impact in terms of cost
between manual and computerized systems on all the of accessing the service. It does not measure the impact on the
performance indicators are found to be statistically significant inherent value of efficient delivery of the service for the
(in most cases at a confidence level of 99 percent), the client. For example, copy of a land title is required for
accuracy of the actual estimate of the difference in direct costs obtaining a farm loan, getting a bail in a court case, insuring
could be improved by the use of larger samples. A larger crops and for purposes of checking the veracity of the record.
sample size in subsequent studies would also permit analysis A farmer would attach different values to efficient delivery of
191
land title depending on the purpose for it was required.
Subsequent studies should also attempt to estimate the impact
on indirect economic value.
REFERENCES
[1] R. Heeks. (2008, May 8). Success and failure in eGovernment projects.
Available: http://www.egov4dev.org/success/index.shtml
[2] R. Basant, S. Commander, R. Harrison, and N. Menezes-Filho, “ICT
adoption and productivity in developing countries: New firm level
evidence from Brazil and India," IZA Discussion Paper No 2294,
Germany, Sep. 2006.
[3] “World public sector report 2003: E-government at the crossroads,”
United Nations, New York, 2003.
[4] “Information and communication technologies: A World Bank Group
strategy,” World Bank Group, 2002.
[5] bridges.org (2005, May). E-ready for what? E-readiness in developing
countries: Current status and prospects toward the millennium
development goals. InfoDev [Online]. Available:
http://www.infodev.org/files/2049_file_InfoDev_E_Rdnss_Rpt_rev11
May05.pdf
[6] “Value creation in eGovernment projects: An exploratory analysis
conducted for the Danish presidency of the eGovernment work group of
the Directors General,” European Union, 2003.
[7] “eGovernment: Realizing the vision,” Third Annual eGovernment
Benchmarking Report, Accenture, 2002.
[8] “Is IT worth it - Presenting the public sector business case,” Gartner,
2002.
[9] E. Lau, “Electronic government and the drive for growth and equity,”
OECD E-Government Project, 2005.
[10] “eGovernment economics project”, DG Information Society and Media,
European Commission., 2005.
[11] “WiBe 4.0 - Recommendations on economic efficiency assessments in
the German federal administration, in particular with regard to the use
of information technology,” Federal Ministry of the Interior, Germany,
2004.
[12] R. M. Peters, M. Janssen, and T. van Engers, “Measuring e-Government
Impact: Existing Practices and Shortcomings,” in Proc. 6th International
Conference on Electronic Commerce, Delft, The Netherlands, 2004.
[13] “State level e-governance projects in India: Overall assessment of
impact on citizens,” Centre for E-Governance, Indian Institute of
Management, Ahmedabad, unpublished.
192
Abstract—We examine the extent to which user-based subsidies of the program can be enhanced. For example, better social
can promote the sustainability and development impact of targeting of the coupons could have brought in more women or
telecenters, where sustainability is defined in financial and social under represented groups. From a financial sustainability point of
terms. We do this by looking at a coupon scheme used by the view, explicitly targeting new users (as opposed to people already
USAID funded eCenter network in Kyrgyzstan. The network using the center) could have also been more effective. In terms of
consisted of partnerships with existing commercial computer impact, coupons could have been distributed, for example,
centers which provided fee-based ICT services to their specifically to entrepreneurs looking to start a new business. A
communities. The eCenter program temporarily provided method of incentivizing eCenter management to perform such
subsidized coupons for Internet access and computer training to targeting is probably required.
users of these centers.
Using a mixed-method approach, we found that user-based Index Terms — Telecenters, Sustainability, Subsidy,
subsidies have to a certain degree aided financial sustainability by Kyrgyzstan, Central Asia
bringing new users to the centers, some of who will, conceivably,
become long-term customers bringing ongoing revenues. The
subsidies also helped to “kick-start” operations during the initial I. INTRODUCTION
stage of the project. However we found that the distribution of
the coupons did not lead to any significant enhancements to social As a method of improving access to information and
sustainability by, for instance, encouraging users from under- communication technologies (ICTs), the telecenter is popular
represented social groups (though we did find that women were among donors and governments in many different countries [1,
marginally more likely to take advantage of the program). 2]. The concept of the telecenter emerged from a community
Moreover, we found that the distribution of both Internet and driven movement in Scandinavia in the 1980s [3]. For poor
training coupons favored more regular users of the eCenters. rural communities, telecenters can provide access to
Finally, looking beyond sustainability to impact within the communication and content. This can include access to market
community, we found that the coupon program had a limited
and crop prices, to financial information/services, and
development impact on participating communities. For example,
the population of users who reported economic benefits from communication with friends, family, and business colleagues
using the eCenter (eg. subsequent employment or starting a new [4]. Telecenters can also support the delivery of government
business), did not make more use of the coupon program than the services [5]. Today, telecenters can be found in many countries
user population at large, suggesting that the coupon program and are referred to by a plethora of terms: tele-cottages, public
itself did not account for this benefit. information access points, public internet access points or
We argue that, even in the presence of such modest positive multi-purpose communications centers. While each of these
effects, user-based subsidies still offer an intriguing model. We terms represent slight differences in the object being discussed,
believe that if the eCenters had narrowly targeted particular the common characteristic of telecenters, as used in this paper,
participants for the coupon programs it is likely that the benefits
are publicly accessible ICT resources in one or more physical
locations that are available with or without fees.
As with any other development intervention, a critical
Michael L. Best is an Assistant Professor with the Sam Nunn School of objective is to make the telecenter sustainable. Another
International Affairs and the School of Interactive Computing, Georgia objective is to address the problem of poverty and other socio-
Institute of Technology, Atlanta, GA, 30332 USA. (e-mail:
economic concerns within the community at large; thus a
mikeb@cc.gatech.edu).
Dhanaraj Thakur is a PhD student with the School of Public Policy, successful solution should also have real impact [6, 7]2.
Georgia Institute of Technology, Atlanta, GA, 30332 USA. (e-mail: However, many telecenter projects have not realized either of
dthakur@gatech.edu). these objectives for a variety of reasons such as a lack of
Beth Kolko is an Associate Professor with the Department of Human
income, inappropriate services, little or no local content [8],
Centered Design & Engineering, University of Washington, Seattle, WA,
98195 USA. (e-mail: bkolko@u.washington.edu). inadequate infrastructure (power, roads, telecoms, etc.)
1 2
This paper is based on the results of an evaluation provided to the In fact, a successful telecenter program might be one that has had
Academy for Educational Development (AED) of the USAID funded eCenter significant impact (e.g. spun-off a few small businesses) but has no long-term
project in Kyrgyzstan in June 2007. sustainability (e.g. the program transitions and closes after a few years).
193
particularly in remote areas [9], lack of political, social, or model built upon the work of Heeks and Bhatnagar to help
managerial sustainability [10] or the challenge of identifying explain why projects that succeed initially, can still fail to
local and global partners that can move the telecenter to a enjoy long-term sustainability.
larger scale [11]. These challenges are enormous and have While the problems of sustainability are fairly well
often relegated telecenter initiatives to the “forever pilot documented (and theorized) the ultimate goals of community
syndrome” [11 pg.11]. impact (and the measurement and evaluation of such impact)
Thus, sustainability and impact remain the two principal may be less conclusively observed. To date most literature has
questions related to rural telecenters. In response, various focused on formative or process evaluation as opposed to
strategies have been developed to advance the sustainability summative or impact evaluations [21, 22]. And the literature
and impact of telecenter projects. One approach is the that does focus on social or economic impact assessments of
provision of subsidies for telecenter users. These are often telecenters to their broader community have yet to demonstrate
applied for a limited period and enable the use of ICT services an absolutely conclusive link (see [23] for a review).
by community members who might not have otherwise been From the above discussion and in tandem with previous
able to do so. While not necessarily a new approach, there is research, [see for example 5, 10, 24, 25], we have selected two
little research that examines the connection between user- main dimensions from which we examine sustainability:
based subsidies and sustainability or impact. The question that 1. Financial – This includes the financial independence,
we wish to explore in this paper is to what extent can user- business performance, and solvency of the project
based subsidy programs promote the sustainability and impact over time.
of telecenters. We do this by examining a user-subsidy coupon 2. Social – This refers to the equitable distribution of
scheme used by the USAID funded eCenter network in benefits among target groups, equal access and use,
Kyrgyzstan. and locally relevant content/services.
The rest of the paper is divided into several sections. First, In addition, we study the center’s development impact as it
we discuss issues surrounding the impact and sustainability of extends into the communities themselves. We can observe this
telecenters. We then articulate the methodology used in in both economic and social forms such as new educational
addressing our research question. Finally, we describe the opportunities, community empowerment, job creation, local
eCenter project, present our results and analyses based on our economic development, etc.
definition of sustainability and impact, and posit some Various initiatives have strived to manage the inter-related
conclusions. issues of impact and sustainability. The literature consists of
many general prescriptive reports and descriptive case studies
of such initiatives [see for example 8, 26-28]. Others have
II. THE IMPACT AND SUSTAINABILITY OF TELECENTERS reported on specific or novel approaches such as incorporating
As with any other development intervention, a critical business incubators into the telecenter and sharing resulting
objective is to make the telecenter sustainable. Typically, the profits with the telecenter organization [29]. Another approach
intention is to make a telecenter financially independent and is to provide subsidies to a telecenter initiative that is typically
solvent. This can be in terms of meeting maintenance costs, the applied to initial infrastructure or staffing costs [13].
recovery of initial investments, acquiring sufficient human Alternatively, user-based subsidies can lower the costs that
resources/staff, or adequate service delivery [12, 13]. the users themselves pay for services they seek. The hope is
Sustainability can also be viewed from a social point of view. that if well conceived, user-based subsidy programs can
Thus, sustainability hinges on having local champions, enhance both financial sustainability (by creating an early flow
sufficient community acceptance/awareness and involvement of income while developing a customer base over the long
in the running of centers, a range of users that is non- term) and social sustainability (by providing subsidies that
discriminatory and balanced, and beneficiary participation in target under-served or over-looked populations, for example
project design [4, 9]. Additionally, there are political factors to women). Additionally, if the subsidies are for activities that are
be considered such as accessing local and national political likely to lead to economic, social or political growth within the
support for the project [5]. This includes having a policy and community at large then the impact of the telecenter should,
regulatory environment which is conducive to the development ultimately, be enhanced.
of telecommunications infrastructure and a commercial This approach has been used in a variety of settings
internet market [14]. including the Cotahuasi Internet Cabina project in Peru in
Investigators have proposed a number of theoretical 1997. In that case, the use of the telecenter’s services by local
frameworks to improve our understanding of sustainability community leaders was paid for by donor funds [4]. Another
including the critical success factor (CSF) and critical failure example of user-based subsidies was the PC3 project in
factor (CFF) models [15, 16]; the ‘design-actuality’ [17] or Bulgaria. Pre-paid coupons were distributed in communities
‘design-reality’ gaps [18]; scenario analysis for long-term where the PC3 centers were located. The main goal was to
sustainability problems [19]; economic and financial promote the centers while quickly developing a client base.
sustainability models [20]; and political and institutional This helped to reduce the financial risk faced by the new PC3
models that underline the lack of commitment on the part of centers [30]. The eCenter project in Kyrgyzstan employed a
political leaders and public managers. Previous works by one similar logic in the use of coupons as an incentive both for the
of us (Best) [5, 10] have presented a sustainability failure local eCenters and clients.
194
Table 1 – Targeted and Actual sample size of user surveys by
eCenter
III. METHOD Targeted Sample Actual number of
We employed a mixed-method approach to study the effect eCenter Size collected user surveys
of the coupon program on the sustainability and impact of the Naryn 70 79
eCenters. Our research consisted of site visits to all telecenters Bosteri 50 40
that were part of the eCenter program; 7 in total. These visits Karakol 75 87
were done in collaboration with local researchers and took Karasuu 67 70
place between March and June 2007. Our research instruments Nookat 66 72
included a user survey and interviews with all the center Talas 22 27
managers, available members of their staff, and local Ivanovka 18 19
businesses. These data were supplemented by a review of user Total 369 394
logs, project reports and updates, and business proposal
guidelines.
The user survey consisted of sixty-two questions developed
around the issues of sustainability and impact. Specifically it IV. BACKGROUND – THE ECENTER PROJECT IN KYRGYZSTAN
explored how the computer training and Internet coupons were Kyrgyzstan is a small Central Asian country that was part of
used, the ways users engaged with ICTs at the eCenters, and the former Soviet Union. Although it is landlocked with
the perceived economic impact of using the centers. The limited resources, it has achieved economic growth of around
majority of the questions were close-ended with a few open- 4% between 2000-2005 [31]. Nevertheless, the poverty rate
ended questions to capture opinions on issues related to the continues to be of major concern. The national estimate of
centers. people living below the poverty line was 43% in 2005 with
To determine our sample size we estimated the overall larger percentages in rural areas [32]. In addition, the
population size as the number of users at each eCenter over the unemployment rate was estimated at 8.1% in 2005, with 13%
period for which data was available: January 2006 (start of the in urban areas [32]. This implies that a significant part of the
project) to January 2007. We defined users as those who population, though employed, is still poor.
participated in the subsidy (coupon) program of the centers for In 2002, the government approved a “National Strategy for
either Internet access or computer training. Given this ICT Development in the Kyrgyz Republic” as part of its plan
approach, we approximated the total user population at 9,497 to use ICTs to address development issues. In general, this
people and thus ensuring a confidence level of 95% required a emphasis follows what Ure [33] notes is a more open approach
sample size of 369 users. to the diffusion and use of ICTs by the Kyrgyz government
To identify subjects from each of the centers, user contact when compared to its neighbors. However, the growth of the
lists were obtained from center staff. Users were then Internet has been hampered by the monopoly held by the state
randomly selected from each list. In many cases, however, telecommunications company Kyrgyztelecom, which is deeply
these lists were incomplete with either missing contact in debt to the World Bank and seeking to return to
information or incorrect contact details. To account for this profitability. One consequence of this is that Internet
difficulty, research teams substituted or augmented random subscription costs are high relative to average incomes. Recent
sampling with subjects obtained by opportunistic sampling at estimates put Internet user rates at a relatively low level of
the eCenters during site visits. In addition, the researchers 13.3% [34]. In addition, PC ownership was estimated at 10.4%
worked through the social network of users at each site in [34]. Given the larger social and economic context, Internet
order to find both previous and current users of the eCenters. access via private, home-based means is not likely to grow
Thus, user surveys were first collected using standard random substantially anytime soon and outside the capital, public
sampling techniques when possible and then with a access is limited, expensive, and usually not fast. Given the
combination of convenience and snowball sampling. problem of poverty and the acknowledged importance of
In addition to sampling the appropriate number of users we information resources for economic development, the need to
also needed to ensure that our sampled subjects were provide alternatives to private use and to enhance public ICT
representative of the population across each center. To account venues was viewed as important.
for this we stratified the sample size according to the
proportion of users from each location. The eCenter Project
Table 1 gives the targeted sample size, and actual number of The eCenter project was launched in Kyrgyzstan in July
subjects surveyed, for all seven eCenters; in almost all cases 2005. It was funded through the Last Mile Initiative of the
we were able to over-sample the population. The one center United States Agency for International Development (USAID)
that was under-sampled (Bosteri) was closed for renovations that seeks to promote greater access to information and
during the data collection period, complicating researchers’ communication technologies, particularly in rural and
attempts to contact users. underserved areas. The goal of the project was to augment and
network a group of telecenters across the country with the aim
of promoting local economic development. Each telecenter
was established within a pre-existing business. In this way, the
195
Nookat
centers sought to improve local access to ICTs, stimulate local for those communities. These figures help to sketch the overall
business creation, improve computer skills, and increase competitive landscape of the centers. By looking at the number
opportunities in non-traditional employment training and job of employees we also get a sense of the relative size of each
creation among the local population. The local project establishment. Note that while this table focuses on Internet
manager in Kyrgyzstan was the Civil Initiative on Internet provision there were also other businesses providing related
Policy (CIIP)3, a Kyrgyz non-governmental organization that business services such as printing or photocopying.
focuses on the promotion of civil society interests in the
development of national ICT policy. Table 2 – Summary characteristics of the eCenters
The project ran from July 2005 to September 2006 with a total eCenter Number of Competing Population
budget of US$390,000. During this time, a group of seven Employees Internet Estimate -
eCenters were established across the country (see Fig.1): Centers 1999
1. Karakol Karakol 7 10 64,322
2. Bosteri Bosteri 4 0 12,000
3. Naryn
Ivanovka 6 0 20,000
4. Nookat
5. Karasuu/Osh4 Talas 3 2 32,638
6. Ivanovka Nookat 4 0 30,000
7. Talas Karasuu 3 0 19,143
(Osh) (80) (220,000)
Each of the eCenters provides a variety of fee-based Naryn 4 10 40,050
services including Internet access and email, printing,
scanning, copying, faxing, multi-media services and IP- User-based subsidies
telephony. The exact suite of services offered varies from There were two main components to the eCenter program.
center to center. Each center also delivers a curriculum of First, there were subsidies granted to the eCenters for the
computer literacy courses which consist of several modules: provision of computer literacy/accounting courses and second,
Windows, Microsoft Word, Microsoft Excel and using the free or discounted Internet access was offered to local
Internet. Additionally, some centers offer accounting courses community members. The subsidies for computer literacy or
and one offers leadership training. accounting courses came in the form of reimbursements to the
Table 2 lists the estimated number of competing Internet eCenter operator for the training costs of those who passed
cafés within each of the seven communities which had each course. People were able to participate in the courses by
eCenters. It also includes the total estimated population size receiving coupons. Similarly, Internet access was subsidized
through the use of five-hour coupons that were distributed to
3
See http://www.gipi.kg/ for an overview of CIIP. users. The subsidy level on both the computer training courses
4
The Karasuu eCenter moved to Osh city in April/May 2007.
196
and Internet coupons changed over time. Subsidy amounts The managers themselves were quick to praise the eCenter
were reduced 20% each quarter; thus while the project paid for program and saw a direct connection between their success
100% of the cost of a training course in the first quarter, by the and the use of the coupons. Some of the managers estimated
fourth quarter it paid only 20% of the cost and required the that 70-90% of their current customers would not have come
user to pay 80%. to the center without having been part of the coupon program.
Interestingly enough, a significant percentage of users did not
Partnering with local businesses actually rely on coupons. Of those surveyed, only 54% and
Each eCenter was established as part of an existing local 43% reported actually receiving coupons for Internet access
business rather than through the creation of new centers. and computer training respectively. This implies that perhaps
Suitable local business partners (including pre-existing the managers had an exaggerated belief of the effect of the
telecenters) were selected through a competitive bidding coupon program on their businesses. When asked if they
process based on criteria such as related prior business would have used the eCenter services if there were no
experience, existing level of investment, relevant telecom coupons, approximately 47% of respondents said yes and
experience, and the potential for further expansion of 19.5% said they would not have used the center in that case. It
programs. Subsequent to a center’s selection, CIIP provided should also be noted that the subsidies were only provided for
technical support on the use of the subsidy program and a limited period. Thus, by the end of the project users were
delivered training to the business owners and their staff. paying for the full cost of Internet access and computer
While some of the selected eCenter sites were already training.
providing Internet services, others were engaged in computer The existing competitive environment (Table 2) appeared to
graphics and photocopying services, gaming, computer repair influence the scale of impact of the coupons. For example, we
or providing computer training courses. found that if the manager identified more competitors in the
There was one final component of the eCenter project – a immediate community, the eCenter users were more aware of
land grant program where suitable land for investment in office and more likely to have used other cyber cafés. If there were
space and technological parks could be linked to each eCenter. fewer competing centers in the catchment then users were
It was envisioned that such investment would be supported by more likely to have used the eCenter prior to the coupon
the success of the eCenters; this component of the project, program. In other words, when the level of competition was
however, did not achieve expected outcomes and was heavy the coupon program was more successful in drawing in
eventually dropped. new users.
While there are several interesting aspects to the eCenter In general, 66% of those who received Internet coupons had
project as a whole, we have focused only on the user-based used the eCenter before while 33% of those who had received
subsidies in this study. The following sections present our a coupon had not used the eCenter before. The results are very
findings and analysis of the user-based subsidies component of similar for training coupons. Thus, during the subsidy period,
the eCenter project. coupon use was associated with the introduction of new users,
which can contribute to financial sustainability.
It should be noted that the decision to partner with existing
V. THE SUSTAINABILITY OF THE ECENTER PROJECT IN and successful local businesses was also important for the
KYRGYZSTAN5 success of the project. Generally, there is little evidence to
suggest that the one-off donor grant approach can work [24]
Financial sustainability and so some have argued for a business centered approach to
All the eCenters enthusiastically implemented the coupon telecenters [13]. The rationale behind this approach is that the
program and several have since sought to expand their offering profit motive can promote financial sustainability. In the case
of ICT services by increasing the scope and size of their of the eCenter project, it was in the interest of the local
businesses. In all cases, the businesses stated that their business partners to make sure that their centers remained in
financial situation was better off following the change to an operation and were successful. The coupon program supported
eCenter. On average, according to the managers this transition the partnership with local businesses by ensuring a sufficient
increased their revenues by an estimated 57% (with a low of number of customers during the initial subsidy period. It was
15% reported at Karakol and a high of 100% reported at critical therefore in “kick starting” the operations of the centers
Talas). In addition, almost all of the center managers agreed after which several of the local business partners were able to
that their clientele increased (some very significantly) after use this momentum to expand their businesses and tailor their
becoming an eCenter. The main sources of revenue varied suite of ICT services to their community needs.
from center to center and were based on local demand. For
example, Internet access was reported to be the main source of Social Sustainability – Types of Users
revenue in only two of the seven centers. Other revenue We examine social sustainability primarily in terms of the
sources included computer training, typing services, IP equitable distribution of benefits, access and use within the
telephony and the sale of mobile phone credit. community. Ideally, a more balanced and representative user
base should represent broader community acceptance of the
center and limited social exclusion. Table 3 below summarizes
5
Statistical results mentioned in this paper are all significant to at least the the general user distribution across the various centers. We
0.05 level.
197
found that the majority of users were women except in Nookat been exploited completely. Thus while the eCenters have
where the more traditional culture in the south seems to limit financially benefited from the coupon program; the long term
female participation at the eCenters. The majority of users in social sustainability of the project may have not.
the sample were typically young, suggesting the propensity to Two factors could help explain this type of coupon
use ICTs is linked to age particularly where the general distribution among users. First, the implied goal of the project
diffusion of such ICTs is low (i.e. rural areas). This young user was to focus on persons who had the least access to the
group is consistent with a nationally young population (31% Internet and computer courses. However the way the coupon
below 15 years in 2005 - [35]) and also partly explains the low program was operationalized did not appear to be systematic
marriage rates among our sampled population. In general, or consistent across centers. Moreover, the eCenters received
older customers used more of the basic ICT services such as no particular incentives to narrowly target users and so some
photocopying or printing and requested the assistance of may have simply distributed their coupons to people already at
younger persons or staff to help them when using the Internet. their center. Thus, there was a significant difference among
Older users (above thirty years) were, however, just as likely receivers of both Internet and training coupons from center to
to attend training courses. The education rates are similar to center though there was no discernable pattern to these
those nationally. Also, the user distribution was approximately differences.
proportional to the religious and ethnic composition of each Some centers seemed to offer coupons to whoever was
community easily available. Indeed, some were more concerned with
issuing the coupons rather than targeting and as such relied on
Table 3 – Summary of user characteristics informal means of distribution. One example of this is a
eCenter Male Female Avg. Bachelors Married neighboring Internet café owner in Karakol who said he
% % age or higher % received a training coupon for the computer literacy course.
% He was already exposed to more advanced computer courses
Karakol 44 56 21 48 10 but still chose to attend the eCenter course since it was free at
Bosteri 25 75 18 10 3 that time. Alternatively, more formal means of distribution
were used at other centers. For example, in Naryn the intention
Ivanovka 53 47 18 5 0
was to target as wide an audience as possible. This was done
Talas 33 67 23 15 19 through advertising in local media and actually did result in
Nookat 67 33 24 24 39 lines literally going out of the door of the eCenter. However, it
Karasuu 41 59 21 30 14 is unclear what type of users this form of marketing attracted.
Naryn 31 69 24 47 25 Second, the eCenters were encouraged but not compelled to
Total for give one Internet coupon per person, so as to increase the
all users 43 57 21 32 19 reach of the program. This suggestion, however, was not
generally followed. For example, in Nookat users were given
one coupon per quarter since, according to the manager there,
Social Sustainability – Targeting and Distribution of Coupons
it was difficult to continually find new Internet users.
Based on the above characteristics, we found few
Alternatively, in Karasuu the strategy was one coupon per
statistically significant differences among users who received
individual. Thus both targeting and distribution seemed to be
coupons and those who did not. Among those who received
dependent on the individual eCenters and was the result of a
Internet coupons approximately 51% were female. This could
compromise required to enable a private business to undertake
have had a marginal effect on the larger proportion of female
socially oriented goals.
users at the centers overall. Other variation in user
characteristics (ethnicity, educational level, etc.) among those
users who benefited from the coupon program and those who
VI. DEVELOPMENT IMPACTS OF THE COUPON PROGRAM
did not were not statistically significant. Also, there was no
difference between those who got Internet coupons and those Internet Access
who did not in terms of their perceptions of the importance of Although this represents a preliminary and indirect impact
the Internet to their jobs or schools. In terms of the training of the program, the Internet coupons did, of course, increase
courses there were also no discernable demographic access to the net among users. This access can be viewed as a
differences between those who received these coupons and form of impact. Indeed, in most cases overall access increased
those who did not. in participating communities. For example, several of the
What was different was whether or not the respondent had centers are located in rural areas with limited access to ICTs.
used the eCenter prior to receiving the coupon. As mentioned Thus, the introduction of the eCenters helped to meet the latent
above, 66% of those who received Internet coupons had used local demand for ICT services and training. Given the
the eCenter before. Similarly, 60% of those who received relatively high subscription costs for the Internet, this also
training coupons had already visited the eCenter. This suggests meant that the eCenter could have been the only source of
that prior users were in a better position to learn about the Internet access for some users. This was the case in at least
program and participate. The concern is that while the coupon three of the communities.
component of the project is now complete, the opportunity to For many, the communication function of the Internet was
attract larger numbers of new users to ICTs might not have most important, particularly where regular post mail services
198
were infrequent. The most common online activities were Development Impact – Local economic development
sending/receiving emails and instant messaging. Other In terms of local economic development, we observed two
important uses included reading news, doing school related types of impacts related to the introduction of user-subsidies at
work/research, downloading music and participating in chat the eCenters. First, the eCenters have stimulated the local
rooms. Several users proffered experiences of how accessing market for ICT services. For instance, there are approximately
the Internet had helped them keep in touch with friends and ten other commercial computer centers in Naryn that were
family abroad, support their work and even religious established after the eCenter opened its doors. Second, the
education. opposite is also possible. In Nookat, prior to the introduction
Finally, in addition to providing increased quality or of the eCenter there, there was one other business offering
availability of Internet access, the eCenters have also been an Internet services. This closed soon after the opening of the
important place for users to gain Internet-related skills. eCenter as it could not compete with the initial subsidies being
Approximately 70% of those who received Internet coupons provided for Internet access. That other business now operates
said that they had learned to use the Internet at their eCenter. as a café.
One of the objectives of the eCenter project was to support
Development Impact – Computer training local business development. Of all users, only 5% said that
The acquisition of relevant skills is a clearer form of they were able to utilize either the facilities or courses at the
community development impact and the eCenters provided eCenter to establish a new business. Most of the businesses
computer training courses which were well received by were located in the services sector and a few specifically in the
participants. Perhaps more than anything else, participants in ICT sector. Of these users, there were no significant
training courses were quick to inform us of the benefits of differences in terms of gender or levels of education. They
receiving the computer literacy training. The courses were were not necessarily in a higher income group either; for
basic and allowed the centers to bring those with no example they were no more likely to own a car than those who
understanding or experience with computers up to at least a did not start a business. Finally, in terms of coupon use, there
foundational level of use. Completion was contingent on was again relatively little difference between those who
passing an evaluation which tested their ability to use the reported starting a new business and those who did not.
computer and some of the Microsoft applications they had Whether we refer to the efficacy of the training courses or
studied. The majority of trainees were able to pass the test; the improved Internet access, one qualifying factor to note is the
failure rate was approximately 10%. For many users there was initial limited diffusion of ICTs in the targeted communities.
a sense of pride in completing the course. This initial lack of ICTs in these mostly rural/semi-urban
For many, these courses provided the basis for the communities meant that the potential for the subsidies to have
acquisition of job-related ICT skills. Eighty-five percent of all some modicum of impact was there from the start. We cannot
respondents stated that they acquired important job skills from be sure whether the development impact would be similar in
their eCenter. Furthermore, some users reported actually other contexts. However, we can indicatively point to the
getting jobs (usually clerical) based on these skills. example of the eCenter in Osh city where there are some 80
Approximately 15% of all users reported finding a job as a other cybercafés, various documentation centers, and in
result of using the eCenter. As noted earlier, the country has a general better infrastructure for the delivery of ICTs. In an
high national poverty rate of 43% and an unemployment rate environment such as this, the eCenter modified its strategy to
of 8.1% (2005). Therefore such impacts from the program, focus on the provision of training courses as there were few
though small, will be valuable to both the unemployed and the organizations providing this service. Thus, the business focus
working poor. We did not find, however, a link between users of the eCenters coupled with the goal of providing relevant
who participated in the coupon program and those who ICT services to targeted communities can potentially create an
reported finding jobs due to training at the center. impact in a variety of contexts.
Approximately 61% of trainees who received user-subsidy
coupons for their classes said they would pay for the service if
there was no subsidy, demonstrating the value placed on the VII. SUMMARY AND CONCLUSIONS
services and experience provided by the eCenters. Many of the The eCenter program provided subsidized coupons for
eCenters are in fact planning to continue and even expand their computer courses and Internet access to users of already
course offerings past the lifespan of the subsidy program. For established commercial computer centers. Thus, the center
at least one eCenter this is also the main source of revenue. owners provided the physical infrastructure for each eCenter
Content was an important factor in promoting the demand for and the subsidies helped them to access, and ostensibly expand
the computer literacy courses. Much of the design and content the local market for training and Internet services. The project,
for the courses came from the initial demand studies done therefore, represented a merger between public goals of
prior to the start of the project. However, a few users increasing ICT access and literacy and private goals of
suggested that having completed these courses they now want increasing profit and market share.
the center to offer more advanced computer courses. We focused on the subsidy component of this project, as an
Continually modifying content to meet the needs of users will innovative approach to telecenter development. As such we
be a factor in supporting long term development impact. attempt to address a gap in current research by exploring the
relationship between user-based subsidies and the
199
sustainability and impact of telecenter initiatives. In this case to grow their businesses. Future user-subsidy programs should
we examined sustainability in terms of its financial and social examine ways to incentivize the local center managers to
dimensions. By financial sustainability we look at an eCenter’s design and implement targeting programs that enhance social
financial independence and solvency; by social sustainability and financial sustainability as well as overall community
we are primarily concerned with the distribution of access and impact. Additionally, some broad parameters and techniques to
benefits to the community. There are of course other this targeting could be developed centrally and offered to
dimensions that could have been included, but we wanted to participating center managers.
limit our analysis in such a way as to make it focused and Even with decades of worldwide experience constructing,
useful. operating and evaluating telecenters, the research reported
In terms of financial sustainability, we found that the here makes clear that the often sited preeminent goals of the
coupon program brought some new users to the centers during telecenter movement – socio-economic sustainability and
the subsidy period. Self-reporting indicates that a third of substantial community impact – remain difficult to obtain.
Internet coupon users and 40% of the training coupon users Further experimentation with models and approaches, along
were new. More importantly, 19.5% of users stated that they with close observation and continued independent assessment
would have not have used the services of the eCenter without work, is required if we are to realize these ultimate ambitions.
the coupons. The coupons also helped the local businesses
thrive with customers during the initial subsidy period; it was ACKNOWLEDGMENTS
critical, therefore, in “kick starting” the operations of the We thank Medina Aitieva and her students from the American
centers. Thus, we argue that the coupons have to some degree University of Central Asia, members of our local research
aided financial sustainability by bringing new users (and team. We also thank the three anonymous reviewers for their
therefore additional revenue and an expanded long-term helpful inputs.
customer base) to the centers and creating a stable source of
income during the initial stage of operations. Obviously,
temporary subsidy programs may actually weaken long-term
REFERENCES
financial self-sustainability as the program winds down. But in
this case, while we observed the end of the program we did not 1. Siochrú, S.Ó., Global Governance Of Information And
find evidence for an immediate financial downturn. Communication Technologies: Implications For Transnational
In terms of social sustainability, the distribution of the Civil Society Networking. 2003, Social Science Research Council:
coupons did not lead to any significant differences in the type New York.
2. Gómez, R. and P. Hunt, eds. Telecentre Evaluation - An
of users except a marginal increase in female participation. We
International Perspective : Report of an International Meeting on
find that the distribution of both Internet and training coupons Telecentre Evaluation. 1999, International Development Research
favored more regular users of the eCenters. This implies that Centre (IDRC): Montreal.
the opportunity to expose as many new users as possible to 3. Fuchs, R., Little Engines that Did: Case Histories from the Global
ICTs may have been missed and this could have limited the Telecentre Movement. 1998, IDRC: Ottawa.
4. Proenza, F.J., R. Bastidas-Buch, and G. Montero, Telecenters for
wider social impact of the project. One of the main reasons for Socioeconomic and Rural Development in Latin America and the
this was the lack of effective targeting strategies among the Caribbean. 2001, Inter-American Development Bank: Washington
centers. DC.
Finally, the coupon program had a limited development 5. Kumar, R. and M. Best, Impact and Sustainability of E-
Government Services in Developing Countries: Lessons Learned
impact on participating communities. One preliminary impact from Tamil Nadu, India. Information Society, 2006. 22(1): p. 1-
was the enabling of Internet access for users and the learning 12.
of Internet related skills. The coupons also facilitated 6. Mercer, C., Telecentres and transformations: Modernizing
computer skills training. However, among those who reported Tanzania through the internet. African Affairs, 2006. 105(419):
p. 243-264.
economic benefits (subsequent employment or starting a new 7. Dagron, A.G., Prometheus Riding a Cadillac? Telecentres as the
business) after acquiring these skills, there was little or no promised flame of knowledge. Journal of Development
difference between those who had been coupon recipients and Communication: Special Issue on Telecentres 2001. 12(2).
just regular users who had not availed of the program. 8. Colle, R., Memo to Telecenter Planners. Electronic Journal of
Information Systems in Developing Countries, 2005. 21(1): p. 1-
We suspect that as the targeting and distribution was left up 13.
to the local eCenters, there might have been less emphasis on 9. Roman, R. and R.D. Colle, Themes and Issues in Telecentre
sustainability and impact and more emphasis on short-term Sustainability, in Development Informatics Working Paper Series.
profit. We argue that both the sustainability and impact of the 2002, Institute for Development Policy and Management,
University of Manchester.: Manchester, UK.
coupon program could have been improved with a more
10. Best, M.L. and R. Kumar, Sustainability Failure of Rural
focused targeting strategy. For example, better social targeting Telecenters: The Sustainable Access in Rural India Project.
of the coupons could have included more women, particularly Information Technologies and International Development, 2008.
in the more traditional areas of the country, or other vulnerable 4(4): p. 31-45.
groups. From a financial sustainability point of view, targeting 11. Fillip, B. and D. Foote, Making the Connection - Scaling
Telecenters for Development. 2007, Washington, DC: Academy
new users could have also been more effective. In terms of for Education Development.
impact, coupons could have been distributed, for example, 12. Harris, R.W., A. Kumar, and V. Balaji, Sustainable Telecentres?
specifically to young entrepreneurs looking for skills or access Two Cases from India. 2000, Infobridge.
200
13. Proenza, F.J., Telecenter Sustainability - Myths and 33. Ure, J., ICT Sector Development in Five Central Asian
Opportunities. Journal of Development Communications, 2001. Economies: A Policy Framework for Effective Investment
12. Promotion and Facilitation, in International Conference on
14. Whyte, A., Understanding the Role of Community Telecentres in Strengthening Regional Cooperation for Managing
Development - A Proposed Approach to Evaluation, in Telecentre Globalization,. 2005: Moscow, Russian Federation.
Evaluation - An International Perspective : Report of an 34. CAICT. 2007 CAICT Survey: Topline Results 2007 [Retrieved:
International Meeting on Telecentre Evaluation, R. Gómez and P. September 5, 2008]; Available from:
Hunt, Editors. 1999, International Development Research Centre http://depts.washington.edu/caict/surveys/2007toplines/TopLines_
(IDRC): Montreal. 2007.pdf.
15. Heeks, R. and S. Bhatnagar, Understanding success and failure in 35. UNDP, Human Development Report 2007/2008: Fighting climate
information age reform, in Reinventing government in the change: Human solidarity in a divided world. 2007, Palgrave
information age, R. Heeks, Editor. 1999, Routledge: London. p. Macmillian: New York.
49-74.
16. Heeks, R., Understanding e-Governance for Development, in i-
Government Working Paper Series No. 11. 2001, Institute for
Development Policy and Management: University of Manchester,
UK.
17. Heeks, R., Information Systems and Developing Countries:
Failure, Success, and Local Improvisations. The Information
Society, 2002. 18(2): p. 101 - 112.
18. Heeks, R., Most eGovernment-for-Development Projects Fail:
How Can Risks be Reduced?, in IDPM i-Government Working
Paper no.14. 2003, Institute for Development Policy and
Management: University of Manchester, UK.
19. Aichholzer, G., Scenarios of e-Government in 2010 and
implications for strategy design Electronic Journal of e-
Government 2004. 2(1): p. 1-10.
20. Best, M. and C. Maclay, Community Internet Access in Rural
Areas: Solving the Economic Sustainability Puzzle, in The Global
Information Technology Report 2001-2002: Readiness for the
Networked World. 2002, Oxford University Press: Oxford.
21. Hudson, H.E., From Rural Village to Global Village:
Telecommunications for Development in the Information Age.
2006, Mahwah, New Jersey: Lawrence Erlbaum Associates.
22. University of Washington Center for Internet Studies. CIS
Literature Review Database on the Impact of Public Access to
ICT. 2008 [Retrieved: September 9, 2008]; Available from:
http://references.cisresearch.org/index.php.
23. Kuriyan, R. and K. Toyama. Review of Research on Rural PC
Kiosks. 2007 [Retrieved: September 16, 2008]; Available from:
http://research.microsoft.com/en-
us/um/india/projects/ruralkiosks/kiosks research.doc.
24. Harris, R.W., A. Kumar, and V. Balaji, Sustainable Telecentres?
Two Cases from India, in The Digital Challenge : Information
Technology in the Development Context, S. Krishna and S.
Madon, Editors. 2003, Ashgate: Aldershot, Hants, England ;
Burlington, VT, USA. p. 124-135.
25. Bailur, S., Using Stakeholder Theory to Analyze Telecenter
Projects. Information Technologies & International Development,
2007. 3(3): p. 61-80.
26. Jensen, M. and A. Esterhuysen, The Community Telecentre
Cookbook For Africa - Recipes For Self-Sustainability. 2001,
United Nations Educational Scientific and Cultural Organization:
Paris.
27. UNDP, Telecottage Handbook: How to establish and run a
successful telecentre. 2006, UNDP, Europe and the CIS:
Bratislava.
28. Badshah, A., S. Khan, and M. Garrido, eds. Connected for
Development: Information Kiosks and Sustainability. United
Nations Information and Communication Technologies Task Force
Series. Vol. 4. 2003, United Nations Information and
Communication Technologies Task Force: New York.
29. Khelladi, Y., What Works:The Infocentros Telecenter Model, in A
Digital Dividend Study. 2001, The World Resources Institute:
Washington DC.
30. Tifft, J.D., Summative Evaluation - Bulgaria Public Computer
and Communication Centers (PC3s). 2003, USAID: Washington
DC.
31. IBRD, Joint Country Support Strategy (JCSS) for the Kyrgyz
Republic (2007-2010). 2007, The World Bank: Almaty.
32. IBRD, Kyrgyz Republic Poverty Assessment. 2007, The World
Bank: Washington DC.
201
Manuscript received September 23, 2008. This work was supported in part II. INDIAN LANGUAGE BRAILLE ENCODING SYSTEM
by Communication Empowerment Laboratory, Indian Institute of Technology,
Kharagpur and Media Lab Asia, Mumbai. The Bharati Braille system is the standard technique of
T. Dasgupta is with the Indian Institute of Technology, Kharagpur, West representing Indian language texts to Braille [1]. The system
Bengal India (corresponding author to provide phone: +91 9433852325; e- uses 6 dot cells to represent each character. The combination
mail: iamtirthankar@ gmail.com).
A. Basu is with the Computer Science and Engineering Department, Indian of these 6 dots can generate 26-1=63 different Braille
Institute of Technology, Kharagpur, India (e-mail: anupambas@gmail.com). characters.
1
Official language of Bhutan
202
It has been observed that for all Indian languages, the 2) The system does not provide any auditory feedback to the
corresponding Braille code is the same. In other words, a user neither it gives a very good GUI for visualizing the
distinct Braille cell may correspond to different Indian transliterated output.
language character. However, a single Indian language 3) Editing of the transliterated output is very difficult and
character may require one or more than one Braille cell for its the system does not provide any interface to create or edit
representation. This process is illustrated in Table 1. input text document.
4) Working of the system depends heavily upon the user’s
Table 1: Braille representation of different Indian language
characters. knowledge, understanding and intelligence. As a result,
the system provides high cognitive load to the user which
Braille Bengali Hindi may result in increasing error rate.
Apart from computer based transliteration of Braille, a Braille
k ক क writing tutor system has been presented in [14]. The prototype
tutor system uses an E-slate device to capture a student’s
action and tries to develop the Braille writing skills.
. খ ख
IV. THE ILBT SYSTEM ARCHITECTURE
" R ঋ ऋ The architecture of our proposed Indian language text to
Braille transliteration (ILBT) system is shown in Fig. 1. The
diagram shows the essential components of the present
One of the most interesting phenomena found in Indian system. There are two major goals of the system:
language scripts is the usage of composite or conjugate • Providing audio feedback to the Indian language text.
characters. Conjugates are often constructed by the sequential
• Forward or reverse transliteration of text documents.
concatenation of two or more characters with a special
character called halant. The construction of conjugate The input to the system may be a text document written in
characters follows certain rules. Some of these rules are any popular word processor, or Indian language texts entered
illustrated in Table 2. through a keyboard. Based on the user’s requirement the
Table 2: Conjugate construction rules with examples taken from
system can generate either a speech output or can transliterate
Bengali language (C=Consonant, H=Halant, V=Vowel) the text to Braille. The details of the major components of the
system are discussed in the subsequent sections.
Rule Example Braille
A. The Text-To-Speech Engine
CHC ঙ+◌্ +ক = @+K A text-to-speech (TTS) engine is at the core of the speech
output generation module. It is converts strings of text to their
CHCV ক+ ◌্ + র+ ি◌ = িk @KRI corresponding human voice equivalent. In order to develop
the proposed transliteration architecture, we used a text-to-
CHCHC স+ ◌্ + ত+ ◌্ + র=st @S@TR speech engine called Shruti [15] that uses diphone
concatenation to synthesize speech in Indian languages using.
স+ ◌্ + ত+ ◌্ + র+ ◌্ + ◌ী The Shruti text-to-speech engine currently supports two
CHCHCV
= stী @S@TR9
Indian languages namely Hindi and Bengali.
We built a wrapper around the TTS engine so that it can be
used by a number of applications including our transliteration
From Table 2 it can be observed that the conjugate
system. The wrapper also provides a number of other vital
characters, as constructed by clustering of consonants and
features like
vowels, may have an entirely different visual representation.
1) Conversion of text in Unicode or iTrans to ISCII 2 which
However, the corresponding transliterated Braille is
is the native representation format of the TTS being used,
represented by a sequence of Braille cells for each of the
2) Pronunciations of individual and conjugate characters.
characters.
B. Input Document Type & Keyboard Input Methods
III. RELATED WORKS In order to create an Indian language document, a text editor is
The Sparsha toolset [2] is the only system encountered by required for accepting the regional language text entered
us so far that can transliterate Indian language texts to Braille. through the keyboard and perform different operations on it,
The system provides some unique features like transliteration like formatting, printing, and saving the text. Further, it is also
of mathematical symbols and tactile graphics to Braille. desirable that the editor should support different font
However, this system suffers from several limitations like, encodings (like, Unicode, ISCII, ASCII, and iTrans) which
1) The system takes significant amount of time during
transliteration of large texts. 2
Indian Standard Code for Information Exchange
203
are
User I/P Through Keyboard Keyboard
Keyboard/Mouse Input Key
Input Hook
Methods
Operating
Regional Font Rendering System Mouse
Language Definitio Engine Hook
Fonts
Glyphs
TTS
Embosser
Interface Engine
Fig. 1: System Architecture of the Speech enabled Indian Language Text to Braille Transliteration System
popularly used to create Indian language text documents. representation of any regional language text. This is
There exists no such editor that can support all the above especially important for Indian languages. Though this may
mentioned list of font encodings together. For example MS- not seem important for a blind user, it is actually essential
Word 2007 supports Unicode as well as ASCII encoding for him in order to communicate with other sighted people.
and the most common editor supporting ISCII is the iLeap The OpenType font format provides a large number of
editor. A solution to this problem is creating a new text features essential for supporting regional languages,
editor that can support all the mentioned font encodings. especially complex scripts [16]. Due to this reason it was
However, the developing a new text editor will require a found to be suitable for the mentioned purpose. In
substantial time and effort. Further, it expects a user to OpenType fonts all information related to the proper
learn the new system which is very difficult especially if rendering of the font may be included within the font file.
the user is visually impaired. Hence, in our proposed GSUB (Glyph Substitution) and GPOS (Glyph Positioning)
architecture we allow a user to create a text content in any are two tables responsible for storing such information.
of the commercially available regional language text editors These tables play a critical role in displaying regional
available in India. However, In case of unavailability of language text as, a lot of character reordering, combination
such kind of editors, our system provides a very basic form of characters and glyph reshaping may be necessary. Glyph
of a regional language text editor that is integrated along substitution is the process of replacement of one or more
with the ILBT system. Currently the editor provides some glyphs by a new glyph like while writing conjugates. Glyph
basic functionalities as compared to a full-fledged text positioning refers to process of shifting a glyph to produce
editors like, creating a new document, opening of the proper visual representation (E.g. while writing vowels
documents in multiple tabs, saving documents, searching in Indian languages called matras) as shown in Fig. 4.
regional language texts etc. However, though the visual representation of the text
changes in the above cases the original string of characters
C. Indian Language Font Rendering
remains unchanged. Both the above mentioned tables find
Proper Indian language fonts and the font rendering extensive use especially in Indian languages
engines are necessary for producing correct visual
204
D. Input and Output of the system in Fig. 3. From the figure we can see that if a character is
The proposed ILBT system can accept English text in the preceded by a halant then during transliteration, position of
form of plain text files and Microsoft Word documents. the character and the halant gets swapped. This
Apart from English the Braille transliteration system, as phenomenon can be represented by the rule:
described, can take Indian language text as input. The input
text document of the system can be of any type like: Precede(X, halant) Æ swap(X, halant)
• Unicode text – generated by any standard editor
supporting Unicode [17]. Another interesting feature found in Indian language text
transliteration is the handling of matras. A vowel may
• ISCII (Indian Script Code for Information occur either as a matra or it may appear as a distinct vowel.
Interchange) [18] documents generated by If the vowel occurs as a matra then a special symbol
applications like iLeap [19]
appears before the vowel else no special symbol appears.
The output of the ILBT system can be of two types. A This is illustrated in Fig. 4.
user can select a particular section of the text document to
get the corresponding Indian language speech output, or a
user can obtain the corresponding Braille output on a large
0981 0027 0045 090F 0947 0035
variety of commercially available Braille embossers 0982 003B 0046 0036 0000 0036
[20][21]. The current system has been tested on the 0983 002C 0047 0917 0000 0037
following Braille embossers: 0985 0041 0048 0038 0000 0038
• Index Basic-S 0986 003E 0049 0907 093F 0039
• Index Basic-D
a b
• Index 4X4 PRO
• Romeo Pro 50 Fig. 2: Code Table Structure of both Forward (a) and Reverse
• Juliet Pro (b) Transliteration Engine
E. Forward Transliteration
The input to the transliteration unit is either an English
text or an Indian language text document. The text স + ◌্ + ত + ◌্ + র + ◌ী = stী
documents are either based on Unicode, ASCII or ISCII
encodings.
As discussed in the previous sections, different Indian
language characters shares the same Braille representation.
In other words mapping rules to transliterate one Indian @ +S +@+ T+ R+9
language can be simultaneously used to transliterate any
other Indian language text to Braille. Thus, in order to Fig. 3: Transliteration of Bengali conjugate characters.
implement this, we build separate code tables for each of
the languages and based on the input language the
corresponding code table is used. An example of the code
table structure is shown in Fig. 2.
The code tables are stored in a hash table due to which
mapping of input text to Braille takes a constant amount of
ক + ি◌ = িক ক + i = কi
time resulting in a very first transliteration output. The said
. I . A I
method of implementation also makes the system highly
scalable and allows the inclusion of more languages in Special Character
future if required.
Transliteration of English text to Braille is a relatively Fig. 4: Illustration of Rule that Handles Vowels and Matras
straight forward process, where there is a direct mapping There are number of other rules present which are required
between the English character and the Braille character. for the Indian language text to Braille transliteration. The
However, Indian language text to Braille transliteration is rule files are written in such a way that a new rule can be
not as straight forward as English. Transliteration rule for a added at any point of time without disturbing the original
particular Indian language character depends upon its system.
previous and the next character. This process is illustrated
205
transliteration Error
4) School for the blind, Dibrughar 8 8
5) Curriculum and Professional Support Division (CAPSD)
Number of
6 6 6
Ministry of Education, Bhutan
As a result of this deployment, the system is undergoing 4 4
3
4
186 181
File Characters Text type
150
135 135
Doc 1 400 Only Bengali text 100 103
117 123
93
71
Doc 2 1000 Only Hindi text 50 46
62
These Braille documents are read by a blind person and the Fig. 13: Comparison of the computation time taken for Indian
transliteration errors are marked manually by the users. We language Braille transliteration in Sparsha and ILBT.
use the existing Sparsha toolset as a lower level baseline for
our system. The transliteration output as generated by the two From the result it can be noted that the ILBT system
systems is compared automatically by a program. The performs considerably better than the Sparsha toolset. The
comparison is based on two parameters a) transliteration reason behind such a marginal difference may be due to the
similarity and b) Output alignment format. Fig. 12 shows the usage of separate hash table for each of the Indian languages.
result of the above experiment. We can observe that the The usage of hash table reduces the searching time complexity
transliteration accuracy of ILBT system is better than the to a constant time as a result the overall transliteration time
Sparsha system. Most of the errors are occurred due to some reduces drastically.
Operations Time Move the mouse to the Bengali menu item (P) 1.1
(in sec.) Click on the Bengali menu item (BB) 2*.1=.2
Key press and release (K) 0.28 Double click to select the file from the open 4*.1=.4
Move the mouse to an object on screen (P) 1.1 dialog box (BBBB)
Hand from keyboard to mouse or vice versa (H) 0.4 Table 7: Action sequence for “Mouse Movement and Selection”
method for Sparsha system
Mental Preparation (M) 1.2
Action Performed Time
Type string of n characters (T(n)) n*K (in sec)
Move the mouse to the file menu (P) 1.1
We define the task of a user as “Transliterate a text document
and view the print preview of the Braille output”. To Click the file menu (BB) 2*.1=.2
accomplish this task, the ILBT system offers the following 1.1
Move the mouse to the open menu (P)
thee different methods:
• Mouse Movement and Selection Click the open menu (BB) 2*.1=.2
• Mouse Click Operations Double click to select the file from the open dialog 4*.1=.4
Table 5 and Table 6 show the different action sequences and box (BBBB)
the corresponding estimated time of the Mouse Movement and
selection method for the ILBT and Sparsha system. One of the Move the mouse to the language selection dialog 1.1
key observations we made from table 5 is that the mouse box (P)
pointer movement operation is taking around 0.9 sec higher
Click on the language bar to select language (BB) 2*.1=.2
time that the mouse click operations. Hence, we tried to
enhance our present method of Mouse Movement and Click OK button to select input text language (BB) 2*.1=.2
Selection so that we can reduce the number of mouse
Click OK button of the message box (BB) 2*.1=.2
movement operations. Consequently, we came up with a
solution by which a user can perform most of the operations Move the mouse to the transliterate menu (P) 1.1
on the ILBT system with the help of a three button mouse.
The various mouse click operations and their corresponding Click the transliterate menu (BB) 2*.1=.2
GUI commands are explained in Table 3 of section V. Move the mouse to the “text-Braille” button (P) 1.1
We try to perform the same task of “Transliterating a text
document and view the print preview of the Braille output”. Click the “text-Braille” (BB) 2*.1=.2
This task has been performed using the Mouse Click Move the mouse to the OK button of the 1.1
Operations method as explained above. Table 7 presents the transliteration dialog box (P)
different actions and the time required to accomplish the given
transliteration task. Move mouse to the OK button of the message box 1.1
From Table 8 we can see that accessing the system with the (P)
help of mouse click events considerably reduces the overall Click the OK Button (BB) 2*.1=.2
execution time of a user. We summarize the third level of
evaluation between the Sparsha toolset and the ILBT system Move the mouse to the print preview menu (P) 1.1
in Table 9. Click the print preview menu to get the preview of 2*.1=.2
Braille document (BB)
Estimated Time 11
211
Table 8: Actions sequences and estimated time for the “Mouse person.
Click Operation” method
Abstract—There is a wealth of information collected about having more information is definitely better, understanding and
national level socio-economic indicators across all countries each visualizing this data becomes a harder problem.
year. These indicators are important in recognizing the level of Many websites are utilizing this large collection of socio-
development in certain aspects of a particular country, and are
also essential in international policy making. However with past economic indicator data to visualize global inequality. Popular
data spanning several decades and many hundreds of indicators websites include CISEIN [25], Gapminder [9], NationMaster
evaluated, trying to get an intuitive sense of this data has in [17], UC Atlas [24], and WorldMapper [5]. These websites
a way become more difficult. This is because simple indicator- utilize a number of classic visualization techniques including
wise visualization of data such as line/bar graphs or scatter line graphs, bar graphs, scatter plots, and geographic maps to
plots does not do a very good job of analyzing the underlying
associations or behavior. Therefore most of the socio-economic allow users to view this raw data in different ways. The tem-
analysis regarding development tends to be focused on few main poral data is almost always visualized using animation. These
economic indicators. However, we believe that there are valuable visualizations take the first step to allow users to investigate
insights to be gained from understanding how the multitude of a variety of questions: How does one country compare with
social, economic, educational and health indicators relate to each other countries in the same geographic region or with similar
other.
The focus of our work is to provide an integration of statistical GDP? How are different socio-economic indicators related to
analysis with visualization to gain new socio-economic insights each other? What policies can be implemented to improve
and knowledge. We compute correlation and linear regression health nationally and globally? However, the simple indicator-
between indicators using time-series data. We cluster countries wise visualization of data falls short of providing a deeper
based on indicator trends and analyze the results of the clustering understanding of associations between various indicators and
to identify similarities and anomalies. The results are shown on
a correlation or regression grid and can be visualized on a world countries.
map using a flexible interactive visualization system. In this work, a team of computer scientists and sociolo-
This work provides a pathway to exploring deeper relation- gists have worked together to create a novel integration of
ships between socio-economic indicators and countries in the statistical tools and visualization with a view to gain new
hands of the user, and carries the potential for identifying socio-economic knowledge. Our goal is to leverage mostly
important underpinnings of policy changes.
the familiar and well-known statistical (correlation, linear re-
gression, and clustering) and visualization techniques (scatter
I. I NTRODUCTION AND M OTIVATION
plots and geographic maps) to investigate deeper relationships
Visualization for the purpose of providing intuitive and between socio-economic indicators and countries. Is the in-
deeper understanding of global inequality is an important tuitive understanding provided by raw indicator visualization
problem. Several websites supporting these visualizations us- supported by the results of correlation and linear regression
ing raw data are becoming increasingly popular. However, analysis? Are the causality claims obtained through complex
seemingly easy to understand relationships between variables multi-regression models, often used in socio-economic lit-
visualized using line graphs, bar graphs, and scatter plots, erature, validated or contrasted by correlation or regression
can sometimes provide incomplete information and may even analysis? We view our system as a first step towards building
lead to misleading or erroneous conclusions. In this work, a bridge between the simple approach of using a raw indicator
we propose using statistical tools combined with visualization visualization and the high-powered causality or other policy-
to provide a deeper and more complete understanding of based models.
relationships between the global socio-economic indicators. Our system features an easy-to-use interface where the user
There is a large amount of data collected across all countries can interactively select and visualize multiple countries and /
annually over a range of socio-economic indicators by various or indicators. We have coupled it with the Globalization-Health
agencies including World Bank [2], United Nations [26], Nexus Database [21] to analyze the relationship between
UNESCO [7], and [18] . For example the World Development various health indicators. Furthermore, we have contrasted the
Indicators Database [2] has data that covers 225 countries and observations of both raw and statistical visualizations with the
regions, spanning 40 years for more than 500 indicators. While causal relationships between these health indicators obtained
Submission Id : 94
213
using a sophisticated econometric model by Cornia et al. [4] the sorted correlation matrix (right diagram of Figure 2)
This integration allows us to get a much better and deeper evokes interest by social scientists, and the resulting mapping
understanding of the similarities, anomalies, and evolution of of one of the columns of the correlation matrix on to the
indicators and countries. geographic map (Figure 3) is of great interest to all. We
have also developed a user interface that allows easy selection
II. R ELATED W ORK of indicators and countries from a variety of databases and
There has been considerable advances in visualizing geo- visualizations to create customized visualizations (including
graphic information data using a variety of novel techniques zooming and data mining features that allow users to gain
[13]. A majority of these techniques include using a combi- access to detailed underlying raw or computed data) that may
nation of texture and color to create a palette that can be used be helpful in analyzing the data at hand.
to display multivariate data [14], [15], [16]. Due to challenges Our main focus is to investigate whether the integrated
associated with understanding animated data, spatiotemporal statistical-visualization system can provide any new socio-
geographic data has been visualized using wedges, circles, and economic knowledge or insights. We applied our system
rings [23] and mashups [27]. Distortions of geographic areas to investigate deeper questions regarding health variables.
using rectangles, cartograms and a combination of cartograms In Section V, we present three examples of the results of
with pixelmaps [19] have also been used to convey the values our investigation. Due to simple and familiar visualizations,
of socio-economic indicators. Additional efforts to visualize social and computer scientists could share and understand the
geographic data include geographically weighted scale varying results equally well to create a meaningful dialogue. Many
visualization [6], diffusion-based density equalizing maps [10], of these investigations validated the understanding obtained
and two-tone pseudo-coloring to visualize one-dimensional through simple means, but the system produced some new
data [22]. [12] presents interactive feature section for identify- and surprising results and is also helpful in quantifying the
ing subspaces together with interactive hierachical clustering intuitive understanding.
to assist visualization.
III. V ISUALIZATION
While many of these techniques appear promising and are
very impressive from a visualization standpoint, most social Global socio-economic indicators can be captured in a 3D
scientists and users are unfamiliar with these techniques and volume as illustrated in Figure 1. Although one can attempt
remain wary of depending on these techniques to gain a better to view all the data in 3D, social scientists are much more
understanding of data. accustomed to familiar 2D visualizations. In this work, we
Integrating a statistical model with visualization has been first describe the typical visualizations associated with the 1D
also explored in the literature. Carr et. al. presented a way to and 2D of this 3D volume.
integrate statistical summaries with visualization by the use
of linked micromap and conditioned choropleth maps for spa-
tially indexed data[3]. The concept of using glyphs to visualize
a correlation matrix has been explored in [8] . Andrienko
et. al. use an iterative interactive approach to classify and
identify patterns in spatial data, by using visualization and data
mining [1]. Guo et. al. have presented an approach to cluster
and sort large multivariate datasets based on self-organizing
maps [11]. While these are general visualization toolkits, our
application is more tailored towards the needs of our target
audience, social scientists, and specifically intended to study
country/indicator based patterns relative to each other.
In the integrated geographic statistical-visualization system
that we have built, we are investigating relationships between
causality, simple statistical relationships between indicators
and countries, and intuitive understanding as obtained through
simple visualization. We have chosen to use the causality Fig. 1. The 3-dimensional volume of indicator (I), country (C)and time (T)
model for global health indicators recently proposed by Cornia data, with (i) a vertical 2D slice highlighted which shows times series data
for all indicators for a specific country, (ii) a horizontal 2D slice showing
et al. [4], that we describe in further detail in Section IV-C. time series data for a single indicator over all countries, and (iii) a vertical
In order to contrast our results with those obtained by Cornia, 2D slice for all indicators for all countries at a specific time.
we have integrated our system to draw data from the GHND
database [21] that has been used by Cornia et al. in their The three 1D slices of the 3D volume of global socio-
study. We have also integrated other databases including the economic data are C-slice (one indicator, one time, all coun-
World Bank indicators. We have introduced a visualization of tries), T-slice (one indicator, once country, all time), and I-slice
correlation and regression matrix (left diagram of Figure 2), (one country, one time, all indicators). The C-slice is typically
that has been used mostly by computer scientists; However, visualized by mapping the indicator values on a geographic
2
Submission Id : 94
214
map using pseudo-color and is perhaps one of the most popular countries to support or contrast the findings based on statistical
geographic visualizations. The T-slice is commonly visualized or regression analysis.
as a time-series line graph. The I-slice is usually presented as In addition, we have also supported multiple overlaid scatter
a table. plots (see Figure 9) where users are allowed to pick individual
In addition to these 1D slice visualizations, visualization or some group of countries.
of 1 12 D data, that is 2 or more layers of these 1D slices, 3) IC Slice for a Time Period: The Indicator-Time slice
is very common. We will refer to two layers of C-slice as represents all the indicators across all the countries for a given
a 2C-slice. A 2C-slice may represent two indicators, one year (Figure 1 (iii) ). We are not aware of any effective way
time, and all countries or one indicator, two times, and all of visualizing the whole 2D slice of this data. Again, it will
countries. In the first case, the data is ideal for visualization on be useful to reduce the dimensionality of indicator space.
a geographic map using bivariate display techniques, although As we will see soon, statistical tools allow us to quantify
there is no one commonly accepted technique except perhaps the relationship between two rows or columns of the given
for side-by-side display of two geographic maps. In the second volume of data and visualize them providing us with better
case, although animation is commonly accepted, technique understanding of relationships between the indicators or the
of small multiples is often employed in practice where two countries.
static images are displayed side-by-side. Gapminder [9] has
IV. S TATISTICAL T OOLS
developed a technique of visualizing a 4C-slice of 4 indicators
using a scatter plot where 2 variables are mapped on the x-axis A. Correlation
and y-axis and two additional variables are depicted through In this work, we utilize correlation in at least two ways
glyph size and glyph colors. – to compute correlation coefficient between two indicator
A 2T-slice may represent two indicators, all times, for one trends for a given country (shown as A in Figure 1) and
country or one indicator, all times, for two countries. This 2T to compute correlation coefficient between two countries for
slice is typically visualized using line graphs or bar graphs. a given indicator trend (shown as B in Figure 1). The first
An NT-slice can also be visualized in similar ways within the approach is useful if we want to analyze how two given socio-
space constraints. economic indicators have varied over time with relation to
A 2I-slice may represent one time, two countries and all each other. We compute their correlation for each country,
indicators or two times, one country and all indicators. The which gives us a single correlation value per country which
purpose of these 2 slices are very different. In the first case, is visualized on a geographic map. This analysis can be used
the goal is to compare the two countries, while in the second to answer questions such as ’Is an increase in immunization
case, the goal is to examine all the indicator trends for the same always correlated with a decrease in infant mortality?’
country. This data is typically presented in a tabular format or In the second analysis, we compute the correlation trends
if a subset of indicators is chosen, then this subset can be (over a period of time) between countries for a single indicator.
visualized using classical visualization techniques including Since each country pair would have a correlation value, we
line or bar graphs. visualize these results using a correlation matrix (see Figure 2).
A cellrow,column in the matrix represents the correlation between
A. Slice Visualization
the indicator trend of a countrycolumn and the indicator trend
1) TC Slice for an Indicator: The Time-Country slice (Fig- of a countryrow . A single column of this matrix corresponds
ure 1 (ii)) represents the data for a single indicator spanning to correlation of the indicator trends of a specific country with
all countries over a period of time. Most websites visualize the indicator trend of all the other countries. This analysis can
this slice using an animation of a world map where countries be used to cluster countries based on indicator trends. We can
are pseudo-colored based on indicator values for that point in get a better understanding of how close those countries match
time. up by sorting and visualizing the correlation matrix by that
2) TI Slice for a Country: This slice (Figure 1 (i)) is useful particular country (right diagram of Figure 2). This analysis
in understanding the evolution of socio-economic trends within can also be used to determine whether a particular indicator
a country. The full 2D slice includes all indicators and is trend, for example, the increase in life expectancy, has been
difficult to visualize. A useful task in this case would be to uniform in all parts of the world, and allows easy identification
reduce the dimensionality of socio-economic indicator space of anomalies (see Figure 3).
by identifying a subset of key indicators for a chosen country. We use Pearsons product-movement correlation coefficient
This would mean restrict oneself to a few rows of the TI [20] as our correlation estimator. The correlation coefficient
slice, typically by choosing specific indicators. These indicator gives a measure of positive and negative linear correlation,
trends are then visualized with line graphs and bar graphs. ranging from +1 to -1.
One can also investigate relationships between a pair of
indicators for a specific country by employing a scatter plot B. Regression
by graphing one along the x-axis and the other one along the In addition to computing the correlation between two indi-
y-axis at different time periods. In this work, we have used cator trends for a country, we have also computed the linear
this type of scatter plots (Figures 6, 7, and 8) for individual regression fit for these indicator trends by taking one of the
3
Submission Id : 94
215
Fig. 2. (left)Correlation matrix for LEB (Life expectancy at birth) for years 1980-2005 between 207 countries; (right) The
same correlation matrix sorted by the column for Sweden, indicating correlation of other countries with Sweden for LEB from
1980-2005.
Fig. 3. Correlation of other countries with Sweden for LEB from 1980-2005 shown on a world
map.
indicators to be the independent and the other the response consumption, smoking rates, unbalanced diet, migration rate,
variable. DPT immunization rate, wars, disasters, etc. Impact of these
yi = β0 + β1 xi + εi , independent variables are studied on a cluster of health vari-
ables including u5MR (infant mortality under 5), IMR (infant
where yi is the dependent or response variable, xi is the mortality rate), and LEB (life expectancy at birth).
independent variable and εi is the residual. One would expect
To improve the goodness of fit, improve the robustness
highly correlated indicators to lead to a good linear regression
of the estimates, and avoid multi-collinearity problems, some
fit and the regression coefficients β1 and β0 (which is also
variables were dropped, normalized or modified. One such
referred to as the intercept and the slope in the case of linear
variable is log (physicians/1000 people) which was divided
regression) can be used to understand the relationship between
by log (GDP per capita) to obtain an index of availability of
the two indicators. Together with the correlation visualization
distribution of health personnel relative to the GDP/c norm.
for the two selected indicators, we also provide a visualization
of the regression coefficients on a geographic map. The estimation was carried out for all the countries together,
and also for four different groupings of countries – high
C. Causality income, middle income, low income, and transitional countries
Recently, Cornia et al. [4] proposed a causality model and for two different time periods, 1960-2005 and 1980-
for global health indicators investigating five different impact 2005. Obtained results were examined for their statistical
pathways for health. These pathways are material deprivation, significance better than 1%, between 1 to 5%, between 10
progress in health technology, acute psychological stress, to 15%, and not significant.
unhealthy lifestyle pathways, and socio-economic hierarchy- Results most relevant to our work include statistically sig-
disintegration. Each of these pathways are measured by a nificant dependence of u5MR on DPT immunization rate for
cluster of socio-economic indicators that include income, all the countries as well as for all the four subgroupings of the
income inequality, unemployment rate, inflation rate, illiter- countries mentioned above and the dependence of LEB on Log
acy rate, health expenditure, number of physicians, alcohol (Physicians/1000 people)/ Log(GDP/c). In this work, we chose
4
Submission Id : 94
216
Fig. 4. Correlation coefficients between U5MR (under 5 Fig. 5. Under 5 mortality rate at zero intercept, after linear
mortality rate) and DPT immunization rate for years 1960-2005. regression between U5MR and DPT immunization rate shown
This world map depicts that U5MR is negatively correlated with only for high and low income countries from 1960 - 2005. This
DPT imm. for most countries as expected. Anamolous countries, map brings out relatively high u5MR for low income countries
such as Germany, Kazakhastan, and Congo are easily detected at comparable level of DPT immunization.
in this visualization
Fig. 6. Scatter plot between U5MR vs DPT Fig. 7. Similar scatter plot for Germany. Fig. 8. Similar scatter plot for Kazakhstan.
for Congo, from 1960 to 2000. Deviation Deviation from the norm is due to variation Deviation from the norm is due to political
from the norm is due to war. in health polivy during 1980-2005. changes.
Fig. 9. Scatter plot of u5MR vs. DPT showing 9 countries, 3 from each income group. These scatter plots reaffirm the general clustering og
high, middle, and low income countries into three separate clusters, characterized by low, middle, and high u5MR at comparable DPT levels.
5
Submission Id : 94
217
D. Clustering
Classification and clustering of countries and indicators
based on similarity is a common and useful endeavor. Existing
solutions deal mainly with classifying the countries based
at a fixed point in time, an example would be the world
bank classification of countries in to ’high’,’middle’ and ’low’
Fig. 10. Correlation coefficients between LEB (Life expectancy income groups. In our system, we can provide results of
at birth) and log(Physicians per 1000 people)/log(GDP per clustering, using K-means algorithm, on any socio-economic
capita) for years 1960-2005. This correlation is positive for indicator such as life expectancy and immunization rates.
most countries including high income countries as expected.
Also we allow clustering over a period of time, based on
the correlation results of indicator trends. For example, if
we could cluster countries that had an life expectancy trend
similar to Sweden for the period 1980-2000. This allows
grouping of countries with similar characteristics over a period
of time, as opposed to just a single year. In most of our
examples discussed later, we have clustered countries into four
categories, although our system allows choosing the number
of clusters.
Anomalies in the relationship between u5MR and DPT imm (Physicians per 1000 people)/ Log (GDP per Capita) over the
is also brought out by Figure 4. These anomalies appear as period 1960-2005.
negative or close to zero correlation for some countries. These We first discuss the derivations of the causality model
countries include Congo, Germany, and Kazakhstan. Scatter regarding the relationship between these variables. Cornia et
plots of relationships between u5MR and DPT for these 3 al. [4] derive that the regression coefficient between these
countries are shown in Figures 6, 7 and 8 respectively. Reversal two variables for middle, low, and transitional (Eastern block
or decrease in DPT immunization rate in Congo from 1990 countries) are 11.2796, 14.2350, and 8.6528, being significant
to 2005 is a result of war. Reversal of decrease in DPT at 1% level for middle income countries and being significant
imm rate between 1990 to 2000 in Germany is due to a between 1% to 5% level for low income and transitional
variation in health policy that has been checked since 2000 countries. The relationship between these variables is also
resulting in continuance of the desirable trend. Finally, the significant at 1% level for all the countries together with
increase in u5MR in Kazakhastan from 1990 to 2005 is due even higher regression coefficient of 36.89. Surprisingly, the
to political changes in the country. In summary, the correlation regression coefficient between these two variables is negative, -
visualization on the world map quickly leads us to anomalies; 28.9, also significant at 1% level. These regression coefficients
supporting scatter plots quickly helps us in validating the are visualized in Figure 13, where the negative regression
anomalies and leads us to causes of these anomalies and points coefficient is mapped to the red color, while the other three
towards possible challenges or recommendations for changes coefficients are mapped to yellow, light green, and dark green
in health policy. in increasing magnitude of the regression coefficient. These
We now examine the relationship between the same vari- causality results are in contrast with the correlation coefficients
ables, u5MR and DPT, using linear regression between the two visualized for all the countries in Figure 10. The dark green
variables. After a linear fit, we compute the y-intercept, that is, colors in Figure 10 illustrate that the relationship between the
level of u5MR at a hypothetical zero DPT level. These levels two variables are positive, as expected, that is increasing the
of u5MR are then visualized only for the high and low income number of physicians (compared to GDP per capita) ’results’
countries (excluding the middle income countries) in Figure in an increase in LEB. While the causality model in Figure
5. This visualization brings out the sharp contrast between the 13 points to a hypothesis that in a multi-regression model,
two groups of countries. the overall increase in LEB attributed to other factors such as
This observation is further validated by picking 3 sample Log GDP/volatility, female education, alcohol consumption,
countries from each of the three groups – low, middle, and and smoking, etc. is in fact offset by physicians to bring the
high income – and then visualizing the relationship between model in line with the rest of the countries. This example illus-
u5MR and DPT on a scatter plot in Figure 9. This supporting trated the utility of statistical visualization in bringing deeper
visualization using raw numbers further validates the obser- understanding and checks against the more sophisticated but
vation that the low income countries are typically clustered harder to understand multi-regression causality models.
towards the high range of u5MR and also saturate at higher Correlation and regression computations and visualizations
levels of u5MR than the middle or high income countries. bring further insight into the relationship between the two
This observation leads to the conclusion that DPT can help variables. Figures 11 and 12 show the countries with negative
reduce u5MR only up to a certain point in low and middle correlation coefficients and those with negative regression
income countries and additional health measures need to be coefficients respectively. In this case, we observe that cor-
undertaken to reduce u5MR further. Although this observation relation and linear regression results agree with each other
may seem obvious after these visualizations, the causality strongly. Furthermore, the negative relationships between the
model described by Cornia et al. [4] focus mostly on the two variables are present predominantly for transitional (East-
regression slope and not making any of the observations listed ern block countries). This is, again, a surprising result, since
above since their multi-variable regression model does not the causality model by Cornia et al. [4] computed a positive
accommodate the simple intercept view of linear regression. regression coefficient for these countries with high statistical
Nevertheless, it is to be noted that most users, when browsing significance. These visualization based observations lead us
raw data using popular websites such as Gapminder and UC to believe that the relationship between these two variables
Atlas are intuitively looking for simple relationships between is more complicated than a simple causal one and requires
variables and the closest statistical analogs are typically cor- further investigation.
relation and regression analysis. In the examples discussed
C. Clustering
so far, simple visualizations including scatter plot, correlation,
and regression visualization go a long way to provide valuable We now present our third and final example, using clus-
information regarding the relationship between these variables. tering, to illustrate the utility of integrating statistical tools
with visualization. To this purpose, we classified countries
into four categories using many different indicators. Figure
B. Correlation, Regression, and Causality
14 presents the visualization of countries classified into low
We now present a second example of relationship between (red), lower middle (yellow), higher middle (light green), and
LEB (life expectancy at birth per 1000 children) and Log high income (dark green) countries based on GDP in the
7
Submission Id : 94
219
u5MR(T)
GDP(T)
as compared the trend of Sweden. Results of comparison of
LEB(T)
DPT(T)
Phy(T)
u5MR
GDP
LEB
DPT
BRIC countries are presented in a table.
Phy
Brazil 2 1 1 1 2 1 1 4 1 1 The table along with the visualization show that India is
Russia 2 3 1 2 2 4 1 4 1 1 slightly behind other BRIC countries in u5MR while Russia
India 1 1 2 1 3 1 3 3 2 1 is slightly behind in the u5MR trend. In LEB, again, India is
China 1 1 1 1 2 1 1 3 1 1
slightly behind other BRIC countries while Russia is slightly
TABLE I behind in the LEB trend. Put together, in the two health
T HE RESULTS OF USING KM EANS TO CLUSTER COUNTRIES FOR A
SPECIFIC YEAR (2000) AND ALSO BASED ON TRENDS FOR A SPAN OF
indicator trends, u5MR and LEB, India lags behind other
YEARS (1980-2005). C LUSTERS 1 THROUGH 4 REPRESENTS THE ’ BEST ’ BRIC countries, but can catch up if it maintains its trend,
TO ’ WORST ’ CLASSIFICATIONS RESPECTIVELY. while Russia is at the greatest risk of falling behind in the
health indicators.
In terms of action or independent variables that impact
health, DPT immunization rates for all BRIC countries are in
the lowest two categories, Brazil and Russia being the lowest,
and India and China next to the lowest. However, with respect
to the DPT trend, Brazil, Russia, and China are in the strongest
category (green), while India is somewhat lagging behind. In
terms of the presence of physicians trends, all four BRIC
countries are in the strongest category, while India slightly
behind in the year 2000. Put together, all BRIC countries are
likely to improve their health indicators by increasing the DPT
Fig. 14. Clustering of countries based on GDP in year immunization rate and need to maintain their strong growth
2000 trend regarding physicians.
VI. C ONCLUSIONS AND F UTURE W ORK
In this work, we proposed an integration of statistical
computing with visualization to glean deeper understanding
of global socio-economic indicators. We utilized correlation
and linear regression to quantify relationships between pairs
of variables and between pairs of countries. We utilized these
tools to investigate static data for a fixed time period as
well as dynamic trends over a large time period. Current
Fig. 15. Clustering based on GDP trends between state-of-the-art global inequality websites provide visualization
Sweden and other countries for years 1980-2005. support using raw data without the use of any statistical
tools. Using three different examples, we demonstrate that
correlation, linear regression, and causality models can bring
year 2000. We will follow the results for the emerging BRIC out similarities and anomalies and provide better understand-
(Brazil, Russia, India, and China) countries. In this case, India ing of relationships between the variables by validating our
and China are still low income countries while Brazil and intuitions based purely on raw data visualization or sometimes
Russia are lower middle income countries. However, when yields insights that are counter-intuitive or surprising. These
we view the classification of GDP trends for the four BRIC observations or conclusions carry important implications in
countries over the period 1980-2005 in Figure 15, we observe policy making both at national and global level.
that India, China, and Brazil are classified into the same This research has opened up several new exciting oppor-
(and strongest) category as most of the developed nations tunities. Which countries can be grouped together? Based
including USA, and most European countries, while Russian on which indicators? Which socio-economic indicators can
GDP growth is the next lower category. For the purposes of be clustered together? Can we reduce the dimensionality
the trend classification, we computed the correlation between of indicators so that a profile of a country is captured by
GDP trends between all the countries and Sweden. We chose some principal socio-economic indicators? What lessons can
Sweden because it was consistently in the top for most of a nation learn from a similar group of nations? Ideally, we
the indicators that are we investigating in this study, including would like to build a system so that the empowered users can
GDP, LEB, etc. explore relationships between countries and between variables
Figures 20 to 23 classify all the countries on the four using appropriate statistical tools combined with visualization.
variables – u5MR, DPT imm, LEB, and Log (Physicians per We believe that this exploration can always be used to validate
1000 people)/ Log (GDP per Capita) using the static data from or contrast the proposed policy decisions and may also lead to
the year 2000. Figures 16 to 19 classify the countries based on important underpinnings of national or global policy decisions
the trends of these four variables over the period 1980-2005 that are not immediately obvious.
8
Submission Id : 94
220
Fig. 16. Clustering based on U5MR trends between reference Fig. 17. Clustering based on DPT immunization trends between
country (Sweden) and other countries for years 1980-2005. reference country (Sweden) and other countries for years 1980-
2005.
9
Submission Id : 94
221
10
222
Fig. 1. Population centers in northwestern Uganda Fig. 2. GSM network coverage in northwestern Uganda
(darker areas indicate better coverage)
primary factor in the proliferation of self-contained cellular
telephone towers which uplink via satellite or fixed underlying difficulties, not least of which is cost. A naïve
wireless transmission. Other than the existing wired model, approach to the problem might be to send a series of
there have been efforts to connect ATMs to financial messages back and forth between the ATM and the
institutions using long-distance wireless links[6] which financial institution – one to request a withdrawal, one to
also provide telephony and internet services, but the case issue an identity authorization challenge, another to supply
for capital investment of this sort for areas in which cellular the response, yet another to issue an approval, one to
telephone service already exists is more difficult to make as acknowledge the approval, etc. etc. While the absolute cost
some degree of connectivity is already present. Any of these messages is not extraordinarily high, the cost does
solution which proposes to address this problem must take not scale well as the user base increases. If the institution
into account lack of network capacity as well as keeping pays for the messaging, the cost eats into any financial gain
cost of deployment and operation low enough to be the client may realize from interest on his savings if he is
practicable for national banks in the region to afford. expected to pay, and in either case this approach does not
We use the wired ATM metaphor rather than simpler scale well as loss rates may be expected to increase if
kiosks which read smartcards with encoded balances or message traffic increases substantially while network
paper passbooks because these methods are vulnerable to capacity remains fixed.
counterfeiting, theft, loss, and damage and have poor
3 DESIGN
provisions for centralized auditing.
Our initial objective is merely to provide some form of
ATM service to rural customers with small amounts of
2.2 The Scope of our Approach personal savings. Our simulated testbed contains 46
The approach presented in this paper is aimed squarely population centers of sizes varying between 2001 and
at reducing cost and making maximum use of existing 70872, totaling 637952 in total. We simplify this by
capacity in the form of the installed base of GSM network considering households as accountholders, using the
cells. As the network’s reliability, latency and bandwidth average household size across Uganda of 4.7, yielding
are not subject to any guarantees, we create a protocol 136541 households if we consider each center separately
which is designed to be resilient against extended outage and round down.
conditions and slow message delivery while simultaneously Having ruled out wired infrastructure, our next
maintaining an extremely low message budget and high observation is that long-distance[10] wireless is feasible as
quality of service. a physical network layer, but has the disadvantage of
requiring potentially extensive infrastructure deployment
2.3 The Technical Challenge depending on the geography of the region, distances
between nodes, and the network topology necessary to
The use of the installed GSM base is a major benefit to
scale with growth of the size of the network. Newer
the design of a solution to this problem, but there are
224
technologies such as WiMAX[30] may allow connectivity techniques currently used in other devices, such as credit
between adjacent villages with commodity hardware, but card swipe readers. A possible solution to this problem is to
how much effective bandwidth a rural WiMAX mesh use a similar model to the M-PESA system and locate
installation will have is still an open question, and there is ATM kiosks within vendors’ shops, allowing the vendors
still the question of deployment lead time and network to charge a small fee for their use.
maintenance costs. GSM cellular technology is widespread We assume that any losses incurred through physical
in the developing world, and reaches into some of the most attacks on the machine are covered by the financial
sparsely populated regions. While the General Packet institution, as with machines in urban locations. In the
Radio Service (GPRS) is part of GSM Release 97 and same vein we neglect considerations about powering the
newer (colloquially referred to as 2.5G and 3G), it is not a machine (solar is an option for low-powered embedded
viable option for data transport as much of the installed computers), how deposits will be marked, validated, and
base in rural areas is comprised of cheaper 2G cellular conveyed to the host institution, and the technical details of
hardware. Due to these factors, we choose to use SMS. the tokens used to identify the users of the machine other
Once connectivity is established, a natural design than that these tokens contain some cryptographically
principle for our system is that we want to reduce network secure writable medium, one such device being a smart
traffic as much as possible, as this represents a marginal card (ICC).
cost to our system, and cost is precisely what we wish to Finally, we assume that we may neglect certain
reduce. In opposition to this is a desire to maintain overall transaction metadata, such as the exact timestamp on each
responsiveness of the system as perceived by the customer. transaction, or that several transactions occurred within a
This perception is directly related to satisfaction rate of given timeframe, preferring rather to reduce the
withdrawal requests, and tied only weakly to the communication budget required to send information about
immediacy of funds availability from deposits, as existing a given transaction by reducing timestamp granularity to an
microdeposit schemes often take up to a month[13] to hour or day-level timescale based on the aggregation
register updated balances. A final critical design principle window of transactions and potentially aggregating
is that, given the intermittent network communication multiple transactions which occurred close together in time
enforced by our traffic reduction, we need to intelligently into one larger transaction (e.g. five transactions that
distribute balances in such a way as to prevent any user happened at the same ATM on the same account within 15
from being able to game the system by visiting multiple minutes of each other may be represented as one
ATMs within a given update window and withdraw more transaction which is the sum of those individual
than his account balance. transactions, at an approximate time.)
ATM computes the same hash upon receipt, and an echo In our testbed, the geography is advantageous as it offers
containing the ATM’s hash value is sent back. Any ATM only a few inter-town river routes, and no tarmac-paved
which returns an inconsistent value is flagged as roads. We also use the intuition that (1) accounts with
‘disconnected’ as above. higher available balances are more attractive to
Messages which are received but fail checksum and overwithdraw, (2) people with higher balances are similarly
cannot be repaired are discarded and considered lost. more likely to have access to faster transport such as boda-
bodas, (local motorcycle taxis), matatus (intra- and inter-
3.4 Security
town minibuses), and interdistrict buses or to own or be
SMS over GSM is susceptible to a variety of attacks[7]. able to borrow a private vehicle, (3) that fraud associated
In particular, it was indicated in a personal communication with higher-balance accounts costs the host institution
to the author by a member of MTN Uganda staff that SMS more, and (4) holders of high-balance accounts are less
is transmitted in plaintext throughout its network, and this likely to need to withdraw all or most of their available
is the default behavior[7]. As this is the case, and as both balances at once.
Originating Address (OA) spoofing and realtime sniffing of We consider the maximum feasible distance covered by
SMS messages are available in the wild, it is necessary for walking over a 3 hour period (~10km) and the maximum
the system to have mechanisms to protect the secrecy of feasible distance covered by driving along unpaved roads
balance/account information and to provide some measure during that same period (~125km). Given that the average
of authentication of messages sent between the ATMs and daily unskilled rural labor wage is in the neighborhood of
the CA. $1.2 per day[13] (approximately Ushs2100), we arbitrarily
In order to maintain effective payload size while set the ‘high balance’ mark at Ushs900000 (USD ~5000).
providing this type of protection, a stream cipher is used, Accounts with balances of 450000 and under have a
which produces ciphertext of the same length as the maximum of 100% of their account balance distributed
plaintext. This system uses RC4 due to its ease of across all the ATMs in the system, decreasing linearly to
implementation, but any such encryption system will serve. 75% at 900000 and beyond.
Each ATM maintains a separate 256-bit key, and the CA Only ATMs within a specific distance of the primary
maintains a keyring of all of these keys, using them as ATM are considered for this distribution, specified by the
appropriate in communication to a given ATM. formula
A 32-bit timestamp is embedded into all messages in
order to mitigate the problem of replay attacks; a timestamp (dmax/2 – d0) * min(1,b/c) + d0
will only be accepted once and any subsequent message
from a given source bearing the same timestamp will be where dmax represents the greatest distance between any
discarded. In order to mitigate substitution attacks, this two ATMs in the network, d0 represents the minimum
timestamp is placed between arbitrary (version, account, radius to be considered (40km in our case), b represents the
amount) tuples and will be padded with 12 bits of 1’s in balance of the account in question, and c is the cutoff ‘high
order both to maintain tuple offsets and to indicate which balance’ mark at which maximum distribution spread is
block contains the timestamp. Version number 15 then reached. This encapsulates our intuition that we want to
becomes reserved as not to restrict the range of account distribute the balance for those with higher balances over a
numbers which may be used. larger area, and the linear increase in radius produces an
increase in area covered proportional to the square of this
3.5 Balance Distribution value.
The proportion of the available balance held in the
The practice of available balance distribution among
primary ATM varies according to the formula
ATMs is a hedge against users who may wish to game the
system by trying to take advantage of network outages or
bprimary = b * pmax – ((pmax – pmin) * min(1,b/c))
periods of higher-than-normal latency in order to withdraw
more than their account balance. We assume for the
where pmax and pmin are the maximum and minimum
purposes of this practice that ATM use multifactor
proportions, respectively (.8 to .4 in our simulation). These
authentication where at least one factor is a biometric, such
numbers are based on assumptions we make about the
as a fingerprint. The hash of this biometric is stored on the
behavior of clients as listed below, and the formula again
card in order to defend against card duplication.
reflects our intuition that the further the client is likely to
We begin by associating each given user with a
travel, the further afield the balances should be distributed.
‘primary’ ATM, which is the machine closest to that user’s
The balances to the ATMs within the effective radius
residence, and then consider the other ATMs which are
defined above other than the clients’ primary ATMs are
present within one 3 hour period’s travel of that location.
distributed proportionately to the inverse of the square of
228
the distance of the machines from the primary ATM. Any proportion that is allocated is drawn from the bottom up:
available balances less than Ushs250 and any residues of any available balance outside the primary ATM is
allocated balances modulo Ushs50 are removed and added consumed first, and then primary ATM balance is
to the primary ATM’s available balance. reallocated until the correct proportion is reached.
These two formulas together have the effect of
concentrating more of a user’s balance nearer to his home 4 SIMULATION
ATM as his balance decreases. In practice the constants
and proportions are adjustable to ensure that any balances 4.1 Simplifying Assumptions
under a given amount are held entirely at the user’s primary There are several factors present in the real-world region
ATM. upon which our simulated testbed is modeled which we
ignore completely. One is familial seasonal migration from
3.5.1 Migration place to place – for the one-year duration of our simulation
The initial germ idea for this project called for support we assume that people’s residences remain fixed, i.e. the
for highly migratory populations to have continuous access primary ATM of a given account never changes; individual
to funds. We solve this problem by introducing a notion of members may go on excursions but never the entire family
carrying ‘virtual’ available balance on a credential token at once. Another is that of immigration to the region or
used to access the system, such as a smart card. This other sources of population growth such as refugee influx
or organic growth of the number of households through
‘virtual’ balance would involve ‘locking’ a specific amount
marriage. We assume that all members of our simulated
of available balance on one ATM and writing a device-
environment start life with accounts, that is, we have a
signed token for the amount onto the card including a
100% adoption rate, with some proportion of accounts
unique transaction ID allowing the user to take the card to beginning with a zero balance. As a final detail, we have
some specific other ATM and withdraw the amount found that we do not have enough information on the
‘locked’ plus any available balance on the local ATM. This overall prevalence of bitwise errors within SMS delivery,
could be useful for trips where larger amounts of capital as the literature has various accounts, and have no
are required but carrying cash is dangerous, e.g. seasonal information at all on its prevalence in the rural context,
trips to buy seed for planting or taking animals or produce where distance may increase and technology may be older
to market. Using the extra balance would cause the local or cheaper but there are far fewer sources of
machine to create a new signed token with any remaining electromagnetic noise, so we ignore it, assuming that those
‘locked’ balance and inserting it into the original machine messages not correctible by the checksum are incorporated
would unlock any such remaining portion and treat the into the loss rate.
spent portion as a withdrawal. This has the advantage that
if the card is lost or stolen, funds which are locked on the
card but have not been unlocked anywhere across the ATM
network can be recovered from the institution.
3.5.2 Multihoming
We wish to model one commonplace situation, in which
a head of household or other member may work in a
remote region for most of the year and return for several
months or a season. In such cases, the person who is
working away from home typically wants to send money
back to his household – this is the problem that approaches
like M-PESA address. We model this situation by
explicitly adding the single ATM closest to the worker to
the set of ATMs considered for balance distribution, if it is
not already in the set. While the typical usage of such an
ATM will be solely to deposit funds left over after cost of
living is paid for in cash, we arbitrarily allocate 20% of Fig. 4. Proportion of downtime versus standard deviation outage burst
balance available at the primary ATM to this ATM. The length for 4 values of λ.
rationale is the idea that the breadwinner represents, on
average, 1/5 of the total size of the household based on the
average family size of 4.7 and assuming a single primary 4.2 Simulator Overview
source of income. This is trivially extended to The simulator uses minute-scale timeslices over the
accommodate arbitrary family configurations. The course of a calendar year and contains metaphors for
229
customers, ATMs, the CA server, and the SMS network specific characteristics with which we attempt to
connecting the various components of the overall system. reasonably emulate a real population:
In our experiments we change various parameters of the 1) Primary ATM, allocated randomly using a uniform
system to see what effect they have on the cost model and distribution
customer request satisfaction rate. We consider a 2) Weekly savings rate, generated by taking an arbitrary
withdrawal request successful if, having specified some minimum savings rate (Ushs200) and adding it to
amount desired which is less than his account balance, the Ushs1000 multiplied by the absolute value of a random
machine both has enough cash on hand to satisfy the element drawn from a Gaussian distribution with mean 0
request and has enough of the balance allocated to the and standard deviation
given machine to allow the machine to dispense that cash. 3) Visit frequency, a step function generated by another
random Gaussian with mean 0 and standard deviation 1, by
4.3 Network Behavior
assigning 7 days if the random value is within 1 standard
Network behavior parameters are the factors most deviation, 14 if within 2, and 28 if more than 2
directly related to the cost model. To model bursty outage 4) Withdrawal strategy (whether the client will withdraw
behavior, we use a Poisson distribution to model burst a proportion of his balance or of his weekly withdrawal
arrival and a truncated normal distribution (with mean at 0 rate, half of the clients being in each category), selected
in which we only consider positive values) to model burst using a uniform random variable
length, and change λ and σ2, respectively for each to test 5) Probability of withdrawing from an ATM other than
the robustness of the system under various outage models, his or her primary ATM, between 0% and 20%, generated
as well as how the level of customer satisfaction changes as using a Gaussian distribution with mean 10% and standard
the network and therefore the application built atop it deviation 5%. Any random value selected outside of 0-20%
becomes less reliable. We would also like to vary latency is reselected
with a truncated normal distribution and μ set to 36 6) 0-4 arbitrarily placed seasonal brief migrations where
seconds as the experimental data in 3.2.1 indicated, and the 10% of the population has one or more migrations.
σ2 likewise set to 5 to show how the effects of variable Presence of a migration is decided by a uniform random
propagation delays; however the timescale we are using variable, and the number of migrations is generated in an
renders this variability negligible. As such we set μ and σ2 analogous manner to 3)
to 1 minute, which biases slower than experimentally and 7) multihome (second primary) ATM, where 10% of
detected, which cannot artificially improve results. Finally, the population has a secondary home as described above.
we add an additional Poisson distribution-based loss model Both the presence of a second primary ATM and its
for ‘black hole’ loss where messages simply disappear after location are selected using a uniform random variable.
having been successfully sent to the SMSC. For simplicity, we register deposits proportionate to the
weekly savings rate each time a client visits an ATM for
withdrawal. Deposits are processed every two weeks, and
only deposits received at least one week prior to processing
are applied to balances.
4.5 ATM Parameters
We use a parameter n to determine how many ATMs to
use to saturate our network, i.e. one ATM per n users.
When this figure causes the number of ATMs to exceed
one per population center, we use the notion that a given
account can only use one ATM within the center and any
others are treated as nonexistent, to obviate complexity of
balance distribution among the multiple ATMs while
maintaining semantic equivalence. We also use a ratio p to
specify the proportion of the sum of the total available
balances from a given ATM that the ATM should have on
Fig. 5. CDF of customer weekly savings rate in Ushs. hand in cash, which affects cash restocking. Finally we
specify how many days the server waits between balance
4.4 Household Behavior redistributions, tb, which affects message traffic volume as
The 136541 households in our testbed are assigned to well as funds availability following a deposit.
population centers according to real-world population To limit the number of variables in the simulation, we
levels. Each account is held by a household having certain chose to fix the frequency of cash refills at once every two
weeks.
230
Fig. 6. Loss versus Cost per User in US Dollars, averaged across all runs Fig. 7. Loss versus successful transaction proportion, averaged across all
of the simulator. runs of the simulator.
Fig. 8. Cash load proportion versus successful transaction proportion, Fig. 9. Number of users per ATM (n) versus Cost per User in US Dollars
averaged across all runs of the simulator. averaged across all runs of the simulator.
Fig. 10. Days between balance redistributions versus cost Fig. 11. Days between balance redistributions versus satisfaction rate
will be investigation into whether decentralization can [9] C. Olston and J. Widom, “Offering a Precision-Performance
Tradeoff for Aggregation Queries over Replicated Data,” in 26th
reduce network traffic and increase robustness against International Conference on Very Large Data Bases (VLDB). Cairo,
network failures, particularly at the cellular network cell Egypt, 2000, pp. 144-145.
hosting the institution. As balance distribution is dependent [10] R. Patra, S, Nedevschi, S. Surana, A. Sheth, L. Subramanian, and E.
on geographical location, it may prove to be wise to have Brewer, “WiLDNet: Design and Implementation of High
Performance WiFi Based Long Distance Networks,” in 4th USENIX
nodes collude with their nearest neighbors rather than Symposium on Networked Systems Design & Implementation.
receiving updates only from a central authority. Cambridge, MA, 2007.
Other potential extensions of the project are increases in [11] L. Payne, “Rebuilding Communities in a Refugee Settlement: A
Casebook from Uganda.” Oxford: Oxfam, 1998.
granularity or changes in the strategy of balance [12] G. Peersman, P. Griffiths, H. Spear, S. Cvetkovic, and C. Smythe, “A
redistribution and tracking of loss and corruption rates of tutorial overview of the short message service within GSM.”
received messages in order to better allocate message Computing & Control Engineering Journal, 11 (2), pp. 78-89.
[13] S. Staschen, “Microsavings, Microcredit and Microinsurance:
budget. A redistribution strategy which is more adaptive to Financial Products of Small Farmer Co-operatives Ltd. In Nepal.”
the past patterns of the user is also likely to yield positive Cologne: AEF University of Cologne, 2001.
results in terms of customer satisfaction. G. Taylor and M. Bekabye, “Employment Creation and Labour-
Based Technology in Roadworks: Ugandan Case Study, 1999.”
As crafted, the system could be trivially extended to [Online] Available:
allow for payouts of microcredit installment loans or http://www4.worldbank.org/afr/ssatp/Resources/HTML/rural_transp
salaries for rural workers. ort/knowledge_base/English/Module%202%5C2_4b%20Case%20St
udy%20-%20Labour%20Based%20Techs.pdf
[14] C. Tseng, J. Jiang, R. Lee, F. Lu, C. Ouyang, Y. Chen, and C.
7 CONCLUSION Chang, “Feasibility study on application of GSM-SMS technology to
This paper presents ATMosphere, a protocol designed to field data acquisition.” Computers and Electronics in Agriculture, 53
(1), pp. 45-59.
provide ATM services to the world’s rural poor. The [15] S. Wisniwski, “Microsavings Compared to Other Sources of Funds.”
system as simulated is based on currently deployed Eschborn: GTZ, 1999.
technologies and is implementable today using low-cost [16] H. Yu, and A. Vahdat, “Design and Evaluation of a Continuous
Consistency Model for Replicated Services,” in Fourth Symposium
hardware. The result is a solution which has low start-up on Operating Systems Design and Implementation. San Diego, CA,
and marginal costs, represents low relative risk to financial 2000.
institutions, and provides sorely needed access to savings [17] H. Yu, and A. Vahdat, “Efficient Numerical Error Bounding for
Replicated Network Services,” in 26th International Conference on
and capital markets for the marginalized. Very Large Data Bases. Cairo, Egypt, 2000, pp. 144-145.
[18] M. Zeller, “Models of Rural Finance Institutions,” in Paving the Way
8 ACKNOWLEDGEMENT Forward for Rural Finance: An International Conference on Best
Practices. Washington, DC, 2003.
We would like to thank various personal friends living in [19] ETSI GSM 3.40: ‘Digital cellular telecommunication system (Phase
rural Uganda for firsthand assistance with baseline testing 2); technical realisation of the short message service point-to-point’,
of MTN’s cellular network. We would also like to thank European Telecommunications Standards Institute TC SMG, Version
4.13.0, May 1996.
Jinyang Li for her comments. [20] ETSI GSM 3.41: ‘Digital cellular telecommunication system (Phase
This material is supported by the National Science 2); technical realisation of the short message service cell broadcast
Foundation under Grant No. CNS-0831934. (SMSCB)’, European Telecommunications Standards Institute TC
SMG, Version 5.2.0, May 1996.
[21] GSM Doc 28/85: ‘Services and Facilities to be provided in the GSM
REFERENCES System’ rev2, June 1985.
[1] E. Brewer, “Technology Insights for Rural Connectivity,” in [22] MTN Uganda Ltd (MTN-UGANDA) – Interactive Coverage Map,
Wireless Communication and Development: A Global Perspective. 2007. [Online] Available: http://www.gsmworld.com/cgi-
Marina Del Rey, CA, 2005. bin/ni_map.pl?cc=ug&net=mt.
[2] L. Elser, A. Hannig, and S. Wisniwski, “Comparative Analysis of [23] MTN Uganda: A Bright Y’ello World, 2007. [Online] Available:
Savings Mobilization Strategies,” Eschborn, GTZ, 1999. http://www.mtn.co.ug.
[3] W. Enck, P. Traynor, P. McDaniel, and T. La Porta, “Exploting open [24] MTN Uganda: Tariff Guide : Mobile Solutions : Business, 2008.
functionality in SMS-capable cellular networks,” on 12th ACM [Online] Available:
Conference on Computer and Communications Security. Alexandria, http://www.mtn.co.ug/business/talktime_tariffs.htm.
VA, ACM, 2005, pp. 393-404. [25] The National Cheque Standard of Uganda, 1999. Available:
[4] A. Hannig, and S. Wisniwski, “Mobilizing Microsavings: The http://www.bou.or.ug/NPSS/Cheque%20Standard.pdf.
Experience of Seven Deposit-taking Institutions,” Eschborn: GTZ, [26] Safaricom: M-PESA, 2007. [Online] Available:
1999. http://www.safaricom.co.ke/index.php?id=745
[5] G. Ivatury and M. Pickens, “Mobile phone banking and low-income [27] Quality of Service (QoS) Indicators for Mobile Services, Unknown
customers: Evidence from South Africa,” Washington, DC, 2006 Date. Available: http://www.ncc.gov.ng/TRS/QoS_Benchmark.pdf.
[6] A. Jhunjhunwala, A. Ramachandran, and A. Bandyopadhyay, “n- [28] Thuraya Satellite Communications, 2008. [Online] Available:
Logue: The Story of a Rural Service Provider,” The Journal of http://www.thuraya.com
Community Informatics,1 (1), pp. 30-38. [29] Uganda Bureau of Statistics, 2007. [Online] Available:
[7] S. Lord, “Trouble at the Telco: When GSM Goes Bad,” Network http://www.ubos.org
Security, 2003(1), pp. 10-12. [30] WiMAX Forum – Technology, 2007. [Online] Available:
[8] C. Olston, “Approximate Replication,” Doctoral Dissertation, http://www.wimaxforum.org/technology
Stanford University, Stanford, 2003.
233
2) *box Internals
In addition to being able to accept route numbers via a
keypad and display the provided route number to the driver,
internally the *box must also be able to determine its location
(via GPS satellites) and send a time-stamped SMS to the
server. For our prototype device, we chose the Telit GM862-
GPS module. We chose to implement the *bus system with
this very compact module because it provides a low cost
solution with the additional convenience of combining GPS
Figure 6 Overview of the *bus System and GSM capabilities on one chip (see Figure 8). Because no
additional work is necessary to interface the GPS and GSM
The notion of a route number is a good fit for the components, developing on this chip is relatively simple. The
marshrutka system in Bishkek where buses have set routes Telit module communicates with the keypad via general
(although no set timings along those routes), indicated by purpose IO to obtain the driver's currently selected route
numbers displayed in bus windows. Changing routes is done number and stores the number in memory. An ATMega16
easily by typing in the new route number. We decided to micro-controller is used to coordinate the communication
require drivers to specify their route via the keypad interface between the Telit chip, keypad and the LCD. The module
rather than forcing the server to do the more difficult job of continuously obtains its GPS location information from GPS
attempting to determine it automatically based on the path of satellites without any action by the driver and reports this
the bus, although this is also a future possibility. information along with its currently selected route number to
One could imagine a simpler device including a hard-coded the server via an SMS message.
route number, although we chose the keypad and LCD
interface for our prototype to allow for ease of testing on B. *bus Server
different routes. The *box was designed to open easily to The primary purpose of the server is to accept SMS
allow for upgrades of the internal components and changing of messages coming from *boxes and bus riders, and to process
batteries, or inspection of device internals to allay fears of them accordingly. System requirements are intentionally
riders or drivers. Drivers can turn the *box off when not in minimal: a laptop or desktop computer running a Java virtual
service and safeguard the privacy of off-duty movements. machine and a MySQL database, connected to a phone
Since we would expect deployment models to build in driver capable of receiving SMS messages. Lower system
accountability, we do not consider theft to be a major concern. requirements allow the server to run using older and less
We built three prototype devices at a cost of about US$200 expensive hardware (such a system may cost around US$200
each, although we believe mass production would bring the to $300). To minimize SMS latency, it is recommended that
cost per box closer to US$50. In the future, as GPS-enabled the phone connected to the server be on the same cell network
cell phones become less expensive, the functionality of a *box as the *boxes, although this is not required. The server does
could potentially be replaced by an application running on a not need to be connected to the Internet, facilitating its
driver’s own cell phone. deployment in areas with low Internet connectivity. Our
Currently, the holder of the *box, in this case the driver or prototype server was implemented using MySMS [19], an
private company that owns the marshrutka, would absorb the application framework built on top of SMSLib [20]. MySMS
cost of the text messages sent out from the system. Our group provides a transport layer, allows SQL queries, notifications,
has considered cost, and the barrier it presents to adoption. remote code execution and rapid application development to
Increased efficiencies may or may not be significant enough be executed over SMS.
for an individual driver to adopt the system, but there are other
entities that might sponsor such a system for users, including a
newly formed transit authority, mobile companies which are
counting on increased SMS traffic from users, or the
employing marshrutka company deciding to adopt the system.
239
many users that will have well-known names. These can
potentially be associated with GPS coordinates in the server’s
database by the server operator. However, the *bus system
was designed to work in scenarios where little or no geo-
coded street information exists. The solution to this problem
is that we allow the clients to geo-code a location, on the fly,
by using the GPS equipment already installed on the *buses.
When a user wants to be able to refer to their current
location in future *bus queries, they can associate a name of
their choosing with the GPS coordinates of their location; they
do this by becoming a secondary user of the GPS on the bus.
The user can instruct the server to associate an arbitrary name
to their current location by waiting until a bus arrives at that
Figure 8 *box Internals. location and immediately sending the following SMS message
to the server:
The server is responsible for handling the following types of STORE <bus-id> AS <my-location>
requests which arrive via SMS messages. bus-id is the *bus id of the bus that has arrived at the user’s
location. In this case, then, a user’s location that she knows
1) Location Update Messages from *boxes only by street name is now given GPS coordinates thanks to
Messages sent from the *boxes to the server have the
the GPS receiver on the bus. The server takes the bus's GPS
following format:
coordinates at the time that the message was sent and
!LOC,<GPS-obtained-time>,<GPS-obtained-
associates it with the location name specified by the user.
location>,<bus-id>,<route-#>
Such geo-coded locations can be retrieved in the future based
The bus-id field is the unique identifier assigned to each
on the location name and the user’s phone number.
individual bus that is equipped with a *box. We would expect
Associating location names with phone numbers prevents
that bus drivers would display this bus id (in addition to their
naming conflicts between users, although other schemes are
route number) in a way that it is accessible to riders. Riders
also possible whereby location names can be shared with other
will need the bus-id in order to associate a name to a location
users, thus organically growing a set of tags for places of
(described later in 3)).
common interest. User tags could also be used to create maps
The server records this information in a database for use in
in regions where few maps exist. This is especially important
responding to user queries.
in Kyrgyzstan where the street names changed after
independence from the Soviet Union and a combination of old
2) Arrival Time Queries from Users
and new names are used. The *bus system could be used to
Information gathered about the current location of buses in
map streets in a way that is comprehensible to all residents
the system could be used to answer a variety of queries. The
regardless of which naming system they use.
server continuously receives and replies to SMS queries from
users. Potential bus riders simply text to a special *bus server
VI. PROTOTYPE DEPLOYMENT RESULTS
telephone number, and the server responds by sending an SMS
back to the user’s phone number. The most basic example is We have implemented prototype versions of the *box
to ask the server to predict the arrival time of the next bus hardware and the server software. In this section, we describe
travelling on a given route number at a given location: some preliminary data collected on the technical performance
NEXT <route-#> TO <location> of the system.
The server replies to the user with a SMS predicting: We deployed a *box on several buses in the University
Next <route-#> arrives to <location> District area of Seattle. The *box sent SMS messages to the
at about <estimated-arrival-time>. server containing its GPS coordinates at an interval of one
These examples demonstrate the types of queries we support message every 30 seconds. Based on this data the server
in English, but the exact format and language can be changed. composed a model of each of the routes. We were interested
The system has been designed so that it can accept queries and in examining the accuracy of our basic route prediction
push out information in alternate character sets – in this case, algorithm as well as the amount of time required for users to
in Cyrillic to support both the Russian and Kyrgyz languages. receive a response to their queries.
While some issues related to a multi-lingual society pose Although we have not yet field tested the system in
significant challenges (for example, when a user sends a query Kyrgyzstan, we have been careful to base the engineered
in a mixture of two languages), the system can be modified to solution on technology that will work in resource constrained
support linguistic variation. environments. Specifically, fieldwork over the years has
demonstrated to the research team that cell coverage and SMS
3) Geo-coding Requests latency in Bishkek is consistent with that experienced in
We expect that there will be some locations frequented by Seattle. Additionally, the relatively open architectural space of
Bishkek means that GPS readings will be easier to obtain.
240
Traffic patterns in the two cities are, of course, different, but
the route prediction algorithm deployed (see next section) can
be modified based on local traffic conditions.
A. Route Prediction Accuracy
1) Route Model
Much research has been done on route prediction algorithms
for public transit and vehicles in general (e.g. [21]). For our (a) Plot of the acquired GPS data for the whole route
prototype, we implemented a simple algorithm based on the (approximately 3 miles) (Google Earth)
concept of building a model by merging together multiple runs
of the route. While one could imagine an implementation that
requires the server to be pre-seeded with detailed sets of GPS
coordinates for each route, our method provides the advantage
of low setup costs for the *bus server, as no special action is
required to enter routes into its database. Furthermore,
prediction models can be made more accurate by merging
additional runs of the route, and more models for varying
traffic patterns can easily be created. More sophisticated
algorithms can be used in future versions, such as ones which
automatically account for varying traffic flow and are tolerant (b) Detailed view of the plot of acquired GPS data
of changing routes. (Google Earth)
We use our route prediction model to provide responses to Figure 9 Accuracy and resolution of GPS data collected at
different message send intervals: red – 10 seconds yellow – 20
user queries about when the next bus travelling on a certain seconds, blue – 30 seconds, purple -40 seconds, green – 60
route is due to arrive at a particular location. To make this seconds.
prediction, the server takes recent coordinates from current One option is to first produce one run with finer granularity,
runs of the route and the location entered by the user and for example, a location update every 10 seconds. This can
attempts to find these points in the model. The time between also be done manually to specify the exact path of the route.
the corresponding locations in the model provides an estimate The method we employ takes points from multiple runs of the
of the amount of time until the bus arrives at the user's route and combines them to produce a model with greater
location. resolution than a single run. Thus, more runs of the same route
The accuracy of these estimated arrival times depends on contribute to overall accuracy of the prediction.
two things: first, the fidelity of the model to the path of the In order to provide accurate estimates of arrival times, a
route and secondly, the timing similarity between the route route model must also reflect the timing characteristics of
model and the buses whose status is being queried. The model buses whose status is currently being queried. The extent to
fidelity is largely determined by the number of points in the which this is true depends on multiple factors. On a bus
model. A point consists of a set of GPS coordinates and the system such as the one in Seattle, where there are set stops,
GPS satellite-provided time of day that the data point was and drivers are expected to arrive at stops at given times,
measured. Accuracy of data points should be quite good (GPS variance can be caused by factors such as weather or
accuracy is within 8-10 meters). Thus, data points collected unusually high traffic. In Bishkek, although marshrutkas
for earlier runs on a given route should provide an accurate follow set routes, there is no schedule of any sort. Thus, we
picture of where those buses were at the time provided. would expect that there might be a wider variance in timings
Number of data points collected per route is determined by for buses in Bishkek. Collecting timing data from more bus
the frequency that *boxes are configured to send location runs and using an algorithm that incorporates time of day and
update messages to the server. More frequent location update day of week would help to improve accuracy in this case.
messages incur more cost and put more load on the server, so 2) Prediction Accuracy Results
it is desirable to keep this frequency low, yet high enough to We deployed *boxes on portions of two Metro bus routes in
give accurate arrival estimates. Seattle and collected data for four runs on each of these routes
Figure 9 below shows a plot of *box location data showing for a total of eight runs. The distance traveled was 2 miles on
the level of accuracy provided by different update message one route and 2.8 miles on the second. Each run consisted of
frequencies. As the figure shows, in general, sending update an average of 11 minutes of collected data. While this running
messages more frequently is more likely to create a model that time is shorter than the length of routes we would expect to
follows the actual path of the route. Particularly around use the system with, it allowed us to collect enough data to
curves and turns, less frequent updates can make it seem as test and provide a proof of concept.
though the bus has cut a corner. To examine the accuracy of our preliminary route prediction
algorithm, we did the following. For each route, we chose one
to three of the four runs to create a model. The remaining runs,
which did not become part of the model, were treated as a bus
241
in motion and we ran our algorithm to generate predictions
about those runs. These predictions were made for five to six
simulated user locations per route, and when the bus was
anywhere from one to nine minutes away from that simulated
user. We provided the algorithm with the model, a time to
treat as the current time of the system, calculated by
subtracting the expected prediction value from the actual
arrival time, and a set of location updates prior to this time,
accounting for location message delays (see VI. B). We
compared the prediction generated to the actual timing data
gathered for that run for a total of 547 measurements of error.
The error in a prediction is the absolute value of the
difference between the actual time it took for a bus to arrive at
the destination and the time predicted by the model. The
mean error from our predictions was approximately Figure 10 Mean error in the predicted arrival time, plotted by the
actual time to arrival (both axes in seconds).
94.9 seconds, with a standard deviation of 83.2 seconds. As
seen in Figure 10, the amount of error in the prediction
We examined 381 messages sent from the *box to the server
increases with the amount of time the bus actually is away
to measure this latency. The average latency, measured from
from the user. The error can be understood as a proportion of
the time it was sent by the *box to the time it was received by
the actual time to arrival of the bus, with the mean proportion
the server, is 19.96 seconds, with a range of 10 to 40 seconds
equal to approximately 37.9% with a standard deviation of
(standard deviation of 10.3 seconds). We consider this an
32.2%. For example, if the bus will actually arrive at a
acceptable latency for our particular application. Although
destination in 10 minutes, we can expect the prediction of this
these measurements were taken in Seattle, fieldwork over the
arrival time to be, on average about 3 minutes 45 seconds off.
years has demonstrated to the research team that cell coverage
In addition to the accuracy of our model, the amount of time
and SMS latency in Bishkek is consistent with that
required for SMS messages to travel through the network,
experienced in Seattle.
both those giving location updates and those from user
interaction, also affects the accuracy and usefulness of bus
2) User Query Latency
arrival time predictions. We examine these factors below.
Queries return information to users in the form of an
B. Message Latency estimated time (e.g. 5:30pm) that a bus is expected to arrive at
1) *box to Server Latency their location. The usefulness of this information depends
The accuracy of a particular prediction depends on both the upon both its accuracy and how quickly it is returned to the
accuracy of the route model and on receiving a recent, and user.
thus accurate, location update from the bus whose status is Factors affecting query processing latency include: server
currently being queried. Ideally, the server would have load, cell phone network load, and system limitations on rate
instantaneous information about the exact location of a of SMS message sends. Based on our measurements of
moving bus at the point it receives a user query about that bus. message latency in the previous section, we would predict the
In practice, the server must base its prediction on the last roundtrip time for the user to receive a response to their query
location update message received from that bus. Both the to be the latency of two SMS messages (assuming time spent
configured *box frequency of sending updates and any delay at the server is negligible), or approximately 40 seconds.
in that message getting from the *box to the server affects Considering that the typical usage scenario would return
accuracy of information about the current bus location. predictions on buses several minutes away, we consider this
Latency of SMS messages can be affected by a variety of latency to be acceptable.
factors, such as cell phone network congestion, signal
coverage for the *box or the server, or the need for messages VII. CONCLUSIONS AND FUTURE WORK
to cross networks. We recommend that the *box and the The *bus system is a direct response to an articulated
server be on the same cell phone network, as they were in our problem in the developing world, and it takes into close
implementation, to minimize this delay. consideration technology usage patterns and technology
infrastructure available in such resource-constrained
environments.
The system recognizes that information is valuable, and that
the lack of information inhibits certain kinds of growth and
efficiencies. However, there are not always central authorities
or institutions with the resources or motivation to supply
information infrastructure for citizens, and the *bus system
provides a mechanism by which decentralized operators of a
242
transportation system can create information resources for REFERENCES
themselves. It is also a system that uses mobile technology in
a way that does not require the direct participation of the [1] G. Weisbrod, “Models to predict the economic development impact of
transportation projects: historical experience and new applications,” The
mobile provider, thus cutting out one potential barrier to
Annals of Regional Science, Volume 42, Number 3 / September, 2008.
adoption. Finally, *bus enables community use of a single [2] S. Démurger, “Infrastructure Development and Economic Growth: An
GPS system to allow for user-generated geo-coding of the Explanation for Regional Disparities in China?” Journal of Comparative
environment. Economics, Volume 29, Issue 1, March 2001, pp. 95-117.
[3] D. Parikesit and K. Czuczman, “Transport, the missing link? A catalyst
Next steps for *bus include testing in Kyrgyzstan in early for achieving the MDGs,” id21insights, #63, July 2006,
2009. Our goal is to test both intra-city transportation in http://www.id21.org/insights/insights63/insights63.pdf
Bishkek, and inter-city transportation elsewhere, perhaps in [4] “Transport's Role in Achieving the Millennium Development Goals,”
Produced by the DFID Transport Resource Centre, August 2002,
another country. Although the *bus system has been largely Available:
built out with the concept of urban transport in mind, the http://siteresources.worldbank.org/INTWDR2004/Resources/22572_Tra
development impact for rural areas is conceivably even nsportsRole.pdf
[5] A. Bradbury, “Transport, mobility and social capital in developing
greater. In areas where distances are long, roads limited, and countries,” Proc. of the Institution of Civil Engineers, Engineering
vehicles intermittent, a system such as *bus could have Sustainability 159, June 2006 Issue ES2, pp. 79-86.
dramatic implications. [6] NextBus, Last accessed: 22 Sept. 2008. http://www.nextbus.com/
[7] Metro Tracker, Last accessed: 22 Sept. 2008.
Future work on *bus includes examining possibilities such http://transit.metrokc.gov/oltools/tracker.html
as incorporating some way to indicate the capacity of buses or [8] MyBus, Last accessed: 22 Sept. 2008. http://mybus.org/
allowing communication between drivers and riders, [9] The World Factbook, CIA, Last accessed: 22 Sept. 2008.
particularly for longer inter-city routes. A richer set of https://www.cia.gov/library/publications/the-world-factbook/
[10] Pew Internet and American Life Project, “A typology of Information
queries may be developed that allows users to subscribe to a and Communication Technology Users”, May 2006. Available:
notification system, so that they only need to send one SMS http://www.pewinternet.org/PPF/r/213/report_display.asp
message to the server in order to receive several bus arrival [11] J. Donner, “Research Approaches to Mobile Use in the Developing
World: A Review of the Literature”. The Information Society 24(3),
time messages. Another potential expansion is to allow users 2008.
to share tagged location data. Further investigation into [12] “Improving Rural Mass Transit in South Africa - Team Smile,” Last
making the system more scalable by using multiple phones per accessed: 22 Sept. 2008.
http://blogs.technet.com/unlimitedpotential/archive/2008/07/03/improvi
server for high density areas and by running stress tests is ng-rural-mass-transit-information-in-south-africa-team-smile.aspx
needed. Lastly, because the current system is based on SMS [13] MSR India SMS Toolkit, Last accessed: 22 Sept. 2008.
usage which generally requires a certain literacy level, further http://www.codeplex.com/smstoolkit
[14] FrontlineSMS. Company Website, Last accessed: 22 Sept. 2008.
work could explore the possibility of voice interfaces. http://www.frontlinesms.com/
Whether moving goods to market or people to medical [15] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana Unwired:
clinics, transportation is a crucial resource for any community. Replacing PCs with Mobile Phones in a Rural Sugarcane Cooperative,”
Allowing individuals to find ways to maximize such shared in Information and Communication Technologies for Development,
2007.
resources is the goal of *bus. Our approach has been a method [16] B. Kolko, E. Rose, and E. Johnson, “Communication as information-
of problem-solving that is bottom-up, that conceptualizes a seeking: the case for mobile social software for developing regions,”
technology device as shared among the community, and that Proceedings of the 16th international conference on World Wide Web,
2007.
recognizes the importance of information in people’s everyday [17] P. Javid and T. Parikh, “Augmenting Rural Supply Chains with a
lives. Our goal with this paper has also been to bring to the Location-Enhanced Mobile Information System,” in Information and
fore the importance of transportation for both economic and Communication Technologies for Development 2006, May 25-26, 2006,
Berkeley, CA.
human development – an area of ICT for development that has [18] M. Chen et al., “Practical Metropolitan-Scale Positioning for GSM
not yet received much attention. Phones,” in UbiComp 2006: Ubiquitous Computing, Springer Berlin /
Heidelberg, 2006.
[19] MySMS, Last accessed: 22 Sept. 2008. http://mysms.sourceforge.net/
ACKNOWLEDGMENTS [20] SMSLib, Last accessed: 22 Sept. 2008, http://smslib.org/
We would like to thank members of the original *bus and [21] A. Karbassi and M. Barth, “Vehicle route prediction and time of arrival
estimation techniques for improved transportation system management,”
MySMS teams of the 2007-2008 CSE477 at UW: Martin in Proc. of Intelligent Vehicles Symposium, 2003. IEEE, June 2003,
Hecko, Naasir Ramji, Jordan Walke, Brad Campbell, Ruibo pp. 511-516.
Li, and David St. Hilaire. We would also like to thank CAICT
team members Cynthia Putnam, Emma Rose, Rebecca
Walton, Erica Johnson, Mark Licata, and also Professor
Medina Aitieva, our local research team, and research
participants. We thank Aidai Seidakmatova, Shairbek Juraev,
and Natalie Linnell for their contributions. A special thanks to
Bruce Hemingway and Jesse Dosher for the design and
construction of the hardware.
243
ABSTRACT -- The healthcare milieu of most developing countries is often characterized by multiplicity of health programs
supported by myriad of donors geared towards reversing disease trends in these countries. However, donor policies tend to
support implementation of vertical programs which maintain their own management structures and information systems. The
emerging picture overtime is proliferation of multiple and uncoordinated health information systems (HIS), that are often in
conflict with the primary health care goals of integrated district based health information systems. As a step towards HIS
strengthening, most countries are pursuing an integration strategy of the vertical HIS. Nevertheless, the challenges presented
by the vertical reporting HIS reinforced by funds from the donors renders the integration initiatives ineffective, some ending
up as total failure or as mere pilot projects. The failure of the systems after implementation transcends technical fixes. This
paper drew on an empirical case to analyze the challenges associated with the effort to integrate the HIS in a context
characterized by multiple vertical health programs. The study revealed the tensions that exists between the ministry of health
which strived to standardize and integrate the HIS and the vertical programs which pushed the agenda to maintain their
systems alongside the national HIS. However, as implied from the study, attaining integration entails the ability to strike a
balance between the two forces, which can be achieved by strengthening communication and collaboration linkages between
the stakeholders.
have electricity. Similar problems of uneven (HMIS), meant to provide information support to all
infrastructure development was experienced in decision making processes of the entire ministry. In
Ethiopia, where according to [18], use of standardized this paper, the term ‘Health Information System
(HIS)’ is used to refer to the HMIS and the term
data formats served as gateways between the paper
‘HMIS unit’ is used to refer to the national level
based systems at the periphery levels and computer MoHSW department, responsible for the health
software at the higher levels of the health system information system.
hierarchy. Moreover, [10] alluded to some of the
The study was done as part of an ongoing action
challenges faced by most developing countries in
research on health information system restructuring
achieving standardized and integrated HIS to include: process undertaken by the Health Information System
conflicting interests between health programs which Program (HISP) in collaboration with the MoHSW
make it difficult to reach a “final” agreement; changes and other development partners in Zanzibar. HISP is
being the only constant, where new needs keep a global action research and development network on
popping up (e.g. HIV/AIDS); and multiple software health information systems by the University of Oslo
and paper tools which are difficult to coordinate and in Norway and other stakeholders, which started in
South Africa in 1994 and thereafter spread to other
change.
developing countries such as Ethiopia, Vietnam,
Whereas, some countries have managed to India, Mozambique, Botswana, Tanzania Mainland,
standardize and integrate some of the vertical Malawi and Nigeria. In all these countries, the
programs information systems into the national HIS, primary goal of HISP is to design, implement, and
ensuring continue reliance and use of the integrated sustain Health Information System through
system by the vertical programs is still a big participatory approaches to support local management
challenge [16]. As presented by [21], some of the of health care delivery and information flows [10]. In
reasons for the categorical programs refusal to 2005, Zanzibar became a node in the HISP network,
support an integrated HIS include: fear that their when the University of Oslo was contracted by the
requirements will not get the attention needed to MoHSW to engage in the restructuring of the HIS.
ensure that their needs are met, and if the programs Case study methodology was drawn upon in the field
have pride of ownership in their existing vertical during data collection and analysis. Case study is an
systems, which meet their needs. The fragmentation appropriate approach for bringing an understanding of
of the HIS after integration, ensuing from the non a complex issue, which could be a program, event, an
reliance or refusal to support the integrated system, activity or a process involving one or more
has not received much attention empirically and individuals and using a variety of data collection
analytically by the contemporary research in procedures over sustained period of time [11]. The
information systems integration. As the literature aim of the research being to develop a comprehensive
depicts, much of the research focused on the understanding of the challenges related to the way
challenges in the process of achieving integration. users received and engaged with the integrated HIS,
The argument of this paper is the need to comprehend case study proved to be a feasible approach.
the nit gritty of what goes on after integration. By
specifically understanding the way vertical program The research was carried out in three health districts;
managers (as users of the HIS) receive and relate to two in Unguja and one in Pemba Island within a
the newly implemented system, can help us answer period of five months (June to November, 2006).The
though partly as to why some health information districts visited in Unguja were Urban and West; both
systems integration in the context of developing of them located on the west region of the island. Most
country fail or run short of expectations, after of the MoHSW main offices such as HMIS unit and
implementation. vertical programs’ offices are located in these two
districts. In Pemba the research was conducted at
III. RESEARCH CONTEXT AND METHODS Chake Chake district, which is located on the south
The research study was conducted in Zanzibar. region of the Island.
Zanzibar is a semi-autonomous region within the
United Republic of Tanzania, which is made up of Semi structured interviews were used, where
two main islands, Unguja and Pemba, and several interviewees were asked open ended questions to
others islets located in the Indian Ocean. Zanzibar has elicit their viewpoints related to the use of the
an area of 2,332 square kilometres (the total area of integrated HIS. At the health unit level, 38
Tanzania is 945,000 Km), and is divided into five informants were interviewed, 19 in Unguja and 19 in
administrative regions, each with two districts, Pemba, where impressions and perceptions of the
making a total of ten districts in the entire region. health workers in relation to the previous and the
Zanzibar maintains its own health system that is newly integrated HIS were gathered. The goal was to
administrated by a semi-autonomous Ministry of learn micro level challenges emanating from the way
Health and Social Welfare (MoHSW). Side by side users responded to the integrated HIS. At the macro
with the health system is an information system level the interviews involved the following
called Health Management Information System informants: District medical officers, District health
246
officers, vertical programs district and general adapted in South Africa and subsequently adopted in
managers and central level HMIS unit officials. other developing countries such as Malawi,
During the interviews different informants’ Mozambique, India, Ethiopia, Namibia, Zambia,
viewpoints in relation to the new datasets and tools Mali, Botswana, Vietnam and Nigeria.
were gathered. Table 1 depicts the number of
informants interviewed. The HMIS restructuring process followed a
participatory and incremental approach in the revision
Name /Position Informants of the previous datasets and creation of new once.
Health unit level staff 38 The incremental approach paved way for the
District Medical Officers 3 involvement of key stakeholders in each stage of the
Vertical programs data managers 7 revision exercise. Among the new datasets included
Programs general managers 2 was the Expanded Programme on Immunization
National level HMIS unit officials 3 (EPI), Disease surveillance, Reproductive and child
District Health Officers 3 health (RCH), STI and HIV, and Maternity dataset.
HISP team members 2 Almost each dataset brought together a number of
TOTAL 58 stakeholders for instance the Disease surveillance
dataset apart from the national HMIS, had two main
Table 1: Interviews conducted at health facility and above
stakeholders, the EPI and Malaria programs. Malaria
A number of documents and software tools were program in Zanzibar is very strong unlike in other
analysed in the field. For instance the district countries due to many interventions by donors
implementation plan was analysed in an attempt to directed towards reversing the malaria trends in the
understand the use of the new system in the country. This made malaria related data more
preparation of the plan. The software tools were valuable purposely because of the need to monitor
analysed to assess compliance to the newly integrated and evaluate the interventions. The disease
system. surveillance being the main source of the malaria
data, the program was involved in the design
IV. CASE DESCRIPTION implementation stages to ensure that the dataset met
The health information system in Zanzibar was the requirements needed. For instance, the program’s
organized haphazardly and mainly shaped by the data manager was involved in the training of the new
organization of fragmented vertical programs with disease surveillance dataset. However, despite of the
their own information systems [14]. The vertical involvement the data manager enacted separate data
programs include the Family Planning (FP), Malaria, collection tool behind the scene and circulated it in
and Expanded Program on Immunization, some districts. When asked why, the answer was very
Tuberculosis and Leprosy, Nutrition, HIV/AIDS and clear,
Safe motherhood programs. Most of the programs’ “They have taken out almost all the age
services are integrated at the health unit level, but group categories and left what they feel will
maintain separate reporting systems. A situational satisfy their needs, but what about us. So we
analysis of the health information system disclosed tried to design it to show them how it should
plethora of problems which included scarcity of look like” (Manager, Vertical program,
resources, gaps in data collection tools, poor analysis August, 2006)
of data, fragmentation of the higher levels, poor Likewise, the EPI Disease surveillance tool (see
feedback and lack of motivation and limited Figure 1) was integrated in the new dataset for
information use [14]. disease surveillance.
As an attempt to improve the situation, a strategic
plan was drafted by a joint team of stakeholders;
comprising both scientific and organizational
researchers, major Ministry of Health and Social
Welfare (MoHSW) donors (DANIDA and WHO),
University of Oslo and some officials from the
MoHSW (ibid). The roadmap detailed and agreed on
the major activities to be undertaken. The University
of Oslo under its action research program called
Health Information System Program (HISP), was
contracted to undertake the task of restructuring the
system by way of integrating the highly fragmented
HIS. HISP as explained hitherto is a research network
that aims at enhancing district health information
systems in developing countries through introduction Figure 1: Monthly disease surveillance report
and local adaptation of open source software. The
software which is known as District Health Although the new dataset was running for more than
Information Software (DHIS) was developed and eight months (till the time of writing), still EPI kept
247
V. ANALYSIS AND DISCUSSION RCH tool where the family planning data elements
The integration of the health information system were minimized. Most of the users argued that they
(HIS) involved standardization of the datasets, data were using the old tools to ensure that information
collection tools, data processing tools and associated owners get all the information they needed. In
work practices; and institutionalization of the addition, by building on their past practices, some
standards to the levels of the health system. district officers kept on enquiring about the separate
Following is the analysis and discussion of the family planning reports from the health facilities.
challenges resulted from the way vertical program
managers received and related to the newly integrated Health officers in some districts drew on their limited
system. knowledge of DHIS software to enact limited use of
the integrated HIS, where preparation of monthly
Limited use of the integrated HIS reports was done by aggregating manually a number
The integration initiative sought to align different of datasets, an exercise which took two to three days
disparate fragmented information systems to form an to finish. While at the same time, all datasets were
integrated data repository at the district level which already entered in DHIS by the data clerk, in which a
gives access of data to different stakeholders. monthly report could be printed out without much
Though, this was achieved to a certain degree, some hurdle. The decision to go back to the manual
other vertical programs whose datasets were aligned systems can be envisioned as to have been made due
with the new initiative enacted limited use of the to the dilemma of not knowing how to use the new
newly integrated HIS by running some of their systems but also by the demand of the need to
previous information systems and tools in parallel compile monthly reports.
with the new system. For instance, the disease
surveillance dataset for EPI program was aligned with Other district health officers enacted limited use by
the national disease surveillance standard. Though the engaging themselves more with a dataset for a
national dataset was in operation for more than eight particular vertical program. This came to play as the
months until the time of writing, the vertical program result of the officers drawing on their past
managers decided to run their disease surveillance experiences of vertical system mindset contrary to the
data collection tool in parallel with the national new integrated HIS where all the datasets needed to
system. This vertical program however had another be afforded equal attention. This was evident from
dataset (on immunization) which was aligned with the what one health officer provided as an answer when
new initiative and which was running smoothly. The he was confronted for poor performance of some
immunization dataset is an independent dataset, with datasets, where he asserted that his responsibility was
the EPI program as the main stakeholder. This is a particular dataset whose performance was good.
different from the disease surveillance dataset which The limited engagement with the integrated HIS was
was formed by fusing together the national disease further observed at the health unit level where health
surveillance dataset and the EPI surveillance dataset. workers dealt with particular datasets more than
To regain control of their disease surveillance system, others. This limited use at the health unit or district
the program managers decided to run their dataset level came to play as users acted on their past
alongside the integrated dataset for disease practices related to vertical programs to conceptualize
surveillance. the new system which entailed equal treatment of all
datasets to ensure availability of comprehensive data
Furthermore, the limited use came to play as the at all levels.
result of the lack of trust by the vertical program
managers to the capability of HMIS unit to ensure Completely none-use of the integrated HIS
sustainability of the integrated HIS. This was evident Some other actors determined to completely shun
from one of the program manager’s assertions that, away from the newly integrated HIS in spite of the
unless we are sure of getting our data, we can not fact that, they were involved in the design process of
abandon our system. Therefore, the managers drew datasets related to their program. HIV/AIDS program
on the history of poor economic status of HMIS unit managers participated in the design of a new dataset
and on their future prospects about the new system’s called STI and HIV dataset, which aimed at collecting
sustainability to make the decision to maintain their data related to both services. However, this dataset
system alongside the new system. was operational for about five months until the time
of writing, but since then the data collected was not
Another health program whose users enacted limited fetched and used by the program. The idea of having
use of the new system was the family planning. In both systems running in parallel may sound logical,
some districts, family planning dataset was running in because the setup gives every stakeholder access to
parallel with the newly integrated system despite the the data; contrary to the previous systems where data
fact that their data elements were integrated in the was vertically submitted to the programs and donors.
RCH dataset. This resulted as users drew on their past However, the setup has enormous implication on the
practices of submitting huge amounts of data to the workload to the data collectors. Rather than
vertical programs unlike in the new standardized rationalizing the fragmented systems and minimize
249
duplication of data, the setup intensify it and HMIS unit to solve availability of data collection
ultimately jeopardize the quality of the data collected. tools problem.
Nevertheless, it must be emphasized that the tensions
The none-use of the integrated HIS, came to play as which resulted from limited and none-use of the
the program officers’ drew on a number of newly integrated HIS were not static, but dynamic in
assumptions and on their multiple needs of data. nature. The level of dynamism was different from one
Some of the assumptions in relation to the new user group to another and from one vertical program
system are that, the new system was for HMIS unit, to another. For instance, the HIV/AIDs which initially
and so their participation aimed at helping them get opted for none-use of the integrated system, at the
their data and that the data collected in the new very end of the field study in an interview with the
system did not satisfy their needs. The assumptions data manager; he sanctioned the software to be
can be envisioned to have resulted from users installed in their computers. His acceptance came due
drawing on their past experiences of collecting to the need to make comparison between the data they
multiple data and on their desire for data to meet collect and the HMIS data, supplementing their data
future needs. in case of some missing data in their systems. Though
the manager took the decision after learning that
The none-use enacted by the HIV program as was the having the new system won’t prevent him from using
case for the limited use explained above, was also their systems, but I see this as a movement from
mediated by the poor economic conditions of the none-use to limited use. Furthermore, the health units
HMIS unit, which led to mistrust of the vertical and districts officers who opted for limited
program to the capability of HMIS unit to ensure engagement by drawing on their previous vertical
sustainability of the system over a period of time system experiences, slowly started to change as they
taking into account its almost total dependency on learned through informal trainings, feedback meetings
donors. and seminars that all the datasets needed to be
afforded equal significance. The change from none-
Tensions in the HIS Integration use to limited use conforms to the argument that,
The resulting picture from the limited and none-use other embedded practices can be changed or replaced
of the newly integrated HIS, is what I dubbed as by others over time [18]. However, contextual
‘pulling effect’ (see figure 2). particularities such as the poor economic status of
National HIS Vertical Programs most healthcare settings in developing countries
sometimes reinforce their existence and hence make
Struggle to Want to maintain
them hard to change.
standardize & their own systems
integrate the HIS (fragmentation)
Advocate on Advocate multiple The need to strengthen communication and
minimum action-led data needs to satisfy collaboration linkages between stakeholders
datasets multiple and The vertical programs in the Zanzibar case have very
No enough funds to heterogeneous
sustain the system donors
strong installed base of information systems
Use of participatory Have funds to reinforced by funds from donors. Moreover, these
approaches to build develop & maintain systems have very well defined and elaborate vertical
trust their systems organizational structures with many people employed
in there, further reinforcing their strengths and the
Pulling Effect tension towards any change attempts. For instance,
Figure 2: Tensions between the national HIS and the vertical the HIV/AIDS program maintains its own
programs administrative structure and employees working on
On one side HMIS unit under the MoHSW is the information system.
struggling to standardize and integrate the fragmented
information systems and on the other side the vertical As pointed out earlier, although most of the vertical-
programs opted for limited and none-use of the newly reporting systems were harmonized and aligned in the
integrated HIS by maintaining their own systems. The integration initiatives, some of the systems were still
upshot of that is a pulling effect on either side where running alongside the integrated system. This state of
the winner is determined by the power (e.g. to argue, inertia could be explained partly by the perceived
funds, human resources, good strategies.), which one results of integration, including fear to lose positions,
of the two sides need to have in order to haul the those with vested interests with the old systems to
opponent. lose them and the mistrust on the capability of the
national HIS in managing and sustaining the
In an attempt to alleviate or eliminate the tensions, integrated system.
the HMIS unit ventured to consensus building
through participatory approaches (PA) like meetings, The argument as implied from the case is that
workshops and seminars to try to strike a balance communication and collaboration between all the
between the two forces. This is exemplified by necessary stakeholders need to be built and
mobilization of concerted efforts and funds by the strengthened as a strategy to deal with the inertia of
250
the vertical reporting systems. The national HIS [23]. This is further reinforced by [21], who argues
however, need to take a stewardship role to ensure that pooling of resources by the categorical health
that comprehensive data is obtained from the programs to the integrated HIS is a difficult
disparate systems in a cost effective way. This further undertaking. The challenges related with the pooling
suggests that some of the vertical systems to run of resources underline the critical need of the
alongside the national HIS, but with the mandate that proposed strategy of strengthening communication
the national HIS take the driver’s seat in ensuring and collaboration linkages and the need to make use
smooth collaboration and communication between the of gateways to ensure flow of information from
stakeholders. This integration perspective is inline vertical systems to the national HIS.
with the concept of accepting to live with a
reasonable level of none integration, since no one, VI. CONCLUSION
including the national health authorities, is in Integration of HIS in developing countries
‘control’ in any strict sense; and therefore a relevant characterized by multiple vertical programs is quite a
strategy cannot be based on a planning or control challenge. It involves economic and political
approach [8]. processes in articulating interests, building alliances
and struggling over outcomes [19]. The tensions
Hand in hand with the need to strengthen
between the national HIS which strive for integration
communication linkages, is the need to use gateways
to link between the national HIS and the vertical and the vertical programs which advocate their
programs information systems which seems strong information systems supported by donors represent
and hard enough to integrate with the national HIS. both challenges and opportunities to the integration
Gateways allow the continuous existence of multiple initiatives. One of the challenges as implied from the
systems, each with their internal organization and study is ensuring compliance of the integrated HIS by
logic (ibid). Consequently, gateways can ensure the vertical programs in the face of poor economic
transfer of data from strong multiple vertical systems
status of most developing countries. However, as
to the national data warehouse, and therefore make
the data available to all stakeholders in a cost implied from this study, strengthening the
effective way. Hence, the use of gateways supports collaboration and communication linkages between
the concept of accepting a certain degree of none the national HIS and the vertical programs represent
integration, while keeping every stakeholder ‘happy’. an opportunity to curb the integration challenges.
While the paper propose this perspective as a strategy
From the case, the communication and collaboration
towards HIS integration, more empirical research is
perspective is exemplified by the approach used by
the HMIS Unit in resolving the availability of data needed to find out more how it can be achieved in
collection tools problem. The problem happened practice.
when the unit failed to fund production of data
collection tools due to financial constraints. To solve The message the paper sends to public health and
the problem, which if left unsolved would have rolled other practitioners in HIS is that misconstruing or
back the entire system; the HMIS unit summoned all under-estimating the non-technical issues of
key stakeholders in a workshop to deliberate on the integration can account for a very significant portion
strategies to solve the problem. In the workshop, it of failures of information systems in healthcare.
was unanimously agreed that each stakeholder While technically, the integrated national HIS
(vertical program) contributes some funds for consisted of datasets catering for the needs of the
production of the tools. As the results, production of vertical programs covered by this study, the programs
tools to be used for a period of one year was made however kept on using their previous systems,
possible through communication and collaboration presenting challenges to the integration initiatives.
between the stakeholders but with the HMIS unit Hence, the paper underscore the need to understand
taking the stewardship role. The national HIS using the socio economic challenges related to HIS
communication and collaboration processes can play integration which goes beyond technical fixes. The
the same role to ensure availability of comprehensive study further reiterates the need to look beyond the
data. integration process, to include the way users receive
and engage with the newly integrated HIS.
The strategy used in Zanzibar of soliciting funds from
vertical programs may point to the idea of donors as VII. REFERENCES
the main supporters of the health programs to pool [1] B. Chilundo, Integrating Information Systems of Disease-
resources together as in the case of the Sector Wide Specific Health Programmes in Low Income Countries: The
Approach (SWAp) [22], to support the integrated Case Study of Mozambique, Ph.D. thesis, Faculty of
national HIS. However, despite the SWAp policy Medicine, University of Oslo, 2004.
mandate of strengthening the local capacity to [2] B. Chilundo and M. Aanestad, Vertical or Integrated
manage funds from donors, evidence shows that Programmes? The consequences for the laboratory
information systems in Mozambique, In Proceedings of the
countries such as Tanzania is facing difficulties in
IFIP TC8 & TC9/ WG8.2+9.4, 2003.
coordinating the funds using the so called basket fund
251
[3] F. Mukama, A study of health information systems at local R. Montealegre and A. Poulymenakou (Eds), Athens, Greece:
levels in Tanzania and Mozambique, Improving the use and 35 -51, 2003.
management of information in health districts, Master thesis,
Department of Informatics, University of Oslo, 2003. [21] G. Land, N. Hoffman and R.Peterson, ”Developing the
[4] J. Braa, E. Monteiro and S. Sahay, Networks of Action, MIS
Missouri Integrated Public Health Information System”, In
Quarterly 28(3): 126, 2004.
O’Carroll PW, YasnoffWA,Ward ME, Ripp LH, Martin EL,
[5] J. Braa, O. Hanseth, W. Mohammed, A. Haywod, and V.
Shaw, Developing Health Information Systems in Developing eds. Public Health Informatics and Information Systems. New
Countries: The Flexible Standards Strategy. In press, MISQ, York: Springer; 2002, 617–64.
Special Issue on IT and Development, 2005.
[6] J. Lungo, Data flows in Health Information Systems: An [22] Brown A, “Integrating vertical health programmes into
action research study of reporting routine health delivery sector-wide approaches: Experiences and lessons”. Institute
services and implementation of computer databases in health for Health Sector Development, on behalf of the Swiss Agency
information”. Master thesis, Department of Informatics, for Development and Co-operation (SDC) 2001, Pages: 30.
University of Oslo, 2003. [23] Smith, M., Madon, S. and Anifalaje, A. ”Integrate Health
[7] J.J.Rubona, Routine health information systems that operate Information Systems in Tanzania: Experience and
in Tanzania”. In the RHINO workshop on issues and
innovation in routine health information in developing Challenges”, EJISDC (2008), 33, 1, 1-21.
countries, The Bolger Center, Protomac, MD, USA 14-16
March 2001. Arlington, VA 22209, USA: MEASURE
Evaluation, JSI Research and Training Institute, 183 –193.
[8] M. Aanestad, ., E. Monteiro, H. Kimaro, E. Macombe, G.
Macueve, F. Mukama, H. Muquingue, L.J Nhampossa, and J.
Lungo, Strategies for development and integration of health
information systems: coping with historicity and
heterogeneity. Working paper at the University of Oslo,
Department of Informatics, 2005.
[9] N. H. Shidende, Challenges and approaches to the integration
of HIS: Case study from Tanzania”, M.Sc. thesis. Department
of Informatics, University of Oslo, 2005.
[10] Braa, J. and Humberto M (2007): “Building collaborative
networks in Africa on health information systems and open
source software development – Experiences from the
HISP/BEANISH network”, IST Africa, 2007.
[11] R. Yin, Case study research: Design and methods (2nd ed.).
Thousand Oaks, CA: Sage Publishing, 1994.
[12] RHINO, The second RHINO workshop on: Enhancing the
Quality and Use of Routine Health Information at the District
Level”, September 29 – October 4, 2003.
[13] World Health Organization (WHO), Health information
systems in support of the Millennium Development Goals,
2006.
[14] Y.H. Sheikh, Improving health information systems at health
districts in Zanzibar, M.Sc. thesis. Department of Informatics,
University of Oslo, 2005.
[15] R. Sauerborn, and T. Lippeveld, Introduction. In: T.
Lippeveld, R. Sauerborn, and C. Bodart, eds. Design and
Implementation of Health Information System. Geneva,
World Health Organisation, 1-14.
[16] M. Galimoto, Integration of Health Information Systems:
Case Study from Malawi, MSc. thesis, University of Oslo
2007.
[17] Sahay, S., E. Monteiro, and M. Aanestad, “Towards a
Political Perspective of Integration in IS Research: the case of
Health Information Systems in India”. Working paper,
Department of Informatics, University of Oslo, Norway,
2006.
[18] Lagebo, B. and S. Mikonnen (2005): “Challenges and
Approaches to Scaling and Standardizing Health Information
Infrastructure in Developing Countries: Case Studies from
Ethiopia”. M.Sc. thesis, University of Oslo.
[19] B. Chilundo and M. Aanestad, Vertical or Integrated
Programmes? The consequences for the laboratory
information systems in Mozambique, In Proceedings of the
IFIP TC8 & TC9/ WG8.2+9.4, 2003
[20] Mosse, E. and S. Sahay, “Counter Networks, Communication
and Health Information Systems: A Case Study from
Mozambique”. In The IFIP TC8 & TC9/ WG8.2+9.4
Working Conference on Information Systems Perspectives
and Challenges in the Context of Globalization. M. Korpela,
Cross Technology Comparison for Information 252
Abstract— This paper introduces a framework to examine the services may not be effective for disaster response. This paper
relative effectiveness and cost-effectiveness of different looks into the relative effectiveness and costs of different
information and communications technologies to deliver a range technologies to deliver different social services in rural
of social services in rural Bangladesh. It focuses particularly on
Bangladesh, focusing on major development sectors such as
major sectors like agriculture, education, disaster response and
healthcare. Expert elicitation (using both local and international agriculture, education, disaster response, healthcare etc.
group of experts) survey on ICT effectiveness by service domains This analysis has several components. First, expert
shows localization as the key determining factor for any opinions on ICT effectiveness by service domain (health,
technological intervention. Community based radio broadcasting education, etc.) for various ICTs (TV, radio, print, telecenters,
(CR) appears as the dominant option among the considered etc.), with international and local (Bangladeshi) experts in
ICTs. Human intervention has been found to be crucial in both
communications and development were gathered via expert
low (20%) and medium-high (60%) literacy populations. Our
study also shows public funded Terrestrial TV is more effective elicitation surveys, taking care to manage anchoring and other
and has higher acceptability in providing information over biases (detailed subsequently). The elicitations were for low
privately owned Satellite/Cable TV channels. Literacy doesn’t (20%) and medium-high (60%) literacy populations. Second,
seem to significantly affect the relative effectiveness of we created an extensive techno-economic model for the
Information Centers with Internet connection over Print and TV various ICTs. Finally, we made estimates of potential
based options. In addition, an extensive techno-economic model
penetrations of the ICTs (reach), to estimate the cost-
for the various ICTs and stochastic evaluation of potential
penetrations of the ICTs (reach) has also found CR to be the most effectiveness per user of the technologies for the particular
cost-effective option. We conclude with a proposed set of policy development domain.
and operating recommendations to enable effective technology The focus here is largely on the delivery of social services
based information services for rural Bangladeshi development. (supply side); analysis of issues such as citizens' rights to
Index Terms— ICTD, Rural Information Service, Community information, citizen participation in governance etc. is beyond
Radio, Telecenter.
the scope of the paper.
The primary objective of this study is to develop a robust
I. INTRODUCTION
framework that can facilitate choices by policymakers,
Figure 1 represents the relative status of the four South Asian countries in literacy, teledensity, newspaper readership and radio
penetration (data collected during 2004-2008 period).[16, 28-34]
5
This challenge is similar to radio or other media, where signal availability is necessarily only one measure, lower being actual listening time.
259
Table 5
Cross Technology Comparison for Multiple Effectiveness Dimensions
Figure 2: Relative effectiveness ranking of the considered technologies in both high and low literacy region, according to the judgment of the
International experts. It considers average effectiveness values of different technologies across all service domains. Individually, the experts
have shown higher variance in effectiveness estimations (details in Section V)
Bangladeshi sample also has much greater variance than the (Agriculture: General Extension). In these figures, the Y and
international sample, especially for technologies not now in X axes represent the effectiveness scores and plausibly
use in Bangladesh [47]. achievable technology penetration respectively. The spread of
This scenario has been reflected in the effectiveness ranking, effectiveness scores given by individual experts for various
which fails to reflect the large potential of RIC with proper delivery systems (from 0 to 100 in the Y-axis) can be seen
web based opportunities. Table 6 summarizes the present together with the highlighted ones representing average
access information of different technologies, in a Bangladeshi values. The size of the average value data points correspond to
setting. the Cost-per Person-Year for each considered delivery
In figures 4, 5, 6 and 7, we see the graphical representations technology. According to our assumptions, the effectiveness
of comparative cost-effectiveness data, from both international score of both expert groups are independent of the potential
and local experts, for a sample application domain technology penetration or cost scenarios in Bangladesh.
Figure 3
Effectiveness Ranking of Different Technologies
(According to Local Experts)
Figure 3: Relative effectiveness ranking of the considered technologies in both high and low literacy region, according to the judgment of the
Bangladeshi experts. It considers average effectiveness values of different technologies across all service domains. Individually,
the experts have shown higher variance in effectiveness estimations (details in Section V).
261
Table 6 to be most effective technologies with which to develop an
Present Access Status of Different Communication Services in efficient disaster response infrastructure in both high and low
Bangladesh literacy communities.
Technology/ Penetration/Access Observation 5: In the healthcare sector, experts emphasize
Information Service Information [16] the higher effectiveness of proactive Face-to-Face
Option interventions by local healthcare workers. CR remains the
Radio 30.4% most effective information service option, in all literacy and
Terrestrial TV 61% service sector scenarios.
Satellite TV 12.4% Observation 6: According to the experts, CR alongside
Newspaper 26% localized print based options are more suitable in providing
Telecenters 1061 information related to financial services among the highly
(total number, literate population. For a low literacy community,
nationwide)[2] broadcasting options are considered to offer superior
information services.
We integrate all these different types of data in a single A. Technology-specific Key Observations
graphical display for each information service domain and in Examining which technology works best for which
two types of literacy based scenarios to: 1) have a information service provides the basis to further examine the
comprehensive look at the relative effectiveness of several effectiveness trends for different types of technologies, their
technologies as information service providers in the target potential impact in future proliferation of certain information
service sectors, and 2) to examine their present positioning in service, the regulatory and policy implications on certain
terms of plausible local or national penetration level. Service technology usage etc. Here technologies here been classified
Sector specific Preliminary Observations within broader groups in order to come to some general
In order to find out the relative effectiveness of different observations.
technologies to deliver a range of information services, we 1) Broadcasting based Information Service Options
first focus on the service sector specific performance in both Observation 1: Radio broadcasting dominates the
high and low literacy scenarios. technology based service delivery mechanism with its literacy
Observation 1: For Agriculture General Extension, CR independent potentials in localized and participatory
was ranked as the most effective technological intervention, communication.
independent of the literacy rate within any community. For Observation 2: CR leads in cost effectiveness in all service
high literacy regions, print based options dominate sectors through ensuring localization and community
broadcasting alternatives. RIC-1 is ranked third overall in participation.
effectiveness, followed by Local Newspaper, Public AM Discussion: CR has been found to be an effective mode of
radio, National Newspaper and others. For communities with communication, independent of service sector type and the
low literacy, Public AM turns out to be the second most state of any target community’s literacy. The apparent
effective way to deliver general extension information among universal efficacy of CR can be attributed to its focus on
the less literate population, followed by RIC-1, Terrestrial TV, localized programming, participatory communication,
RIC-2, Local Newspaper etc. appropriate content development etc[5].
Observation 2: For Agriculture Market Information, CR Observation 3: Broadcasting based information service
remains the top choice. Here in high literacy area, Public AM options are the key behind any effective Disaster Response
radio is considered to be more effective, followed by mechanism.
technological options with higher localization options (RIC-1, Discussion: The effectiveness of broadcasting based
Local Newspaper, Printed Brochure). In low literacy areas, disaster response mechanisms is unparalleled, primarily due to
broadcasting options (Public AM, Terrestrial TV and Satellite their nearly universal acceptability, point-to-multipoint
TV) dominate the effectiveness ranking, followed by RICs and (broadcast) capabilities, and speed in disseminating the
print based options. required information.
Observation 3: In Education (Lifelong Learning), apart Observation 4: Public funded Terrestrial TV has higher
from CR, in higher literacy areas, with a wide range of options effectiveness and acceptability in providing information in
available using the Internet and multimedia, RIC-1 is comparison with privately owned Satellite/Cable TV channels.
considered to be the second most effective delivery Discussion: In this research, we assume that higher literacy
technology followed by Public AM, Local Newspaper, Printed closely correlates with higher economic affluence, an indicator
Brochure, RIC-2 and the other broadcasting alternatives. For for people to opt for “better” communication technologies. In
low literacy regions, broadcasting options overwhelmingly Bangladesh, Satellite/Cable TV is less popular (and has lower
lead the relative effectiveness ranking. penetration) than Terrestrial TV[16]. It shows the positive
Observation 4: When a country like Bangladesh is under externality a nation’s broadcasting regulations can create
threat of imminent calamities [48], radio broadcasting (CR, through mandating the incumbent public broadcasting service
Public AM) and TV (Terrestrial and Satellite) are considered air social service information programs. Generally, private
262
Figure 4
Cross Technology Comparison for providing Agriculture (General Extension)
based Information Service in a high literacy ( ≥ 60%) area (International Experts)
E 100
Face-to-Face
f CR
f
e 80
c Printed
Public AM Brochure
t RIC-1
i
v 60
e Local
n Newspaper
Satellite TV RIC-2
e
40
s National
s Newspaper
Terrestrial
20 TV
0
0% 20% 40% 60% 80% 100%
Likely Achievable Technology Penetration
in Bangladesh
Figure 4 represents the relative effectiveness (judgment of the international experts), penetration and cost-per person-year status of each
considered technologies in providing Agriculture (General Extension) related information service in a high literacy ( ≥ 60%) community. The
X and Y axes correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness
scores for various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted
ones representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. CR is the most cost-effective technology based option here, with tradeoffs of penetration, cost, and effectiveness between the
next choices of face-to-face, Rural Information Center with Internet connection, and local Print media.
3) Rural Information Centers
satellite channels have fewer regulatory obligations to abide Observation 7: Literacy doesn’t affect the relative
by and little or no financial incentives in airing societal effectiveness of RIC-1 over Print and TV based options.
programs, let alone localized versions. Satellite also has Discussion: According to the experts in this research, on
greater challenges of localization than many other average in both high and low literacy scenarios, the RIC-1
technologies based on its very wide footprint. centers are (or, perhaps, can be?) more effective than some
2) Print Based Information Service Options major TV broadcasting and locally published print based
Observation 5: Local Print based options are more effective options. Especially with the help of its information
than the national ones. intermediary (“ informadiary”) mechanism, local database and
Discussion: The general population can be effectively cellular phone integrated applications, RIC-1s have a lot of
reached with locally published newspapers or printed growth potential as a relatively new technological option[49].
brochures in comparison with the big newspapers with Observation 8: RICs without Internet can still be utilized as
nationwide circulations. Community ownership, localization a successful tool for providing community education.
of the content development and easy accessibility are reported Discussion: As we have found in this research, RICs
as some of the key factors responsible for this trend. without web enabled facilities (RIC-2) are less effective than
Observation 6: Print based options can be used as effective RIC-1s as well as the majority of the other interventions in
auxiliary mechanism for societal services that require both high and low literacy communities.11 But in the education
personalized attention. domain, RIC-2 shows higher effectiveness than mainly
Discussion: In lifelong learning, financial service advising, National Newspaper and Satellite TV. Based on discussion
health care and agriculture-based information services in any with the development experts in S. Asia, we found that the
high literacy region, community based print options are more relatively lower demand of updates in lifelong learning
effective than the visual broadcasting media, according to the curriculum is considered to be one of the main reasons behind
effectiveness ranking and comments of the international RIC-2’s better performance in this specific service sector.
experts. Usually NGOs and development agencies in the field
use this type of intervention as a supporting tool in addition to 11
An RIC-1 could always operate without Internet (RIC-2) in theory, but
their broader door-to-door personalized campaigns. in practice, this is rarely operationalized.
263
Figure 5
Cross Technology Comparison for providing Agriculture (General Extension)
based Information Service in a low literacy (≤20%) area (International Experts)
100
CR Face-to-Face
E
f
80
f
e Public AM
c
t Terrestrial RIC-1
60
i TV
v
e
n 40 Satellite TV Local
e Newspaper
s
s RIC-2
20
Printed
National Brochure
Newspaper
0
0% 20% 40% 60% 80% 100%
Likely Achievable Technology Penetration
in Bangladesh
Figure 5 represents the relative effectiveness (judgment of the international experts), penetration and cost-per person-year status of each
considered technologies in providing Agriculture (General Extension) related information service in a low literacy community. The X and Y
axes correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted ones
representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. CR is the most cost-effective technology based option here, with tradeoffs in cost, penetration, and effectiveness in the next
choices of face-to-face, Public AM and Rural Information Center with Internet connection.
100 Face-to-Face
E
f
80
f CR
Terrestrial
e Public AM TV
c
t 60
i RIC-2
Satellite TV
v Local
e Newspaper
n 40 RIC-1
e
s National Printed
s Newspaper Brochure
20
0
0% 20% 40% 60% 80% 100%
Likely Achievable Technology Penetration
in Bangladesh
Figure 6 represents the relative effectiveness (judgment of the local experts), penetration and cost-per person-year status of each considered
technologies in providing Agriculture (General Extension) related information service in a high literacy community. The X and Y axes
correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted ones
representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. Face-to-Face is considered to be most effective. Among the technology based options, CR is the most cost-effective one,
followed by Terrestrial TV and Public AM
.
• With literacy independent access efficiency, RIC-1 can development, management etc. [11, 52] Regular in-
provide a platform for sustainable Information (this service training for personnel and resource pooling in
implicitly assumes the use of informadiaries). The PK remote places (i.e. sharing of content, technicians etc.)
inspired proactive approach reduces the "barrier to entry" also can ensure the operational sustainability of any
factor for local information service facility [46]. information service center. This is vital for long-term
Following the successful example of Nepal [50], a sustainability beyond the initial period.
balanced integration of Face-to-Face, CR, Print and
A. Additional Policy Suggestions
Internet based multimedia activities can address the
majority of the information service needs in South Asian These draw from literature, our analysis and
region. suggestions from the surveys:
• Efforts to demystify the technologies used for information • Effective government policies and regulations need to be
service centers need to be in place to ensure greater in place to create an enabling environment for
efficacy and wider participation of the common people. development-oriented information services. The policies
For example, low cost CR options such as the "suitcase should be effective in removing bottlenecks for the free
radio" can be used in the field for on- the spot flow of information by giving away incentives like
broadcasting and human resource development purposes. lowered or zero tax rates for imported equipments,
lowered license fees etc.
• Both the literature review and collected field data [11, 51]
identified "scarcity of trained human resources" as one of • Information service providers engaged in societal and
the major roadblocks towards establishing an effective development work need separation from competing with
information service network. Experiences from South commercial entities. The absence of such support
America and from Nepal showed the effectiveness of mechanisms can result in unfair competition, pressure on
decentralized support and resource centers for training service providers to overlook societal commitment for
information service providers on technical issues, content
265
Figure 7
Cross Technology Comparison for providing Agriculture (General Extension)
based Information Service in a low literacy (≤20%) area (Local Experts)
Face-to-Face
100
E
f Terrestrial CR
80 TV
f
e Public AM
c
t 60
i
v
e Printed RIC-2
n 40 Satellite TV Brochure
e
s
s RIC-1
20
Local
National
Newspaper Newspaper
0
0% 20% 40% 60% 80% 100%
Likely Achievable Technology Penetration
in Bangladesh
Figure 7 represents the relative effectiveness (judgment of the local experts), penetration and cost-per person-year status of each considered
technologies in providing Agriculture (General Extension) related information service in a low literacy community. The X and Y axes
correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted ones
representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. Face-to-Face is considered to be most effective. Among the technology based options, CR is the most cost-effective one,
followed by Terrestrial TV and Public AM. Rural Information Center without the Internet connection has been considered to be more effective
than the one with the web-enabled options.
greater commercial success and, at times, the abrupt extends to applications such as voice enabled and GSM
closure of such initiatives[4]. based help-lines (for providing healthcare, market
• A “Development Information Service Fund” can be information etc.), more hybrid solutions or innovations
created to provide monetary help for establishing such as Internet Radio.
information infrastructure in underserved communities. • Hybrid solutions between technologies, including face-to-
All the commercial broadcasting and communication face and mobiles, might offer the most promise. This
enterprises can contribute to this fund on a mandatory implies policies and regulations should not be spoiled by
basis. This type of mechanism is the backbone for parallel technology, but, rather, facilitates such convergence.
policies worldwide on Universal Service for telephony
[53]. VII. CONCLUSION
• On modeling and comparing program and labor costs of This analysis provides a comparative framework and an
community based information service providers in initial estimate of the cost-effectiveness across technologies
isolated versus more collaborative environments, we find for delivery of information services for rural users in
found that national, regional and local cooperation in Bangladesh (and elsewhere). Effectiveness analysis using
terms of program and resource sharing should be experts’ judgment and financial data modeling in this paper
encouraged officially, in addition to pooling resources for have demonstrated the strength of Radio based information
learning, training, troubleshooting etc. (given labor is a service options as useful and applied tools for development
significant fraction of operational costs [5]. activities in Bangladesh and S. Asia. Rural Information
• With the trend of transferring everything to be digitized Centers with net connections also turn out to be very
and being transferred to cellular phone media if not the promising as hybrid information access points (under the
Internet, the policy framework should support digital assumption of informediaries for extending their reach). We
information service initiatives, content sharing and mobile find awareness is a challenge in choosing technologies, given
wireless based information services in near future. This the differences between international and local (Bangladeshi
266
experts) – this implies dissemination of data and best practices [22] A. Opoku-Menash, "Radio, Conflict and Political Transition: The
Future of Community Radion in Africa," in Communication for
to be of value. Given the enormous differences in cost, Social Change, Anthology" Hostorical and Contemporary
effectiveness and penetration, often with a tradeoff (there is Readinngs, A. Gumucio-Dagron and T. Tuft, Eds.: CFSC, 2000,
rarely a dominant solution), policy-makers will need to decide pp. 856-963.
[23] A. Dock and J. Helwig, "Interactive Radio Instruction: Impact,
how to prioritize technologies in the form of enabling
Sustainability and Future Directions," World Bank, 1999.
environment. Regulatory and policy decisions can strongly [24] I. Pringle and M. David, "The Kothmale Model: Using Radio to
impact the relative costs of technologies as well as their Make the Internet visible," in The One to Watch: Radio, New ICT
penetration. Nonetheless, the broad portfolio of technologies and Interactivity, B. Girard, Ed. Rome: FAO, 2003, pp. 97-114.
[25] D. R. Holmes, D. M. Kamacharya, and J. K. Mayo, "Radio
available for Information Services for development implies Education in Nepal," in Distance Education for Teacher Training,
great potential for the region, with relatively low costs per H. Perraton, Ed. New York: Routledge, 1993, pp. 136-195.
beneficiary for all the technologies (under $1 per person-year [26] A. W. Bates, Technology, Open Learning and Distance Education.
London: Routledge, 1995.
in scale). [27] T. Hulsmann, "Costing Open and Distance Learning," 2000.
[28] "India's foremost television network - Doordarshan." vol. 2008.
[29] "CIA - The World Factbook -- India." vol. 2008, 2008.
[30] "Future Tense of Dailies ". vol. 2008, 2001.
REFERENCES [31] "CIA - The World Factbook -- Nepal," 2008.
[32] "CIA - The World Factbook -- Pakistan," 2008.
[33] A. Riaz, "Media and Judicial Independence in Pakistan," 2004.
[1] R. Rajora, Bridging the Digital Divide: Gyandoot—The Model for [34] "Teledensity growth in Pakistan almost double in FY 2006." vol.
Community Networks. New Delhi: Tata McGraw-Hill, 2002. 2008, 2006.
[2] "Bangladesh Telecentre Network." vol. 2008. [35] L. Tobing, How To Do Community Radio: A Primer for
[3] "Case Study Series on ICT-Enabled Development:I-NetworkCase Community Radio Operators. New Delhi: UNESCO.
Study Series on ICT-Enabled Development:I-Network." vol. 2008: [36] R. Iannella and K. Henricksen, "Managing Information in the
bridges.org, 2004. Disaster Coordination Centre: Lessons and Opportunities," in 4th
[4] I. Pringle and B. Subba, "Ten Years on: The State of Community International ISCRAM Conference (B. Van de Walle, P. Burghardt
Radio in Nepal," 2007. and C. Nieuwenhuis, eds.), Delft, the Netherlands, 2007.
[5] F. Hussain and R. Tongia, "Community Radio for Development in [37] M. Carnoy, "ICT in Education: Possibilities and Challenges,"
South Asia: A Sustainability Study," in IEEE/ACM International 2004.
Conference on Information and Communication Technologies and [38] L. S. Austin and K. Husted, "Cost-Effectiveness of Television,
Development Bangalore, India: IEEE, 2007. Radio, and Print Media Programs for Public Mental Health
[6] F. J. Proenza, "Telecenter Sustainability- Myths and Education ".
Opportunities," in Bridging the rural knowledge gap: Information [39] R. Kumar and M. Best, "Social Impact and Diffusion of Telecenter
systems for improved livelihoods: FAO, 2001. Use: A Study from the Sustainable Access in Rural India Project,"
[7] "Radio Sagarmatha.." vol. 2008, 2007. ITID.
[8] R. Mainali, "Volunteering and Community Participation in [40] F. J. Proenza, "Telecenter Sustainability- Myths and
Community Radio: Experiences from Nepal.," National Opportunities," in Bridging the rural knowledge gap: Information
Consultation for Practicing and Potential Community Radio systems for improved livelihoods: FAO, 2001.
Operators 2007. [41] S. Finquelievich and G. Kisilevsky, "Community Democratization
[9] "India: Time for Community Radio has come!," UNESCO, 2007. Of Telecommunications Community Cooperatives In Argentina:
[10] F. Noronha, "India still waits: Rural poor not yet ready for the The Case Of Telpin," The Journal of Community Informatics,
promise of radio." vol. 2006, 2004. 2005.
[11] C. Fraser and S. R. Estrada, Community Radio Handbook: [42] "South Asia Economic Report – Social Sectors in Transition,"
UNESCO, 2001. 2007.
[12] J. Antin, "Cultural Assessment for Sustainable Kiosk Projects," in [43] G. M. Faheem Hussain, Rahul Tongia, "Technological
ICTD 2006 UC Berkeley, California, USA: IEEE, 2006. Interventions for BRAC's In-Service Teacher Training Program: A
[13] F. Hussain, "Community Broadcasting in South Asia: Hybrid Comparative Analysis," International Journal of Education and
Solutions for localized Information Access Points (A Bangladeshi Development (submitting for publication), 2008.
Case Study)," in 26th Scientific Conference on the International [44] "Radio Lumbini 96.8 MHz." vol. 2008, 2008.
Association for Media and Communication Research: Media and [45] B. Haiplik, "The BRAC Education Program Teacher Training and
Global Divides (IAMCR 2008) University of Stockholm, Development System Quality Improvement Initiative: Report #2,
Stockholm, Sweden, 2008. Monthly Refresher Training for BRAC Teachers," 2004.
[14] "CIA- The World Factbook -- Bangladesh." vol. 2008, 2008. [46] "Pallitathya Kendra." vol. 2008, 2008.
[15] "Media in Development: Linkages Between Poverty Reduction, [47] S. Rahman, "Internet in Bangladesh," Bangla ICT 2008.
Social Development and Diversified Media in Bangladesh," [48] "The Most Deadly 100 Natural Disasters of the 20th Century." vol.
Bangladesh Enterprize Institute, Dhaka 2006. 2008.
[16] "National Media Survey, 2002," 2002. [49] T. Köhler, M. Reitmaier, and G. J. Schulz, "Telecenters as
[17] "Bangladesh Telecommunication Regulatory Commission." vol. Instrument for Bridging the Digital Divide in Rural Areas," TU
2008, 2008. Dresden.
[18] "Many Voices, One World," Essex 1980. [50] "Community Radio Madanpokhara (CRM) Palpa." vol. 2008,
[19] C. Rodriguez, "From Alternative Media to Citizen's Media," in 2008.
Communication for Social Change, Anthology" Hostorical and [51] "Community Broadcasting Association of Australia ". vol. 2008,
Contemporary Readinngs, A. Gumucio-Dagron and T. Tufte, Eds.: 2008.
CFSC, 2001, pp. 778-786. [52] "Nepal Forum of Environmental Journalists." vol. 2008, 2008.
[20] A. E. Opubar, "If Community Media is the Answer, What is the [53] "Universal Service." vol. 2008: Federal Communications
Question?," in Communication for Social Change, Anthology" Commission, 2008.
Hostorical and Contemporary Readinngs, A. Gumucio-Dagron
and T. Tufte, Eds.: CFSC, 1999, pp. 778-786.
[21] A. Gumicio-Dagron, "Take Five: A Handful of Essentials for ICTs
in Development," in Communication for Social Change,
Anthology" Hostorical and Contemporary Readinngs, A.
Gumucio-Dagron and T. Tuft, Eds.: CFSC, 2003, pp. 976-988.
267
O ver the past few decades, most developing countries have in particular that, with the LLP, the Bolivian government
embarked on decentralization programs with the finally recognized traditional indigenous forms of organization
ambitious aims of: i) increasing government‘s efficiency and and transferred important resources to local governments. It
efficacy; ii) enhancing the quality of service delivery in such was, in short, a formidable attempt to shift power relations
areas as health and education; and iii) to enhance the within Bolivia‘s mestizo-dominated society, so that indigenous
responsiveness and poverty-focused of governments [1]-[3]. peoples and other marginalized groups would have the space
Most of these programs operate under the assumption that to voice their concerns, interests, and needs.
decentralized governments achieve more efficient results and This case study investigates the role ICTs can play in
are less corrupt since they are closer to the people, better able supporting such a decentralization process and in helping to
to understand the local dynamics of economic and social improve the well-being of indigenous peoples. In particular,
development and are thus more accountable for their programs the case study addresses the following questions: (i) do ICTs
to rural communities [4]. These programs are usually contribute to improving the efficiency and efficacy of local
embedded into broader policy reforms of the public sector and governments, thus strengthening their institutional capacity to
go frequently hand in hand with major privatization programs. provide basic services to local communities?; (ii) can ICTs
In Bolivia, the 1994 Law of Popular Participation (LPP) enhance the transparency of local governments and make them
signaled a dramatic shift in national policy from a centralized more responsible and accountable to the needs of local
communities?; (iii) was Enlared Municipal, the centralized
government-led program, able to achieve its main objectives,
and which constraints and limitations did it face in the
implementation of the program?; (iv) in which dimensions
Manuscript received September, 22, 2008. Bjӧrn-Sӧren Gigler is a PhD
Student at the Development Studies Institute (DESTIN), London School of
1
Economics and Political Science, Houghton Street, London WC2A 2AE, UK Bolivian Government in cooperation with the IMF and the World Bank,
(e-mail: b.gigler@lse.ac.uk). ―Economic Policy Framework Paper for 1997-99‖, June 1997
268
(i.e., political, organizational, and social) do ICTs have the of local governments [17]-[20]. Most of these authors have an
biggest influence on local governments and to what extent do optimistic view on ICTs and their role in development and
they enhance the human and social capabilities of the thus recommend that governments proactively embrace e-
program‘s participants?; and (v) did the project lead to any government applications in order reap the benefits from the
institutional changes in local governments, thus altering the use of ICTs within the public sector.
relationship between local government officials and More critical views of e-government highlight that in praxis
indigenous communities? the majority of e-government programs in developing
The case study investigates these central issues, based on countries have failed [21]-[23]. In fact, Heeks [24] ascertains
the empirical evidence from the government-led ―Enlared that the empirical evidence shows that 35% of e-government
Municipal‖ project—the principal ICT program in support of programs are total failures (the program was not implemented
local governments in Bolivia, coordinated by the Federation of or was immediately abandoned after its development), and
Municipal Associations of Bolivia (FAM). 50% are partial failures (major goals were not attained and/or
The overall objectives of the program were (i) to improve there were undesirable outcomes). He introduces the ‘reality-
the transparency and accountability of local governments; (ii) design‘ gap framework to analyze the underlying causes for
to enhance local governance structures and to improve the the numerous cases of failures and shows that the major
relationship between local government and indigenous reason for these failures is the common gap between the
communities; and (iii) to strengthen the institutional capacity current realities within public administrations and the design
and thus the performance of local governments in particular in of the e-government project. Heeks concludes that the larger
relations to the delivery of social services. Thus, the program‘s the design-reality gap is in the following seven dimensions
main focus was the political dimension, as its principal aim (information, technology, process, objectives and values,
was to improve local democratic processes and to strengthen staffing and skills, management systems and structures and
the popular participation of indigenous peoples in local other resources) the greater the risk for-government programs
governments. Nevertheless, the program also included in its to fail [25].
overall goals institutional and social aspects of local Madon demonstrates based on an interpretive set of case
development as evidenced through its third general objective. studies in Kerala India, that e-government programs have
Based on these objectives the project focused on the frequently neglected such critical factors as adequate
following specific activities: (i) implementing an ICT resources, skill-levels, values, beliefs and motivations and thus
capacity-building program for local government officials; (ii) poor people and communities were unable to derive real
developing and promoting the use of a national Portal on benefits out of such investments [26]. The author proposes a
issues related to municipal development www.Enlared.org.bo framework for evaluation the impact of e-governance projects
among local, regional, and national government officials; (iii) based on Sen‘s capability approach in order to be able to
assisting municipalities to develop and implementation their assess the developmental impact of these programs on
own municipal websites at the local level and (iv) helping peoples‘ quality of live. Madon concludes that a critical
formulate a national ICT strategy within the framework of the barrier for the successful implementation of e-government
ongoing UNDP-sponsored ETIC-initiative2. programs is that far-reaching back-end administrative reforms
need to be implemented in parallel for these programs to have
II. THE RELATIONSHIP BETWEEN E-GOVERNMENT, GOOD any significant positive effects on local governance [27].
GOVERNANCE AND DECENTRALIZATION The literature on the e-government and decentralization has
E-government is defined in the literature “as the use of stressed that the principal value of ICTs lays in both in
ICTs to improve the activities of public sector organizations‖ enhancing the participatory (good governance) and the
[10]. More specifically, the main objectives of e-government managerial (improved performance) dimensions of local
are (i) to make government more accessible, effective and government [28]-[30]. Lawrence Pratchett emphasizes that
accountable; (ii) improve the relationship between government ICT projects have the potential to contribute to support local
and citizens; (iii) strengthen the coordination and cooperation governments in the following three dimensions of their work:
within the public administration; and (iv) enhance government (i) enhance local democracy; (ii) promote public policy-
performance in the delivery of public services [11]-12]. making; and (iii) improve the quality of their service delivery
The proponents of e-government frequently argue that the [31]. Most scholars however highlight that local governments
application of ICTs in the public sector can play a catalytic use ICTs primarily as instruments to improve the delivery of
role for greater democracy, improved government services and neglect its uses to enhance participatory
performance and socio-economic development more generally mechanism conducive for good governance [32-35]. In fact,
[13]-[16] particular, the literature frequently points to the an empirical study of the ICT use of 270 local governments in
potential of e-government programs to promote good California highlights that most municipalities do not see ICTs
governance by enhancing the performance and accountability as an instrument for promoting good governance at the local
level, but as an additional communications tool—failing to
2
The Estrategia Nacional de las Tecnologías de Información y explore their value as change agents and their potential to alter
Comunicación (ETIC) is a UNDP-financed initiative which developed the fundamental relationship between government and citizen
through a participatory process including all sectors of the Bolivian society a
National Strategy for the Information Society for Bolivia (www.etic.org.bo) [36].
269
This case study draws on previous e-government studies by
Ciborra (2005); Ciborra and Navarra (2005); Madon (1993, ―Frequently the needs of local communities
2004), Walsham (1993) and Heeks [37]-[42] in its approach to are not considered, the projects are in fact
apply Pettigrew‘s (1985; 1998) methodology of a contextual designed on the desks of the Ministries or of
approach which emphasizes the importance of the economic, the international donors. A consequence of
social, political, and cultural context into which the this approach is that many projects are not
technologies are introduced [43]-[44]. These factors thus will sustainable at the local level and finish once
be essential for assessing the effects Enlared program had on the financing dries up. This program must be
local governments and the well-being of indigenous peoples. one of those unsustainable projects—at least
for my municipality there are currently no
III. THE ENLARED MUNICIPAL PROGRAM concrete benefits visible from this project‖.3
The Enlared Municipal project was implemented in two
distinct phases. The first phase of the project, from March The findings of the fieldwork showed that the first phase of
2001 to June 2004, was implemented by the International the project did not have any lasting development impact on
City/County Management Association (ICMA) and financed indigenous communities, so the research is focused instead on
by the U.S. Agency for International Development (USAID). the second phase of the project. Based on its experience with
ICMA is a U.S.-based international professional organization the first phase of the project, USAID decided to refocus the
for managers, administrators, and assistants in cities, towns, program and strengthened its partnership with Bolivia‘s
and counties and provides technical and managerial assistance, national association of local governments—the Federación de
training, and information resources to its members and the Asociaciones Municipales de Bolivia (FAM)4. During this
local government community (ICMA website). phase—from the beginning of September 2004 to the end of
During its first phase, the project developed an online November 2005—the FAM assumed responsibility for project
central Portal for municipal government coordination. The Federation organized a competitive bidding
(www.Enlared.org.bo), attempted to introduce a national process in mid-2004, which was won by the Swiss-funded
bidding platform for Bolivia, and provided technical support program Programa de Apoyo a la Democracia Municipal
to the Mancomunidad of the Chiquitania for the development (PADEM)5. The program focuses on increasing the capacity,
of an integrated Management Information System for the accessibility, and accountability of municipal governments in
fourteen municipalities belonging to the Mancumidad of the order to strengthen the participation of community-based
Chiquitania [45]. organizations and citizens in local governance. This program
However, the program‘s impact on local governments was was selected, in part, because of its seven-year experience in
limited by its ambitious and poorly focused objectives, its working with local governments on issues related to good
concentrated, top-down organization, and the perception governance, participation, and empowerment6.
among participants that the ICMA was a foreign and distant The new Enlared project team aimed to make the program
organization unconnected with the particular realities of the more responsive to specific local needs and integrated ICTs
Bolivian municipalities. During field visits to the Chiquitania into the ongoing local and regional processes of policy-
region in the Eastern lowlands in from April 20-25 and August making and citizen participation. There was a strong emphasis
2-7, 2005, it became apparent that the project did not have any on finding new ways of achieving the overall development
sustainable impact on local governments and communities. objectives of the project.
Interviews with local indigenous leaders, several government A. Socio-economic, Cultural and Technological Diversity
officials, mayors, and Juan Burgos, the ICT project of Municipalities
coordinator of APCOB—a local NGO promoting indigenous In order to set realistic objectives, the project team focused
people‘s development in the eastern lowlands—revealed that its technical assistance on fifteen municipalities and nine
the municipalities and communities were not aware of the regional municipal associations, covering all of Bolivia‘s nine
suggested Information Management System, and that in fact states.7 A major challenge for the project team, however, was
the fourteen communities of the Mancomunidad did not have
any connectivity to the Internet. The tremendous ―design-
reality‖ gap between project descriptions, consultant reports, 3
Interview with Justo Seoane, Mayor of Concepción, and Ex Vice-Minister of
and the realities on the ground was highlighted by the Indigenous Affairs, April 22, 2005.
4
FAM is the National Association of all the Bolivian municipalities with the
evidence through the results from my fieldwork. Most people exception of the major cities of La Paz, El Alto, Cochabamba and Santa Cruz.
interviewed were not aware that the project existed, even 5
PADEM stands for the Program for the support of the democratization of
though its first phase had been completed less than a year municipalities, which is being coordinated by the Swiss Non-governmental
before the fieldwork was carried out. In the municipality of organization Ayuda Obrera Suiza (AOS), which works towards the
strengthening of local democratic processes and the empowerment of poor
Concepcion, located in the Chuiqutania, approximately five communities throughout Latin America.
hours from the city of Santa Cruz, the newly elected mayor 6
Interview with Carlos Soria, coordinator of the Enlared project. August 25th
clearly expressed his frustration with the overly centralized 2005.
7
The 15 municipalities include: Cobija (Pando), Trinidad (Beni), Montero
planning and execution of the projects: and Comarapa (Santa Cruz); Sipe Sipe and Villa Tunari (Cochabamba),
Monteagudo and Tarabuco (Chuquisaca); El Puente (Tarija); Batallas and
270
that the fifteen municipalities differed extremely between municipal elections in December 2004) and their contact
rural, poor indigenous communities, such as Batallas in La information, as well as the most recent socio-economic and
Paz, or El Puente in Tarija, rural more mestizo dominated human development data on all 314 municipalities. In addition
municipalities such as Comarapa or Montero in Santa Cruz to the provision of content, the Portal provided more
and large urban municipalities such as Trinidad or Cobija— interactive features, such as a weekly forum with national or
the capitals of Beni and Pando, respectively. In addition to the local policy makers including relevant issues of interest such
tremendous difference in the socio-economic, political, and as the Constitutional Assembly or political and social
cultural context, these municipalities differed in their access to exclusion of the extreme poor due to their lack of appropriate
electricity and telecommunications services. To address and documentation.
understand the heterogeneity and the local context of the Concerning the capacity-building program, the project
fifteen participating municipalities described above, the trained 60 municipal staff in the use of ICTs (30 from the
program team initiated its activities in October 2004 with a regional associations and 30 from the 15 municipalities). The
detailed ICT needs assessment at the local level. staff from the municipalities participated in two one-week
Its main objective was to assess the differences in national training workshops, one organized in June in La Paz
awareness, ICT uses, and proficiency of use among local and the other one organized in late August in Cochabamba.
government officials in the fifteen participating municipalities. The content of the training focused on (i) basic computer
As part of this baseline study, the team carried out community skills, such as word processing or use of spreadsheets; (ii)
meetings, interviews, and structured surveys with mayors, introduction to the use of the Internet; (iii) the design and
council members, technical staff of the municipalities and with development of websites and; (iv) an introduction to ‗digital‘
representatives of local Territorial Grassroots Organizations journalism. The main purpose of these training workshops was
(TGOs), members of the Vigilance Committees (VCs), and to create a network of technical specialists in the
other community leaders. In total, 634 people—424 men and municipalities, who then could become the municipal focal
210 women—participated in this baseline study and provided points for Enlared.
valuable information on their information and
communications needs and their current use of ICTs [46]. IV. EVALUATION OF THE PROJECTS IMPACT
The assessment found that there was little awareness and
uses of the Internet in any of the rural municipalities, and that, Based on the above description of the project‘s main
even in the urban municipalities like Tupiza or Uyuni, local achievement or outputs, this section will focus on evaluating
governments were not using the Internet for work. In fact, out the impact the project had on the well-being of rural
of a total of 127 local government officials, only a quarter communities. The evaluation of the project impact is hereby
(24%) used the Internet for work; the overwhelming two- defined as the “systematic analysis of the lasting or significant
thirds majority did not use the Internet at all, and one-tenth of changes— positive or negative, intended or not—in people’s
participants did not even use computers. The results of the lives brought about by a given action or series of action [47].
needs assessment reveal profound differences in ICT readiness As such, the analysis aims to unpack the extent to which this
among the fifteen municipalities studied. project has made a difference in peoples‘ lives and has in fact
B. Main Activities met its overall development objectives by asking the
following: (i) Has the project contributed to strengthen the
Based on the baseline study, the project team focused its
managerial and technical capacity of local governments, thus
activities on the following four strategic areas: (i) technical
allowing them to play a more pro-active role in public-policy
redesign of the Portal; (ii) content development and online
making? (ii) Has the project supported the transparency of
services; (iii) capacity-building in the use of ICTs; and (iv)
municipalities and thus promoted participatory democratic
news desk and communications. Within these different areas
processes at the local level; and finally (iii) Did the project
the project was able to achieve the following specific results.
enhance the efficiency and efficacy of local government thus
First, the Portal www.enlared.org.bo was redesigned using
improving its delivery of services, such as health and
local technology and focusing on providing technical support
education to its community. This categorization is based on
to the Federation of Municipal Associations, as such helping 9
the theoretical framework developed by Lawrence Prachett
municipal associations and individual municipalities in the
described above in the case study [48]. The article applies the
development of their own websites. At the completion of the
alternative evaluation framework based on Sen‘s capability
second phase of the project, the Portal hosted 116 sites from
approach developed by the author in an earlier article to
these organizations. The Portal placed significant effort on
evaluate the impact of the e-government program on people‘s
providing targeted and timely content for policy makers, local
well-being [49].
government officials, donor agencies and the general public
interested in issue of municipal and local development. For
instance, the Portal provided for the first time a A. The program’s Overall Impact
comprehensive list of all newly elected mayors (from the The principal finding presented in this section is that
the program had a very limited positive overall impact on its
Coroico (La Paz); Machacamarca and Pampa Aullagas (Oruro) and Uyuni and participants‘ human and social well-being. The empirical
Tupiza (Postosi).
271
results are based on two user surveys carried out by the project
Knowlede about the EnlaRed Municipal Program
team and a sub-sample from the impact survey (n=22), carried
among local government officials
out as part of my research in July 2005. The first user survey (user survey: n=190)
(n=190), administrated as part of the ICT needs assessment in
October 2004, was designed to generate a baseline about the
N/A, 22, 12%
awareness and proficiency of ICT use in the fifteen
municipalities. The second user survey (n=186), was carried Yes, 70, 40%
indicated that the Internet would not have any impact at all on B.1 The Impact on Enhanced Transparency & Accountability
either their personal or community well-being. The empirical The first critical finding is that the Enlared Municipal
evidence from the impact survey thus emphasizes the program had only very limited effects on the transparency and
participants‘ overall negative perceptions of the Internet‘s vertical accountability of local governments and thus could
effects on their well-being. In contrast with the overall results, not significantly strengthen local democratic processes. The
Enlared Municipal participants were skeptical of the Internet‘s graphic below shows the effect the program had on the major
impact on their individual well-being and on the social four areas of accountability.
capabilities of local governments and their communities. In The disaggregated results illustrate that the program helped
fact, the data analysis demonstrates that the program failed to improve people‘s access to information about their own rights
enhance people‘s individual capabilities and help local and national government policies, but failed to enhance the
governments to improve their performance, accountability, transparency of local governments. In fact, less then one
and delivery of social services to the mostly rural quarter of the respondents (23.5%) believed that the Internet
communities. could make an important contribution in this area, and only
25% were convinced of its value for enhancing people‘s
B. Impact on the Political Dimension of Well-being access to information on government programs.
35.3 34.4
28.6 31.7
The analysis of survey results and interview responses in 30
23.5 25.1
23.1
the previous section shows that the Enlared Municipal project 20
19.6
effect on the political dimension. The data indicate that 35% Fig. 3: Continued lack of governmental transparency and
of the programs participants believed that the Internet had no accountability
effect at all on this dimension. The particular pessimism of
Enlared project participants becomes apparent when their The results point to a principal incongruity in the program:
responses are compared with those from the overall sample, while it had some positive impact through its Portal on
among whom 36% of respondents indicated that they were improving the access to information related to municipal
convinced that the Internet could play a significant role in this development at the national level, it did not improve
area and only 20% expressed serious doubts about its efficacy. information flows between municipalities and local
It is critical to note again that the Enlared Municipal program communities. The program had an overly abstract national
was the only ICT program that specifically prioritized the use agenda and improved the access to information for a small
of ICTs to enhance peoples‘ political well-being. These elite of national policy makers, international donors and the
273
media, however failed to enhance in any significant manner population and attempted to suppress the influence of
the dynamics between local governments and its citizens at the indigenous peoples in local politics. The president of the local
community-level. agricultural producers associations and a Quechua indigenous
This major finding was also confirmed by the semi- leader describes the exclusivity of local politics:
structured interviews with local government officials, NGO
representatives and community leaders in Batallas and ―For us it is very clear that the municipality
Comarapa. In an interview, the ―official mayor‖8 of Comarapa is quite biased against our communities. It is
clearly expressed his skepticism about the project and the role always the same issue. We can wait for
of the FAM in supporting local governments in general: hours in the municipality with our demands,
the mayor and the councilmember won‘t
―We have a very diverse population in listen to our concerns. They are too busy
Comarapa and the surrounding with their political games and to represent
communities; about half the population are the interest of the rich people of the city—
mestizos and the other half are indigenous we just don‘t have any voice in their
migrants from the highlands. Thus for us to decisions and we have no information at all
improve our communications with all the about what type of programs and activities
communities is very important, but I don‘t they are planning for this year.‘10
see how this program can help us to better
deal with this problems. The program has In short, local authorities lacked the will to improve access
developed a website—How does this help us to information for the general population and instead allowed
to improve our communications with more a small circle of mostly urban elites to dominate the local
remote indigenous communities? How do political decision-making process. The mostly rural
these communities benefit from this—the indigenous population was entirely excluded from any
majority of them do not even have access to political decision-making processes and was forced to choose
electricity, then how can they access a the path of street blockages and civil disobedience to gain a
website. I really can‘t see the value of this voice in the political process. Questions remain, however:
type of program for us.‖9 What was the role of Enlared Municipal project in such a
complicated socio-political environment? Did the program
This comment needs to be understood in the context of have any influence at all on the local political process? Why
Comarapa, a municipality that, in spite of relatively favorable did the program fail to promote in any way the popular
socio-economic indicators, was confronted with significant participation of indigenous communities in the decision
social tensions between its primarily rural indigenous making process of local government.
population and its urban mestizo population. In the statement One reason was the programs‘ inability to react flexibly to
above, the senior advisor stresses the challenge of improving the given socio-political context. Instead of working with local
his municipality‘s capacity to communicate with local government officials to make the existing municipal planning
communities in order to enhance the popular participation of a processes more transparent and to gradually enhance the
broad range of people in municipal planning processes. This is popular participation of indigenous communities in local
a critical shortcoming the implementation of the LPP has decision-making processes, the project continued to promote
confronted in most rural communities throughout Bolivia. its original concept to assist the municipality in developing a
While the law has succeeded in significantly enhancing the website which was supposed to enhance its transparency. This
participation of communities in the identification of municipal activity provided by the program clearly did not match the
development projects through broad-based consultations, it local socio-political context nor was it an appropriate
has faced in most municipalities serious limitations in terms of mechanism to address the existing democratic shortcomings in
improving the participation of local communities in municipal Comarapa. Thus, the program did not alter the prebendal
decision-making processes about critical issues of local culture 11 of local government. The same few powerful groups
development such as for instance the approval of planned continued to dominate the decision-making process within
public investment projects (Goudsmit and Blackburn, 2001). local government and the mostly indigenous population
The above quote highlights that the program failed to continued to be excluded and marginalized in political terms.
enhance the transparency of the municipalities and to improve In brief, the program failed (i) to improve the transparency and
its communications with more remote indigenous accountability of local government and (ii) to enhance
communities. In Comarapa a critical issue furthermore was peoples‘ political capabilities to better participate in the
that the local government sided frequently with the mestizo decision-making processes. The empirical evidence presented
8 10
The ―official mayor‖ is the most senior government official in local Interview with Julio Sanchez, president of the agricultural producer
government, who is in charge of the all processes of the public administration association, Comarapa, May 24, 2005.
11
(own translation). A prebendal local political culture is defined by local political practices
9 9
Interview with Jose Luis, official mayor of the municipality in that are characterized by patrimonial, clientelistic and caudillist features
Comarapa, July 15, 2005. (Blackburn, 2000)
274
above, based on participatory observations and unstructured of our own program. Our idea was that we
interviews with key policy-makers and local actors, clearly could share the connectivity costs with
demonstrates that the local stake-holders perceived the several institutions working in Comarapa
Enlared project to be distant from their own realities and and asked the mayor for his support.
vividly illustrates the program‘s inability to meaningfully Unfortunately, the mayor did not see any
improve local democratic processes. value in this program and instead promised
that the municipality will build its own a
B.2 Impact on Enhanced Local Governance
telecenter for the community. Now, two
Another critical aspect of enhanced political capabilities at years later, Comarapa still does not have the
the local government level is the improved capacity of promised telecenter and we have never
municipalities for public policy making in local communities. received an official response from the
Within the increasingly fragmented and disaggregated municipality on our proposal we submitted
structures of local communities many local governments face in writing to the mayor‘s office.‖12
considerable challenges in playing a central role in promoting
local governance structures that are conducive for economic This statement reflects the all too common lack of
and social development [50] The following analysis uses the coordination among local municipalities, local NGOs, and the
concept of ―good local governance,‖ developed by Judith central government. Despite the stated interest from the
Tendler, which refers to a healthy ―three-way dynamics municipal government and the Enlared project, after two years
among local government, civil society, and an active central Comarapa still did not have the telecenter it had been
government‖ [51]. promised. A key aspect in the failure of the Enlared Municipal
At the core of the concepts stands the notion that a good program to promote good governance at the local level was
local governance structure is essential for effective public the frequent discrepancy between the political mandate
policy making. Tendler stresses that coordination between designating local governments the executive agencies of
local governments, local civil society organizations, and development programs like the ICT program and their actual
communities, as well as coordination between local and institutional capacity and technical knowledge to actually
central government, are critical for good governance. Local implement such programs. The following comment by the
government is more of an effective facilitator of partnerships mayor of Comarpa about the planned telecenter illustrates this
between different local actors and stakeholders than a sole point:
implementer of development programs. ―For me it is very clear. We in the
A critical question this raises is what role ICTs can play in municipality are the executive branch of
such a new system of local governance? The case study will government at the local level. It is our
focus on the extent to which the Enlared project was able to mandate to plan and implement all programs
strengthen the capacity of local governments to assume the in our municipality. In the case of the
role of a ―facilitator‖ of local development. In brief, to what planned telecenter, I had several meetings
extent was the project successful in enhancing healthy local with the staff from ICO and explained to
governance in local communities? them that it is our role to coordinate this
The principal finding from both the municipalities of project. I still think that it is a good idea to
Comarapa and Batallas demonstrate that the Enlared project build a telecenter here in Comarapa,
was unable to play a central role in enhancing the capabilities particularly for your youth, however it needs
of local governments in this dimension, primarily because it to be clear that we need to oversee all
did not effectively promote a cultural change within local activities related to this project.‖13
governments to a more democratic and inclusive approach of
local public policy making. Instead, local government This statement clearly demonstrates that the local mayor
continued to manage local development programs in their considered it the prerogative of local government to
traditional authoritarian and bureaucratic styles. coordinate and oversee all development programs in his
Asked about the role the Enlared program played in local municipality. His defensive attitude illustrates the tensions
politics, Claudia Camacho from ICO cited the program‘s between his local administration and the NGO. While local
inability to facilitate a partnership between the local government officials frequently insist that the municipality
government and her NGOs in order to work together on the needs to spearhead all the development programs at the local
construction of a telecenter for Comarapa. level, the severe crisis of the state has led to a situation where
national and international NGOs have frequently filled the
―When we first heard about the Enlared institutional vacuum and are de facto implementing many
Municipal program and that Comarapa was programs, such as in health, education or infrastructure which
selected to participate in its pilot phase, we under normal circumstance would be carried out by the state.
were very excited and thought that we could
closely collaborate with the municipality on 12
Interview with Claudia Camacho, ICT program coordinator of ICO in
promoting the use of ICTs, a key objective Comarapa, May 23, 2005.
13
Interview with mayor of Comarapa, May 23, 2005.
275
The above-mentioned telecenter is a good example of a significant impact on enhancing the inter-agency coordination
situation in which the local government lacks the institutional between local government, civil society organizations, and
and technical capacity to implement the program and a non- local communities; neither did it enhance the relationship
governmental organization brings its own expertise and between local and central government.
resources to provide the population with a needed service.
Another reason for the poor local governance structure in
C. Impact on the Organizational Dimension
Comarapa is the role international donors frequently play in
creating dependencies in both the public sector and civil The overall empirical results from the impact survey seem
society organizations at the local level. The Enlared to indicate that the program had considerable positive effects
Municipal program, for instance, had defined its priorities and on the organizational dimension. The data show that more
operational methods through a highly donor-centric approach. than three-fourths of the program‘s participants (77%) rated
In fact the program did not allow local actors to come together the impact of the Internet on this dimension either as very
in order to adjust the program‘s planned activities to their significant (59%) or somewhat significant (18%). This section
particular social, political, and cultural context and to define a disaggregates these overall results in order to investigate the
way on how to collaborate on its implementation in extent to which the Enlared project did indeed lead to
Comarapa. In fact, the provision of internet services and ICT significant organizational changes in local governments and
training workshops would have lend itself to such a thus had a positive impact on their institutional capability.
collaborative approach, since many organizations within the This issue is critical for evaluating the overall impact of the
municipality had initially expressed their strong interest in program, since one of its core objectives was to enhance the
sharing the connectivity services with each other. An efficiency and efficacy of local governments by introducing
important consequence of the common donor-driven approach ICTs into local public administrations.
is that it provides important disincentives for local actors to First, the analysis uses the ―social capital index‖ in order to
work together and to coordinate their activities with each carry out a deeper analysis of the impact of the Internet on the
other, since each of the organizations is primarily accountable organizational dimension. The graphic below highlights that
to their international donor and not to local stakeholders. while a two-thirds majority of participants (69%) perceived
The example provided above—about the inability of the the Internet as critical to strengthening their horizontal
municipality to react to the proposal by ICO to share the costs networks with colleagues in other municipalities, only a
of the Internet connectivity—demonstrates that programs like minority of respondents (49%) believed that the Internet had a
the Enlared Municipal project focus solely on achieving their significant effect on strengthening the organizational
centrally defined development objectives rather than on capabilities of local government. The participants‘ skeptical
facilitating partnerships among the different local actors. In views about the program‘s impact on organizational structures
fact, the program worked exclusively together with the of local governments is also evidenced by the fact that more
municipality and entirely neglected to reach out to the other than one-third (35%) of respondents doubted that the Internet
organizations within the municipality. could play any role in this dimension. Furthermore, it is
In this sense the Enlared program is exemplary of many noteworthy that the large majority of respondents (64%), in
other development programs in Bolivia in that it was unable to spite of their skepticism about the Internet‘s ability to affect
facilitate a partnership between the different local stakeholders organizational change within government, had a very
that would have significantly enhanced the impact and favorable view of the positive effects of the Internet on local
sustainability of the program [52]. In the case of the Comarapa organizations outside of government, such community-based
municipality, the local government officials even preferred to organizations. This seems to indicate that the participants
delay the municipality‘s Internet access for many months believed that there are major barriers to introducing ICTs into
rather than develop a collaborative relationship with a local public administrations, while they consider this to be easier to
NGO. be achieved outside of government.
In sum, the Enlared program failed to enhance the local The more in-depth analysis of the data provides much richer
governance structure at the community level, since its program insight into the effects of the program on the organizational
had an overly abstract national agenda and its centralized dimension. In fact, the results demonstrate that the program‘s
approach prevented it from affecting in any significant manner organizational impact was limited to establishing a strong
institutional changes at the local level. Furthermore, the social network among its participants, instead of instituting
program did not facilitate effective partnerships between far-reaching organizational changes within local
different local actors and instead narrowly focused on trying to bureaucracies. Furthermore, the observed strengthening of
implement its pre-conceived objectives irrespective of the social capital and horizontal communication networks was the
local socio-political context. In this sense the program result of ad hoc formation of informal networks between the
represents the common shortcomings of similar donor-centric program‘s participants, rather than a sign of improved
programs in Bolivia. Finally, the program failed because it institutional linkages between the different local
lacked the necessary institutional credibility with local municipalities.
government officials based on the severe crisis of central
government institutions. In brief, the program had no
276
majority (62%) of the respondents who had an e-mail account
Enlared: Impact on Social Capital
used it only once every three days.
strengthens horizontal Moreover, the graphic below illustrates the resistance or
69 20 11
netw orks apathy of key decision-makers to these technologies. In fact,
the large majority of mayors and/or council members (82%)
strengthens local
49 16 35 did not consider the Internet important and thus did not use it
gov ernments
for their work. The data indicate that only among technical
staff did the majority (58.2%) use the Internet for their work.
strengthens local
organizations
64 18 18 Technical staff also demonstrated higher rates of e-mail use:
54% of them had personal e-mail accounts, compared to just
0 20 40
Percentage
60 80 100 18% of mayors and council members.
A lot Somew hat None/NA The very low local appropriation of ICTs in local
government can be confirmed by comparing the results from
Fig. 4: ICT—a powerful tool for horizontal networking
the baseline study from October 2004 with those from the user
The above data also have to be seen in the context of the survey from September 2005 described above.
significant barriers the program participants faced in
developing and promoting the use of information systems Internet Use for work by local government officials
within local bureaucracies, because of local government (user survey: n=190)
60
41.8
government officials had a personal e-mail account (such as 50
40 32.7
―.yahoo‖) and none of the participants had an institutional e- 30 10.7
25.5
14.818.3 15.7
20.5
20
mail account (―.gov.bo‖). It is also noteworthy that the large 10 7.1
0
Mayor/Council Technical Member of Total
member specialist oversight
14 committee
Both of these user surveys focused exclusively on issues related to the
awareness, readiness and use of ICTs in local governments (Enlared project Position
A lot Somew hat None/NA
files).
15
The user survey from September 2004 was based on a sample of 190 Fig. 6: Government officials‘ skepticism about Internet use
respondents and the second survey from September 2005 on a sample of 186
participants.
277
In fact, 68% of the respondents from the 2005 survey
indicated that they had been working for less than one year in
the municipality and just eight of the training participants, out Enlared: Impact on Social Services
to reach one of its main objectives—to promote the use of A lot Somewhat None
Abstract—Blood is a scarce resource critical to the developed as a collaboration between CDC, the Georgia
management of a variety of life-threatening medical conditions. Institute of Technology and the participating countries. The
It can also be a medium for the transmission of infections, tool expands on the Excel-based system to facilitate the
including HIV. Monitoring the quality and quantity of available efficient and accurate collection and analysis of more than 80
blood, essential to utilizing it as effectively as possible, has been a
indicators from 14 countries.
challenge in many developing countries. This paper describes the
design and implementation of a web-based tool to monitor the
collection, screening and distribution of blood in developing
II. RELATED WORK
countries. This project was conducted under the auspices of the
US President's Emergency Plan for AIDS Relief (PEPFAR),
Several ICT projects have focused on improving the health
which funds blood safety projects in 14 countries in Africa and
care system in developing countries. These projects have had
the Caribbean. We report results from a usability study,
formulate relevant design principles and discuss prospects for several different themes including development of web-based
long-term sustainability. tools for information tracking, implementations of electronic
medical record systems and telemedicine systems to support
Index Terms—developing nations, health, ICT and education of clinicians and referral of difficult cases for
development, Internet, medical services further review.
In [2] a web based information system supported medical
I. INTRODUCTION record keeping and tracking of HIV treatment in Haiti. While
this work is similar to our work in that there is a centralized
Blood is a scarce resource worldwide. But the scarcity is database and an online interface for certified users to enter and
particularly acute in the developing world. In 2008, the World track data, the population and scale of users is quite different.
Health Organization estimated that although 80% of the In our work we focus on users in 14 different countries
world's population lives in developing countries, these worldwide with varying levels of internet access and technical
countries only collect 45% of the global blood supply [1]. support. We also present evaluation of the system by
Effectively managing this scarcity—deciding who gets blood prospective users which indicated that the system was easy to
and when—is a challenge for health care systems throughout adapt to and use.
the developing world. Blood is also a highly effective medium
for the transmission of blood borne illnesses, including HIV, Blaya et al in [3] created a web based tool, e-Chasqui, to
hepatitis and syphilis. The WHO estimates that between improve the timeliness and quality of reporting tuberculosis
95-10% of all HIV infections worldwide may be linked to laboratory data in Peru. This national system is able to provide
unsafe blood transfusions. In 2004, the US President's access to a hierarchy of users in health centers, regional
Emergency Plan for AIDS Relief (PEPFAR), identified blood laboratories and the national laboratory to test results. These
safety as a key element in its comprehensive HIV prevention results are then used to prescribe treatment regimens for
strategy. Since then, PEPFAR has committed $192 million in patients at the local level.
funding to National Blood Transfusion Services (NBTS) in 14 In analyzing the use of web-based tools, Marquez found
countries (Botswana, Cote d'Ivoire, Ethiopia, Guyana, Haiti, that such tools provide notable enhancements in quality
Kenya, Mozambique, Namibia, Nigeria, Rwanda, South improvements in health care [4]. The web-based tool
Africa, Tanzania, Uganda, and Zambia). To track, monitor and described in this paper is an example of such tools.
evaluate the impact of this funding, the Centers for Disease
Control and Prevention (CDC) worked with global blood
safety experts to develop a set of program indicators and a III. PROBLEM AND BACKGROUND
Microsoft Excel-based data collection tool. In this paper, we
describe a web-based data collection and management tool Monitoring the development of a national blood service
requires data from a diverse set of sources. The WHO, in its
Aide Memoire for National Blood Programmes [5], advises
S. Thomas, A. Osuntogun, and S. Vempala are with the Georgia Institute of
Technology, Atlanta, GA 30332, supported in part by a Raytheon Fellowship developing countries to build national blood strategies around
and the NSF. four "key areas", including: The establishment of a nationally-
J. Pitman is with the Centers for Disease Control and Prevention, Global coordinated blood transfusion service; the collection of blood
AIDS Program, Atlanta, GA 30333 only from voluntary non-remunerated blood donors; universal
B. Mulenga is with the Zambian National Blood Transfusion Service, testing of all donated blood, including screening for
Lusaka, Zambia
transfusion-transmissible infections, blood grouping and
281
compatibility testing; and the reduction of unnecessary These challenges increased the possibility of inaccurate, or
transfusions. These broad recommendations can be out-of-date, reporting. To illustrate, in several countries,
further categorized by technical area: regional or provincial blood service offices would would fill
out and email the Excel-based tool to the national manager, or
1) Policy information is needed to track progress in
save the report on a disk or flash drive and send the disk back
standardizing procedures (e.g., laboratory testing)
to the national manager by road. The national office would
and establishing a blood service's legal authority over then compile all the results from the different provinces and
the national blood supply,
send a final report to the CDC. Each update from a regional
2) Social and demographic information is needed to identify, center required re-entry and re-aggregation of the data, leading
recruit and retain blood donors with low behavioral risk to multiple versions and increasing the chance of error.
profiles for infection with HIV and other transfusion- To address the weaknesses of the Excel-based tool, the team
transmissible infections (TTI),
developed an Internet-hosted database accessible with a
3) Scientific data are necessary to track the prevalence of standard Web browser. Storing the data in a Web-based
infectious markers in donated blood and to monitor relational database provides several key benefits compared to
laboratory performance, ad hoc circulation of a spreadsheet file:
4) Administrative data are needed to measure human 1) There is a reliable, authoritative source of reported data.
resource and logistical capacity, and, in many cases, to (No more wondering which file among several is most
manage large grants from bilateral or multilateral donors, accurate.)
and
2) Data is continuously available throughout the reporting
5) Medical information is critical to ensure that blood is used period allowing immediate modifications and corrections.
appropriately and to monitor patient outcomes. (No files "in transit" or lost via email.)
A blood service could potentially collect data on hundreds 3) The system can support automated, real-time aggregation
of variables. of reported data from multiple sources with less risk of
transcription errors.
With PEPFAR support, CDC and its international partners
developed a limited, but comprehensive, set of programmatic 4) Full change tracking with audit logs is available.
indicators. These indicators are broken down into a subset of
Additional benefits derive from the system's use of a
variables which cover the five technical categories described standard web browser interface:
above.
5) Access is ubiquitous and available on all modern
Collecting these data from 14 developing countries with computing platforms.
different data systems was the principal challenge associated
with this project. To begin to address this problem, CDC 6) End users are familiar with web browsers (e.g., Internet
developed a standardized data collection tool using Microsoft Explorer, Firefox, Safari).
Excel. This format was selected for its ease of use (data
7) Updates and enhancements can be easily managed and
managers in all of the PEPFAR countries were familiar with
deployed.
Excel) and its ability to be transferred via e-mail over low
bandwidth internet connections. A version of this tool, with The specific design principles for the system follow directly
pre-programmed formulae and skip patterns, was field tested from its requirements:
in 2007. While countries reported the tool was easy to use and
1) [Network compatibility] The system must support access
met the programs' PEPFAR-related reporting needs, the pilot
through low bandwidth, dial-up connections, as high
phase also revealed several weaknesses in the use of a
bandwidth connections cannot be assumed for all users.
spreadsheet-based tool, including:
2) [Intuitive interface] The system must provide a user
1) Difficulty in tracking versions of multiple files in
experience appropriate for blood safety staff who may be
circulation as new data were entered and re-saved by
familiar with the use of standard office software (e.g.,
different users,
web browsers, spreadsheets, word processing), but who
2) Countries' inability to quickly modify, clean or correct a lack high-level IT training.
data set after a file was submitted to CDC,
3) [Security] The system must provide appropriate security
3) The potential for transcription and other errors as multiple and access control for aggregated health information.
versions of the spreadsheet were merged (e.g., from
4) [Flexible and adaptable] the system must be easy to
multiple regional centers) prior to submission to CDC.
manage, adapt, and expand. For example, it must be easy
4) PEPFAR requires quarterly reports from each country, but for administrators (many of whom will not be trained IT
blood is collected, tested and utilized continuously. This managers) to add new countries, regions or users, and
meant the Excel-based tool could sit idle for weeks or ensure that data entered by these new sources are properly
months at a time. aggregated. (Based on previous experience, the CDC did
not feel the need to support languages other than English
282
in the initial deployment. The site was designed, however, users are able to view as much or as little of the information as
to accommodate localization to alternate languages should they wish, all on one page, as Figure 1 highlights.
that be desirable in the future.)
The tool was designed via collaboration between CDC,
Georgia Tech, and NBTS data managers in the 14 PEPFAR
countries. The Zambian National Blood Transfusion Service
(ZNBTS) agreed to host the pilot phase described in this
paper. This partnership was made possible through a program
at Georgia Tech which seeks to leverage student and faculty
expertise to address social and development problems in low-
income settings. Funding for logistics during the pilot
project was provided, in part, by PEPFAR and the ZNBTS.
The Georgia Tech team's time and expertise were donated
gratis as part of a course.
After multiple rounds of testing, user feedback and usability
studies (described later), it was evident that the web tool was
highly welcomed by NBTS data managers in Zambia and
elsewhere. Their feedback led to changes both in the user Fig. 1. Selective disclosure allows the user to show only those sections of the
interface and in the architecture of the system. Further, data page that are relevant for the task at hand.
managers expressed a strong desire to use the system not just
for reporting but also as a decision informing tool. We Although it is theoretically possible to implement selective
redesigned the system to allow individual countries and disclosure strictly through PHP scripts executing on the web
individual centers within countries to choose the frequency server, most users would find the resulting lag between their
of data entry (to match the frequency at which data was action and the interface's response to be intolerable. For a
actually being aggregated physically), and to choose the type suitably responsive interface, selective disclosure is
of reports and views of historic data that they desired. This implemented solely on the web browser client, using
was all done while maintaining a uniform quarterly PEPFAR javascript to show or hide the appropriate interface elements.
report for all participating countries. However, not all web browsers can fully support javascript,
and some users may choose to disable javascript even when
their browser supports it. To accommodate these users,
IV. INITIAL DESIGN unobtrusive enhancement in the initial design adds selective
disclosure as an optional, additional layer in the interface. The
Initial prototypes for the system used common web prototype accomplishes this by first creating a page with all
application techniques and technologies. MySQL database interface elements available and disclosed. Figure 2 shows a
tables were defined to collect and store reported information. screen capture for such a page.
Scripts written in PHP provided a web-based front end to the
database, allowing users to enter data and view collected
information. The web-based user interface also provided a
means for system administration, including defining regions
and countries, managing users, and supporting data backups.
A key aspect of the user interface design was its reliance on
unobtrusive enhancement[6]. This strategy creates a layered
user interface, where successive layers provide more
functionality, but are only invoked if the user (in particular, the
user's web browser) supports the required technologies. For
example, many of the input pages consist of a number of
individual data elements. To make it easy for the user to focus
on the information in manageable pieces, the interface
provides selective disclosure--information not relevant to the
immediate task is hidden or de-emphasized so that it does not
distract the user. This approach has significant benefits to the
user compared to alternative strategies (such as making each Fig. 2. When the page loads initially, all sections are fully shown. If the
section a separate page or screen), as users are able to decide browser does not support javascript, the page remains displayed in this form.
for themselves what information is relevant to the task. For The interface then attempts to execute javascript to add
example, a user may wish to refer to other inputs when selective disclosure. If the web browser does not support
supplying a particular value. If the other inputs exist on javascript (or if the user has disabled that option), then the
separate pages, then the user would be forced to jump back page remains as in Figure 2. Although there is no selective
and forth between multiple pages. With selective disclosure, disclosure, the interface remains fully functional. When
283
javascript is available, however, selective disclosure national blood center in Lusaka, Zambia as well as in Kitwe,
immediately activates and the new functionality is available to the center for the Copperbelt region. While in Zambia, the
the user. Figure 3 shows a screen capture for the same page team also conducted the study with data managers in Dar-es-
after the javascript executes. Note that in most cases the Salaam, Tanzania, over the telephone. We next describe each
javascript execution is fast enough that the user will never of the three test sites in more detail.
even notice the initial interface of Figure 2. For most users, the
Zambia National Center, Lusaka. The evaluation
only interface of which they will be aware is the interface of
conducted at the ZNBTS consisted of 6 participants in total,
Figure 3.
some were assigned tasks as country managers while others
were assigned tasks for provincial managers. The participants
consisted of a physician, biomedical scientists, financial
manager, information technology manager and data managers.
The participants were given the web address to the web tool
and were asked to complete the tasks. Two observers from
Georgia Tech were available to administer the evaluation and
provide help only if absolutely necessary. None of the
participants had prior experience with the website so they
were allowed to navigate around the website to determine how
to use the features. Internet connectivity available was via
satellite and evaluators used laptops with wireless
connections.
Zambia Copperbelt Center, Kitwe. In Kitwe, there was
dial-up internet connectivity, although it had not been
functioning for some months. After several attempts during the
first hour of our visit, the problem was discovered and fixed,
Fig. 3. If javascript is able to execute in the user's browser, it activates
selective disclosure and enables the user to hide or reveal sections making it possible for the evaluation to happen. There were 3
independently. participants consisting of a physician and two data managers
who were all familiar with the Excel-based tool. The dial-up
This design illustrates the benefits of unobtrusive
enhancement. Users without the more advanced technology connection had to be reset twice during the entire evaluation.
(in this case, browser support for javascript) still have access Tanzania National Center, Dar-es-Salaam. The usability
to a serviceable interface. The majority of users who do have study for the Tanzania data managers was conducted over the
the necessary technology, however, receive a more advanced phone. The observers called in from Lusaka, Zambia and went
user interface. through the tasks on the task list with each participant
individually; two participants had a copy of the task list by
email and were asked to indicate by speaking aloud how they
V. EVALUATION were navigating the system. Fiber-based internet connectivity
was available in Tanzania which made for significantly faster
A. Methodology
navigation.
Prior to the implementation of the web-based system, there
B. Study Design and Analysis
were several meetings between the CDC and the team from
Georgia Tech. These meetings helped determine the The usability study consisted of 3 components: a pre-trial
specifications of the system including the design of the user- survey, an observed trial where data managers were asked to
interface, the preferred programming language and hosting perform a set of tasks and a post-trial survey.
and management of the website.
Pre-trial survey. A survey in form of a questionnaire was
To evaluate the effectiveness of the tool, we conducted a conducted prior to the trial of the web-based tool to assess the
series of studies involving the end-users. We selected a subset expectations of the levels of ease or difficulty of using the
of the PEPFAR countries to use for evaluation. All the data web-based tool. The survey was conducted via the web by
managers in the National Blood Transfusion System in the using a commercial web-survey tool that provided information
PEPFAR countries are currently expected to use the Excel about the date the survey was taken, the IP Address and
based reporting system and therefore already had some aggregate results from all participants etc. By conducting a
understanding of what kinds of data would be requested. web-survey, we were also able to assess the response time and
the availability of internet access. There were 9 participants in
Zambia was chosen as the lead country for deployment and
the pre-trial survey, all from the Zambia NBTS; 7 from the
evaluation due to the progress in using the Excel based tool
Lusaka and 2 from Kitwe. The pre-trial survey link was sent
and their overall blood safety effort. The team from Georgia
out by July 21, 2008 and all surveys except one were taken
Tech spent approximately two weeks, July 27 to August 8,
between July 22 and July 27. 2008, prior to observers’ arrival
2008, in Zambia studying the local setting, usability of the
in Zambia. All the participants had some prior knowledge of
system, network characteristics and gathering qualitative
the Excel-based tool. The questions in the survey are shown in
feedback. The usability study component was conducted in the
284
Table I. Each participant was asked to rate the ease of each results indicate that user expectations of the difficulty of using
task on a Likert scale of 1-5 when appropriate, with 1=very the web based tool were above the actual difficulty
easy, 2=easy, 3=neither easy nor difficult, 4=difficult and encountered while using the tool. There was a reasonable level
5=very difficult. Questions 4 and 5 were given out to aid in of ease while working with the web-based tool. Participants
qualitative analysis. were also able to find some typographical errors and provide
concrete suggestions for changes.
Table I. Pre-trial Questions
The overall survey results indicate that the system was not
1 You have forgotten your password and need to log into very difficult to adapt to and use. Figure 4 shows the average
the web-based system. How hard or easy do you expect difficulty levels pre- and post-trial for each site. From Figure
to find this task? 4, it is apparent that the difficulty experienced by users in
2a You are only responsible for reporting some of the data, Zambia was significantly lower than what they anticipated.
while other users report the remaining data. You need to
enter the data for which you are responsible without Pretest Tanzania Zambia
disturbing the other data. How hard or easy do you
expect to find this task? 3.0
2b At the close of a deadline, you have only partial
information to report. How hard or easy do you expect to 2.4
find this task?
3 Previously entered data is wrong, and you must provide
1.8
the correct information. How hard or easy do you expect
to find this task?
4 What are some of the advantages you expect from a 1.2
web-based system compared to the old spreadsheet tool?
5 What are some of the disadvantages you expect from a 0.6
web-based system compared to the old spreadsheet tool?
0
Trial and post-trial survey. Discount usability techniques Task 1 Task 2 Task 3
were used to evaluate the usability of the system and the
Fig 4. Pre- and post-evaluation comparisons show tasks were as or easier than
interfaces. This evaluation provided a means of obtaining expected.
insights about usability problems and possible solutions. A
task list was used to obtain quantitative and qualitative The results from Tanzania, show less ease in using the
measurements of the ease of use of the tool. When participants system; this effect may be attributed to the additional
completed all tasks, they were asked to rate the ease of use of difficulty of having to evaluate the system while on the phone
the system on the same rating scale as in the pre-survey. This with the observers. Also, the evaluators may have felt some
evaluation occurred immediately after each participant pressure to accomplish the tasks quickly because the observers
completed all tasks, during the observers visit to Zambia. The were waiting for them to complete each task while holding on
list of tasks to be completed by each evaluator is shown below the phone line.
in Table II. In table III, the average difficulty for each task is given.
Table II: Evaluation Task List Overall, the pre-trial results are higher than the post-evaluation
results, this supports the notion that the interface was well-
1 Your username is country/province@blood-safety.org designed, to the extent that such a self-evaluation can assert.
and the password is the word "password''. Use the The "Export" task, not mentioned in the pre-trial survey was
system to change the password to "country/province''. perceived to be more difficult than the other tasks; it involved
2a Enter the regional information for the current quarter. transferring data from the system to the local computer in a
standard format (XML) which may have been unfamiliar to
2b You are entering data for the current quarter for your many evaluators.
country/province. You only have data for all the non-
Table III. Mean Scores for Selected Tasks
numeric fields. Enter and save this part of the data.
3 There is an error in the HIV prevalence field. Please
Task Pre TZ ZM ZM-NTL ZM-CB
correct it.
4 Export the country/province data to an XML file and 1 2.56 2 1 1 1
view the data using Excel.
2 2.22 2.5 1.17 1.33 1
C. Results
3 2 2 1.29 1.67 1
Analysis of the results obtained from this evaluation, shown
below in table I and figure 4, revealed that most evaluators Export - 3 2.5 3 2
rated most of the tasks between scale level 1 and 2. These
285
D. Qualitative findings of the network infrastructure1 . Because the same infrastructure
was used for all Internet access and, thus, lengthy page load
Utility and maintenance. Respondents to the pre-trial
times were common for nearly all web sites, participants were
survey felt that the web-tool would:
somewhat conditioned to expect this delay. Nonetheless, their
• Result in more timely reporting. dissatisfaction levels suggested that it would be worthwhile to
• Be quicker and less tedious. investigate opportunities to reduce the page delay.
• Be more cost-effective. The revised design significantly reduces the perceived
• Reduce the number of errors. initial loading time through the use of Asynchronous
• Be more user-friendly than a spreadsheet. Javascript and XML (AJAX) technology[7]. With AJAX, a
• Permit easier correction of data entry errors. web page can continue to interact with the server after it is
initially loaded without requiring explicit action on the part of
• Make it easy to share data between users.
the user. By using AJAX, the initial contents for each page can
All respondents were strongly of the opinion that a web- be substantially reduced so that the page loads more quickly.
based tool would be considerably superior to a spreadsheet- While the user views or interacts with the page, it continues to
based report. retrieve additional information from the server. By combining
AJAX functionality with selective disclosure, the revised
Security. On one hand, some users felt data security is a
design provides quicker initial page loading in a manner that is
concern since the data is no longer locally stored and
seamless for users.
vulnerable to malicious attacks; on the other hand, keeping
data on a server insures against data loss in the case of Figures 5 and 6 illustrate the steps involved in loading a
problems with local computers. In any case, it was clear that page using the approach of the revised design. The server first
users were concerned about how secure the data would be. supplies an initial web page that is almost completely empty.
However, the individual sections on the page are marked as
Functionality. While the original purpose of the tool was to
hidden. Even though they contain no content, users do not
allow for data aggregation for the purpose of generating
perceive a problem because the sections are hidden.
PEPFAR-mandated reports, data managers were quick to see
that having a real-time aggregate picture would greatly
enhance their ability to manage blood collection and
utilization. They asked for data analysis, such as historical
trends and comparisons of data from different regions to be
included in the functionality of the system.
Network constraints. Given the state of network
infrastructure in Zambia and other participating countries, the
potential pitfall of relying heavily on the internet was noted by
several data managers. Would pages load too slowly? Would
the system be able to handle connectivity going up and down?
While Dar-es-Salaam has a fiber-based internet connection,
there is no fiber connecting Zambia to the rest of the world.
Thus, all internet traffic goes via satellite, imposing a strong
constraint on bandwidth, and perhaps even more significantly,
on latency. Indeed we did not notice much difference in
response times between Kitwe, with a dial-up connection, and
Lusaka, with wireless connectivity and a dedicated satellite Fig. 5. The web server initially supplies only the skeletal framework for the
connection. full page; however, since the framework contains minimal content the network
can transfer it very rapidly and present the user with the appearance of a full
interface.
IV. REVISED DESIGN As soon as the initial page has loaded, it begins using AJAX
requests to retrieve the actual content for each section. During
Usability testing of the initial prototype indicated that, in
this time, however, the user is still viewing the initial page and
general, the application and its interface were suitably easy to deciding how to proceed. In many cases, by the time users
use. None of the tasks were rated as difficult or very difficult take their first action by selecting to show one of the sections
by any user. Comments made by participants during the study,
on the page, the content will have been retrieved from the
however, indicated that one of the most unsatisfying aspects of server and can be displayed immediately. When that is not the
the system was the time required for the initial loading of each
case, a progress bar tells the users that the interface is acting
web page. The main cause of the lengthy delay was the quality on their requests.
1While on site in Zambia, the developers were able to make measurements of network bandwidth and latency from several locations over 24-
hour periods. The tests measured performance from a local computer in Zambia to the prototype web site operating by a commercial web
hosting company in New York. The tests showed available bandwidth ranging from 12 to 57 kb/s to New York and 88 to 116 kb/s from New
York. Mean latency ranged from 5.1 s to 6.8 s.
286
use. Because the database captures all inputs to the system as
new data insertions (rather than updates to existing data), there
is no need to lock database access to avoid concurrent updates.
Such updates are problematic for any Internet-based
application, and would be especially challenging in an
environment with sporadic connectivity. The system's
approach resolves conflicts during report generation rather
than database modification. These features ensure that the
system meets the needs of both the sponsoring agency and the
local health officials.
V. DEPLOYMENT TIMELINE
non-AJAX design
VI. DISCUSSION
AJAX design Overall survey results indicate that a web-based system was
highly desirable and the system developed here was easy to
adapt to and use. This finding is in line with other experiences
with web-based tools for health projects in the developing
time
world[8]. A very significant observation from our survey is the
t1 t2 t3
need to incorporate local users in the preliminary design
initial page full page is hidden elements
framework can loaded; user are transferred; phases of a project. The original goal of the system was to
be displayed; perceives page page loading is serve as a data collection tool for CDC data managers in
user perceives loading as fully complete Atlanta; while conducting surveys, we discovered a gap in the
page loading as complete
complete
stakeholder perspectives as the local users perspective had not
previously been incorporated in the design. This highlights the
Fig. 7. Although the AJAX design requires longer to load because of extra necessity for human-driven design and research practices as
content, it allows the web browser to present a functioning page sooner than noted by Brand & Schwittay [9]. Local data managers stressed
the non-AJAX design; to the user, this difference gives the perception of faster
loading to the AJAX design. their preference for a tool that would also help in blood
management by analyzing and tracking recorded data. The
One consequence of the revised design is that it requires final design incorporates perspectives of both stakeholders; the
javascript. There is no way to make the page fully functional CDC data mangers and local users. This reflects several
when the user's browser does not have javascript support, so dimensions of Information and Communications Technology
unobtrusive enhancement is not possible. Based on for Development (ICT4D) as outlined in work by Tongia et al.
interactions with prospective users, however, it was [10] by addressing the original goals of the CDC’s Excel-
determined that improving page loading times was more based data entry system while also meeting the needs of the
desirable than supporting non-javascript capable browsers. In local users in the developing countries.
the future, should support for non-javascript browsers be
necessary, a separate user interface could be developed. We highlight two aspects of our design that give us
confidence in its long-term sustainability and might serve as a
Another key element of the revised design directly model for similar projects in the future.
incorporated requests from the blood safety staff in the remote
locations. In many cases, local staff wished to collect and track Data flow constraint. The data entry requirements of the
information in addition to that required by the CDC. Also, system should match the data collection, i.e., if the system
reporting periods were sometimes more frequent than the expects too much data, it will be hard to maintain and impose
quarterly reporting required by the CDC. By supporting an overhead on data collection. In technologically advanced
collection of local information at locally-defined reporting countries, there is often an abundance of easily available data
periods, the system provides a much greater incentive for its and thus it is possible to deploy systems that use vast amounts
287
of data collected rapidly. For a system to be sustainable, support HIV treatment in rural Haiti,” BMJ (Clinical Research
however, its data requirements should match the local data Ed.), vol. 329, Nov. 2004, pp. 1142-6.
flow, even if this necessitates a less complex system.
[3] J. A. Blaya, S. S. Shin, M. J. Yagui, G. Yale, C. Z.
Flexible design. The second most important aspect of our Suarez, L. L. Asencios, J. P. Cegielski, and H. S. Fraser. “A
system design, closely related to the first, is its flexibility, both web-based laboratory information system to improve quality
in data entry and analysis. Web-based systems are, in general, of care of tuberculosis patients in Peru: functional
designed to minimize the constraints inherent with variable requirements, implementation and usage statistics.” BMC Med
rates of data entry. As we observed with the spreadsheet based Inform Decis Mak. 2007;7:33.
data collection system, the potential for errors seemed to rise
[4] Lani Marquez. Are Web-Based Platforms to Support
as the numbers of users, and the numbers of copies in
Quality Improvement Feasible in Developing Countries?
circulation, increased. The web-based tool is less vulnerable to
Experiences from Ecuador, Honduras, Nicaragua, and
these errors: delays between entries will not have as great an
Rwanda, 2007. http://www.qaproject.org/news/ISQUA07/
impact on the overall quality of submissions. Because of this,
ISQua2007%20Marquez%20handout.pdf.
users will be more confident in the database, and more likely
to use it as a decision-informing tool. Our design goes further [5] World Health Organization, "Blood Transfusion Safety",
in that data managers can choose the frequency of entering 2004. http://www.who.int/bloodsafety.
data to match the frequency at which they receive data. For
[6] Jeremy Keith, Bulletproof AJAX. New Riders Press.
example, in Lusaka we observed that blood collection data
February 2007.
arrived weekly (by fax) from each of the other centers. On the
other hand, blood utilization data was not obtained so [7] Jesse James Garrett. "Ajax: A New Approach to Web
regularly and a month-long window was more realistic. The Applications". 2005. http://www.adaptivepath.com/ideas/
system allows users to choose views of the data for analysis essays/archives/000385.php.
and management. The views of interest depend on the local
context; e.g. in Zambia blood types have roughly the same [8] Jennifer Manne and Walter Curioso (2007).
“Technology and health care in the developing world.” Student
distribution across the country, but the incidence of malaria
BMJ, December 2007. http://student.bmj.com/issues/07/12/
(and thus the need for blood) varies by both time of the year
life/436.php
and region. Therefore, reports that show geographic and
seasonal trends in demand for blood would be more relevant [9] P Brand, A Schwittay. The Missing Piece: Human-
than reports that indicate blood type distributions. Driven Design and Research in ICT and Development. In
Proceedings of the International Conference on Information
Blood safety monitoring is an important special case of the
more general problem of monitoring and tracking large data and Communications Technologies and Development, May
2006.
sets which are updated frequently. The latter has many
possible applications in health, emergency relief, water and [10] R Tongia, E Subrahmanian. Information and
other basic resources, etc. In all these cases a web-based Communications Technology for Development (ICT4D) – A
distributed monitoring and aggregation solution seems worthy Design Challenge? In Proceedings of the International
of exploration, in spite of network constraints. We believe that Conference on Information and Communications Technologies
the insights gained from our development of a blood safety and Development, 2006.
monitoring tool could be applied more generally.
ACKNOWLEDGMENT
REFERENCES
I NFORMATION and Communication Technology (ICT) has preparedness and citizen e-readiness to make the whole effort
pervasively intruded into our everyday life. Perhaps, the successful and beneficial to the society. To achieve this, there
most recent picture of this pervasiveness is demonstrated by are many essential components, one of which is the
the aggressive way in which most governments are information technology (IT) literacy among the masses. Apart
automating their service deliveries and citizen interfaces from providing ICT infrastructure, IT literacy is equally
through what is popularly known as e-government. important to alleviate what is popularly known as the ‘digital
divide’.
The Sultanate of Oman, like all its neighbours, is pursuing a
national e-government initiative since the year 2002, along Information technology is now taught in all schools in Oman
with a wider Digital Oman strategy [1]. The country’s as a separate subject from grade one to grade ten in basic
Seventh Five Year Development Plan (2006-2010) lays education. In grade eleven, the ICDL (International Computer
emphasis on upgrading the information technology sector by Driving License) is a required course and each student has the
implementing the national strategy for Oman’s digital society opportunity to learn IT skills [2].
[2]. The Information Technology Authority of Oman, which
spearheads these initiatives, provides a definition for ‘digital Nevertheless, considering the fact that nearly one-third of the
society’ as a modern, progressive society that is formed as a population is beyond the usual ages of formal education 1 ,
there is a need for alternative means of achieving basic citizen
readiness to benefit from the ongoing digital society
initiatives. Perhaps, some non-formal means of achieving IT
The pilot IT literacy campaign part of this study was fully supported by
the Middle East College of Information Technology, Muscat.
1
Sherif M. Aziz is with the Public Establishment for Industrial Estates, Estimated from the data in the Statistical Year Book- 2004 published by
Sultanate of Oman (e-mail: Mohammed.Sherif@peie.om). the Ministry of National Economy.
289
literacy would yield better reach and coverage in the shortest tested questionnaires, a follow-up telephone interview of the
possible time. participants, and focus group exercise to evolve the national
strategy.
In the above background, a group of volunteers under the The focus group consisted of a homogenous group of ten
leadership of the author, and with the support of an persons who had equally fair idea about the IT literacy
educational institution, carried out a pilot IT literacy campaign campaign. They had sufficient understanding about the
in the district of Barkha in Oman. This campaign was planned concepts of e-government and had basic awareness about the
as an action research, which provided ample insights for Digital Oman initiative.
evolving a national strategic framework for achieving IT
literacy. The IT literacy campaign as such is not a new concept.
However, the locale under consideration was not very familiar
Objectives of the research: with such non-formal and non-governmental interventions.
Preliminary assessment of the situation demanded an action
The overall objective of the research was to evolve a basic research, which follows an ‘intuitive proactive’ approach as
strategic framework for a national IT literacy campaign, with against a ‘rational reactive’ one [4]. In the first case,
a view to enhance citizen e-readiness in the context of the practitioners who know or think they know, what needs to be
Digital Oman initiative. done, implement an intervention program first and then see
how well it is progressing. In the latter approach, the
Specific objectives: researcher examines what is occurring with a specific focus on
some known problem and then draws up a program to react to
To design and carry out a pilot IT literacy campaign in a semi- what has been discovered.
urban region as a test case through an action research.
II. THE REAL ACCESS
To evaluate the lessons learned, and other analytical findings
resulting from the aforesaid objective, in the background of The information (technology) era demands a higher degree of
relevant international thinking and literature on the issues of literacy and complexity of skills to function effectively in
access to ICT, IT literacy and digital divide. what is known as the “digital society”. An OECD adult
literacy survey [5] reported that, across 20 countries, one in
To draw inferences from the aforesaid stages, and to evolve a four adults who participated in the study do not possess the
basic strategic framework for a national IT literacy campaign. necessary skills to manage in today’s world. Findings point to
large disparities in the average level and distribution of
Scope of the research and its limitations: literacy skills, both within and between countries. The report
pointed out the need for policies to be directed at the work
With regard to the IT literacy campaign, there was a larger place and family settings so that an overall societal
community of interested parties such as the Wali 2 and his preparedness for the information age is possible. Lack of
office, local community leaders, schools where the initial literacy skills does cause difficulties for living, working and
campaigns were conducted and, of course, the national even survival [6]. It is not the work place alone, which
government. Therefore, from concept to implementation, demands such skills. Even one’s own healthcare demands
subtle socio-political issues had to be handled carefully to skills to access and use appropriate information.
maintain the interest and motivation of all the key actors. In
such a background, apart from coping with the inherent Being information-literate has been defined as "having the
methodological complexity, the very management of the ability to recognize when information is needed, then to be
campaign, data collection, and the research in general able to locate and evaluate the appropriate information and
demanded limiting the IT literacy campaign to a pilot action use it effectively"[7]. These abilities have always been
research in a limited geographical region. The pilot literacy important to lead a successful life. However, as of now, IT
campaign also had to be short to maintain the morale and skill has become an important component of information
motivation of the volunteers. In the light of the above, the literacy, given the fact that the channels and quantum of
field study was limited to the Wilayat of Barkha. information have multiplied phenomenally on account of the
rapid technological advancements in the field of information
Methodology: and communication technology.
The question of digital divide
A mixed methodology was adopted for conducting this
research. This included a literature survey, selection of Digital divide is a term that emerged in the 1990s as an
appropriate theoretical framework, structured and unstructured expression of concern about the distribution of access to and
interviews, action-research through an IT literacy campaign, a benefits from advances in communications and information
survey among the participants of the campaign using pre- infrastructure. Digital divide, in simple terms, is the
differentiation between those who have access to IT and
2
Wali is the administrative head of a district, which is known as the digital information (particularly, the Internet) and those who
Wilayat
290
do not [8]. But, with rapid technological advances, technology The foregoing, however, appears to be idealistic, and is more
itself is a moving target, thereby enhancing the rate of further close to a developed country situation. In a developing
divide between what could be called as the ‘information country context e-readiness requires to address the issues of
haves’ and the ‘information have-nots’. basic literacy, poverty, health, and other social aspects first. In
any case, digital divide is increasingly a matter of universal
The British Council’s info@UK website glossary defines concern, as the personal computer and the Internet have
digital divide as the gap between those who can effectively become critical to economic success and personal
use new information and communication tools and those who advancement. Therefore, understanding its subtle nuances in
cannot. This divide can be the result of one or many different the context of the developing world is extremely important.
factors. Being a member of a socially excluded or
disadvantaged group, or lack of necessary skills to participate The Access for Opportunity Framework
are all important factors [9].
ICT for development has traditionally focused more on
The average OECD country, which has 11 times the per capita technology infrastructure issues. It is true that, for most
income of a South Asian country, has 40 times as many developing nations, creating physical networks is very much
computers, 146 times as many mobile phones and 1036 times beyond their means. However, there are a few other cash-rich
as many Internet hosts [10]. North America is home to only 5 developing nations such as those in the Persian Gulf region,
percent of the world’s population, but represents 28 percent of which are examples to the argument that technology
the world’s Internet users. The Middle East with about 4 infrastructure is a necessary, but not a sufficient condition for
percent of the world’s population has only 1.4 percent of the access to various aspects of a digital society.
world users [11].
It is in the above context that the Access for Opportunity
The foregoing picture is with regard to the existing disparities. Framework, which is detailed in the United Nations Global e-
The rapidity of technological advances is, in fact, pushing the Government Readiness Report 2004 [15], becomes relevant to
‘have not’ countries further behind. But the disparity with study the situation of developing countries such as the
respect to e-readiness is not merely due to technological Sultanate of Oman. This approach emphasizes the aspect that
components. As of now, it is well accepted that the digital the end goal should be access for opportunity, rather than
divide is not merely about technology deprivation, but is also access to ICT.
about many inter-related social, educational, cultural, political,
and economic issues. An ITU opinion poll had shown that The Access for Opportunity Framework lays out the
77.3 percent of those covered by the survey citing poverty as a parameters of real access. The salient points of this concept
major barrier towards integrating to the information society, are:
followed by lack of education (76 percent), and lack of • Physical access to ICT is only the first step towards
infrastructure (72.8 percent) [12]. The UN report [13] states: building real access.
• Access must be blended with relevant and culturally
As countries progress in employing ICTs appropriate content for onward transmuting into
for development, the challenges knowledge
underpinning inequality in access have • The blended knowledge is processed and utilized to
shifted from connectivity issue to create opportunity for economic and social
encompass a wide array of economic, empowerment
social, cultural, and language
barriers…The issue of a digital divide is This framework goes further to outline the Model of Access-
essentially one of a disparity in real Acceleration, which maintains that:
access which is inequality in both
physical access to ICTs and the ability, Technology infrastructure needs to reach
know-how and the culture to use the some threshold level in a given nation
technology well. for access to start accelerating, but only
as long as other access-supporting
The Centre for International Development at Harvard economic, social, educational, and
University defines an e-ready society as one that has the cultural elements are in place [15].
necessary physical infrastructure (high bandwidth, reliability,
and affordable prices); integrated current ICTs throughout Thus, a nation, which is advanced in terms of educational,
businesses (e-commerce, local ICT sector), communities cultural and social support structure, is in a better position
(local content, many organizations online, ICTs used in when the technology infrastructure reaches the threshold
everyday life and also taught in schools); the government (e- access point.
government); strong telecommunication competition;
independent regulation with a commitment to universal The Access-Acceleration Model is based on the basic premise
access; and no limits on trade and foreign investment[14]. that there is a threshold level of `real access’, which when
291
attained, will allow the subject to enter a state of accelerated implies, a felt need for information, a sense of purpose, and
access. The threshold level of access comprises ICT the ability to use information, all of which put together means
infrastructure, penetration of technology, government the realization of opportunities and empowerment due to
leadership, education directly supportive of technology, information availability.
culture of technology, and global language of technology
[15].This model goes further to make the following Access But at least the process of seeking information is a multi-
Framework definitions: dimensional technology issue, one dimension being the basic
or even rudimentary ability to use the related technology.
Access threshold : A static cut off point where the mix of However, the Internet, which is currently the ubiquitous
technological, economic and social systems begins to blend means of global information sharing, is still substantially a
together to provide a synergy which allows positive feedback text-based technology. Therefore, all the aforesaid aspects
cycles of technology for the utilization of knowledge, which become significant only if the person is fundamentally literate
allow for greater economic and social opportunity. Beyond the in a language, which is used in the technology driven
access threshold a country enters the accelerated access phase. information network or environment.
Accelerated Access: A dynamic changing optimal mix of Therefore, being IT literate would have three almost
technology, educational skills, economic and societal hierarchical components viz. (a) basic literacy in a language
conditions, which come together at any given point in time to (b) technical literacy of being able to use the related
produce threshold access, which in turn leads to real access. equipment or computer literacy and (c) being information
literate as explained above (see Fig.1).
Real Access: The equilibrium level of access, whereby an
individual has the required availability of technology,
educational skills, culturally appropriate and relevant content
in his/her language of choice at an affordable cost.
The UN report [15] states that the vast majority of countries Information Literacy
have not yet reached real access or even a threshold level of
ICT infrastructure needed to enter the acceleration stage, and ---------------------------------
most are lagging in the other critical access elements.
A detailed study, which will consider a possible maturity Computer Literacy
process to real access, was beyond the scope of this research.
Therefore, in this study, the concept of real access is viewed ---------------------------------
in a very simplistic manner as an equilibrium of availability of
technology (basic telecom and internet access), education and Linguistic Literacy
skills, culturally appropriate and relevant content in the
language of choice at an affordable cost, and the motivation
(sense of purpose) to use.
IT Literacy redefined
Fig.1 Hierarchical components of being IT Literate
The information age is largely characterized by the
technological innovations that have had far reaching
In regions such as the Persian Gulf, and therefore in Oman,
consequences in the way people work, communicate and do
the first component itself is a matter of concern in that the
business. In fact, the convergence of telecommunication,
abundance of information available in the so called world-
broadcasting and computers has had significant impact on the wide web itself is not in the ‘language of her/his choice’,
society, which is comparable with the previous major societal which is Arabic. On the basis of the above concept, it can also
changes such as the industrial revolution or the advent of the be said that being information literate also indicate a general
printing press. As with the advent of the printing press, the maturity with regard to the human capital, not necessarily in
technological age requires the development of higher levels of terms of mere educational qualifications, but essentially the
skills in the society [16]. It is not enough to be literate in the positive impact of education and technical skills to lead better
sense of being able to read and write in some language. If a quality of life. In other words, such societies will have an
person is to survive in the information age he needs to be ‘information maturity’ to make use of the information and
‘information literate’, which, simply put, means that he has knowledge around them to lead a better life.
the skills to seek, understand, and use information 3 . But it also
3
The American Library Association Presidential Committee on needed and have the ability to locate, evaluate and use effectively the needed
Information Literacy 1989 defines Information Literacy as: “ To be information (ALA, 1989 p.1)
information literate, a person must be able to recognize when information is
292
The foregoing is the other conceptual framework, which has (c) The Internet Bus: The Internet bus was used for promoting
been evolved and used in this research. the campaign, for providing follow-up support to the already
trained participants, for providing basic training to small
III. IT LITERACY CAMPAIGN groups assembling in souks 4 , and to reach out to the interior
villages.
Citizen e-readiness is a basic necessity for any digital society
initiative to be successful, and IT literacy is one of its primary The volunteers were divided into three groups – (a) publicity
components. Perhaps, for quite some time, it could have been group (b) trainers group, and (c) follow-up group. The overall
taken for granted that such IT literacy will come through the approach was that of ‘training of trainers’ within the
formal education system. However, at least for the time being, community, which was expected to result in a chain reaction
this will be inadequate considering the fact that about 34 and speedy coverage.
percent of the Omani population has already crossed the usual
age of schooling [17]. Therefore, there is a necessity for some Lessons learned during the campaign:
non-formal means of making the population IT literate.
However, since this is an issue of the masses, a hit and try Many observations were recorded, discussed on a daily basis
approach can result in a strategic catastrophe. Therefore, such and corrective measures taken. The major issue was that the
a non-formal intervention is best implemented after a pilot campaign could not attract significant number of men and
project so that, based on the lessons learned; a nation-wide women above the age of 30 years, though it was primarily
strategy can be evolved and implemented in the most effective aimed at attracting people who have already passed the usual
manner. age of formal education.
It is in the above background that the IT literacy campaign at Therefore, the Internet bus was deployed to the souks to
Barkha was organized as an action research. attract aspirants from higher age groups. Even with sufficient
cajoling by the volunteers, older citizens were not able to
Overall organisational set-up: relate themselves to this activity. By the second week of the
campaign it was more or less clear that the campaign has to
The campaign had to be legitimized by governmental reach out to the villages in a more intrusive way to attract
patronage. This was obtained from the district administration. older women and men.
This campaign had the support of the Middle East College of
Information Technology, Muscat, by way of providing student Converting the campaign into a family affair – a minor
volunteers and a fully equipped Internet bus. breakthrough:
A core team headed by the researcher, five volunteer leaders, The whole campaign actually worked on the concepts of
and three technical support staff was formed with full voluntarism and local community participation. In order to
autonomy to plan and organize the campaign. About 60 break multi-dimensional (social, cultural and gender issues)
volunteers actively participated in the various activities of the hurdles in reaching out to the household level, the volunteers
campaign. were encouraged to take the Internet bus to the villages, where
she/he can impart basic awareness training to mostly people
Design of the campaign: who are her/his relatives and neighbours, all to be done by
her/him without any outsider being there. This approach
The IT literacy campaign was structured in the following generated more active participation from women, but still
manner: could attract only a very few above the age of 40 years.
(a) Two days class-room sessions: This included introduction Survey among the participants:
to the era of computers, creating a sense of purpose for
gaining IT literacy, and creating awareness about the A survey was conducted among the participants of the IT
opportunities in the information era and in an e-government literacy campaign. A pre-tested questionnaire was
environment. This was followed by general familiarization administered to all those who attended the campaign. First set
with computer. Trained volunteers gave lectures quoting real of data was collected prior to imparting training and at the
life situations. Subsequently, small groups were formed to time of registering for the campaign. The participants were
provide hands-on training. asked to fill and return the second part of the questionnaire
after the training. Subsequently, 45 days later, a telephonic
(b) Walk-in-and-use support for the next two days: Two interview of the participants was done to cross-check certain
volunteers from the neighbourhood of each venue were made data as well as to get some supplementary information such as
available for about three hours per day to provide further actions taken after the campaign.
helping hand to those who have already attended the
campaign on the previous two days. 4
Old market place with very narrow streets, typical of the erstwhile Arab
commercial centers
293
TABLE II
The empirical data obtained from the aforesaid survey as well MOTIVATION FOR ATTENDING THE CAMPAIGN
as from the published documents were used to analyse the
issue of citizen e-readiness in Barkha, in the background of Sl. No Motivational Number % to total
the UN’s Access for Opportunity framework and in the factor participants
context of the Digital Oman initiative. 1 Acquire computer 64 31.2
skills
Details of the analysis and findings of this primary data are 2 To get a decent 74 36.1
given under the Study Findings in section IV. job
3 Can’t live without 55 26.8
computer skills
IV. STUDY FINDINGS
4 Upgrade the skills 20 9.8
Though the campaign had targeted attracting more of those
who have passed the age of formal education, most of the 5 Others 10 4.9
participants (55.1 percent) were from the age group of 16-25 Note: Figures are not mutually exclusive.
years, followed by the age group of ‘up to 15 years’ (Table I).
A rough comparison of this with the age distribution in the Education
national population reveals that the campaign could not attract
all the segments of the local population representatively, the Majority (30.1 percent) of the participants had education up to
maximum distortion being at the two upper age-groups. the 12th class, followed by those with 10th class (14.9 percent)
Incidentally, the first age group seems to be properly and 11th class (16.6 percent) education (Table III).
represented.
TABLE I Nearly 60 percent of these participants with higher secondary
AGE DISTRIBUTION education did not have computer skills. In fact, no correlation
Corresponding was found between the level of education and computer skills
Age representation or degree of awareness about Digital Oman or e-government.
Sl. Number of % to
Group in the national
No Participants total TABLE III
(years) population (%)
* EDUCATIONAL BACKGROUND
1 Up to 74 36.1 34 Sl. No Level of Number of % to total
15 schooling participants
2 16-25 113 55.1 22 (class)
3 26-40 14 6.8 25.2 1 12 62 30.2
4 More 4 2.0 18.8 2 11 34 16.6
than 40 3 10 44 21.5
Total 205 100.0 100.0 4 7 to 9 51 24.9
Note:* Approximate figures computed based on the national population
5 Up to 6 14 6.8
statistics, sourced from the Oman Census Survey Report 2003. The total Total 205 100.0
population is 2.34 million with more than half a million expatriates.
From the above observations, it is inferred that the formal
Apparently, computer is still seen as a modern tool meant for school education in Oman up till now has not made significant
the younger generation. This aspect, seen along with the fact contribution towards general IT literacy. Further, about 34
that more than 30 percent of the participants had higher percent of the Omani population is already outside the scope
secondary education (see Table III), indicate that many of of formal education. Therefore, the non-formal means of an
them were job seekers trying to acquire more computer skills. IT Literacy campaign was an appropriate intervention.
This is also evident from the higher representation (55.1
percent) from the relatively more active job seeking age group Literacy
of 16-25 years.
An analysis of all the hierarchical components of being IT
The expressions of motivational factors for attending the literate (see Fig 1) has been attempted here. The Barkha
campaign reinforce the aforesaid inferences (Table II). survey revealed that 87.8 percent of the people could read and
write Arabic (Table IV). This is fairly a high level of literacy
rate in a developing country. Many Arab countries are likely
to have such high Arabic literacy, as reading the Holy Quran
is almost a must from the religious point of view.
294
matrimony are all relevant contents. The relevance of these
TABLE IV contents also depend on the maturity of the services offered by
LITERACY the different actors in these sectors, be it in the private or
public sector. For instance, an online shopping portal is of no
Nature of Literacy No. of persons % to total significance if the associated online payment and the delivery
including all (including mechanisms are not intact, or if the legal framework is not
family members family mature enough to support e-commerce.
of the members)
participants * The above observations should be seen along with the fact
Arabic (Read) 1471 93.5 that in the Barkha sample only 27 percent had basic computer
Arabic (Read & Write) 1381 87.8 skills. This is out of a sample population of predominantly
English (Read) 889 56.5 youth with nearly 60 percent having higher secondary
Basic Computer Skills 427 27.0 education.
Note: *Total number of family members covered – 1573
From the point of being IT literate, the last component is to
The published national statistics show about 80 percent become ‘information literate’ (see Fig.1). From the survey of
literacy rate among Omanis [18]. Obviously, this is their those who attended the campaign to acquire IT literacy, it is
ability to read and write in the local language, which is not possible to assess the status of their ‘information literacy’.
Arabic. In fact, the figure of 87.8 percent obtained from the Such a survey needs to be carried out differently among a
Barkha sample is almost in line with the national figure. The larger cross-section of the population. However, a sum total
survey also revealed that only 56.5 percent could read of all the above observations points to the possibility of weak
English. Those who had computer skills were even less – 27 information literacy among the population, no matter whether
percent. it is computer-based or not.
The English reading ability in the national population can be Access-for-Opportunity framework – A revisit in the
much below what is reflected in the Barkha sample, as this context of Oman:
sample had a poor representation of people above the age of
40 years. Further, a substantial number of participants of the The UN Access-for-opportunity framework detailed under the
campaign had higher secondary education as well. section REAL ACCESS talk about the importance of a perfect
blend of technological, economic and social systems to
The opportunity to use one’s own language on global provide a synergic situation for ‘real access’. Considering the
information network such as the Internet determines the extent various dimensions of this approach, a slightly modified
to which one can participate in the knowledge society. conceptual framework was evolved to include some finer
However, an estimated 87 percent of documents on the elements such as ‘motivation or sense of purpose’ among the
Internet are in English [19]. The UN Report [20] confirms that masses to make the ‘use of ICT a way of life’ (Fig.2). Such
English continues to dominate the cyberspace material. subtle nuances can look trivial but are often critical
components of many development issues involving a society.
Only 1.4 percent of the world online users are Arabs, whereas In this modified conceptual framework, the elements of
they represent 4.6 percent of world population. This is in spite ‘physical access’ and ‘affordability’ are considered together
of the fact that the Internet use in the Middle East grew 219 along with a via-media solution of ‘public access’ points, all
percent during the period 2000-2004. This is in sharp contrast of which put together can provide a synergic effect in
to the situation of 39.4 percent of world online users using enhancing the ‘access to technology’.
English content, where as only 5.4 percent of the world
population are English speaking [20].
Thus in the case of Oman, though Arabic literacy is high, the PHYSICAL ACCESS
Electricity
issue of ‘linguistic literacy’ has two aspects to tackle in order CONGNITIVE ACCESS
Telephone
Computer
to enhance IT literacy: Education
Internet
IT Literacy
Awareness
The foregoing lacunae are apart from the issue of ‘relevance’ Motivation/Sense of Purpose
website ACCESS
Digital Oman
15 Any step 25 180 Not enough
taken after initiative? Fig.3. Access to Technology in Oman
campaign
5
This is as per the PPP (Purchase Power Parity) Method. As per the Atlas
Note: Incidentally, 60 percent of the participants had higher secondary Method the per capita income of Oman is US$7830, with almost the same
education. global ranking.
297
Cognitive Access has strict filtering of pornographic material; and practically do
not have any other control mechanism.
Only 7.7 percent of the national population has post-
secondary education [18]. It was also shown earlier that in the
Enabling environment - Motivation or sense of purpose
sample population covered by the survey, the school
education has not contributed significantly towards IT
19.5 percent of the participants knew about e-government in
Literacy among the masses.
general, mostly through media reports about global
developments. Only 9.8 percent had heard about Oman e-
The overall picture obtained from the analyses under the sub-
government initiatives. Out of the 40 participants who were
headings Education and Literacy is that of a poor cognitive
aware about e-government, 29 persons still felt that going to
access situation, with weakness at the levels of linguistic
offices personally is better to ensure speedy processing. A
literacy, computer literacy and information literacy, to
look at Table VIII reveals that the responses to most aspects
participate and benefit from a world-wide information
connected with the individual and his capabilities are positive
network.
in nature (descriptions corresponding to Sl. no. 1 to 7). Others,
wherein government has to play an active role in popularising
In the absence of sufficient cognitive access, the digitization
or reaching out with the correct message with regard to e-
of government processes, which is part of the Digital Oman
governance, the people need to gain better awareness and
movement, is unlikely to have much impact on the quality of
confidence levels. But the fundamental attitude of the people
life of the wider cross-section of the Omani society.
is positive enough to mature into an information society,
provided the related government interventions are appropriate,
Addressing the issue of linguistic literacy can have two
timely and motivating.
possible approaches:
The ‘real access’ issue in Oman
(a) A long-term strategy of enhancing Arabic
content
All the above findings put together, it can be concluded that,
(b) A short-term solution of popularizing the
among the various components of `real access’, the Omani
learning of English so that what is available in
population is almost at the threshold of access to technology,
English in the world-wide-web is readily usable,
but are yet to reach an advantageous position with regard to
and relevant
cognitive access, content access, and the required sense of
purpose to partake and benefit from the digital society era
Even computer literacy can be considerably enhanced through
(Fig. 4).
a multi-pronged approach of:
All the above analyses also reveal a situation of very limited Poor awareness ; some
awareness of global
PCs exist.
government web sites, which are mostly in Arabic, do not about local initiatives. No
evidence of motivation or
users.
The survey among the participants of the campaign revealed Fig.4. Assessment of ‘Real Access’ Situation in Oman. Grey
that among those who have visited one website or the other boxes indicate challenges.
(i.e. 26 % of the Barka sample) only 32 percent felt that they
got what they were looking for.
Majority (91.7%) did not have any apprehension about
cultural erosion due to the Internet. This country, of course,
298
V. KEY CONCLUSIONS As is evident from the figure, what has been proposed is a
multi-pronged and holistic strategic framework, with the
This section tries to highlight the key conclusions of the study. following broad considerations:
The predominant picture that is emerging out of this study is a. There need to be a thrust in maximizing social
that the access to technology cannot be considered as a inclusion, which does not necessarily mean
deterrent to citizen e-readiness in Oman. However, the issue making everybody IT literate; but it would
of cognitive access capabilities need to be addressed from simply mean maximizing social inclusion
different angles and a non-formal means of training (such as through a multi-pronged approach, including the
the IT literacy campaign) is imperative to alleviate possible use of intermediary mechanisms.
digital divide. b. Those above the ‘job seeking’ age group have to
be reached out through mobile training facilities
Some of the specific conclusions are: in a sustainable and friendlier manner.
c. An appropriate communication plan is a
a. The issue of ‘linguistic literacy’ has two problems to prerequisite to create the right type of awareness
tackle. Firstly, the inability to use the relatively huge about all what is happening in the name of
English content available in the World Wide Web. digital society.
Secondly, though Arabic literacy in Oman is high, d. Attempts shall be made to involve Sheiks 6 as
the web-content in Arabic is limited. champions who would take the ownership of IT
b. Therefore, for the masses to be IT literate and derive literacy initiatives at the local level.
meaningful benefits of a digital society, the language
access barrier may be considered as a fundamental It may be emphasised that, even to achieve IT literacy and
deterrent. thereby e-readiness among the people, the overall
c. Addressing the issue of linguistic literacy requires a environment should be conducive and encouraging to create a
long-term strategy of enhancing Arabic content as sense of purpose among the people to gain these new skills.
well as a short-term solution of popularizing the This can be achieved if the related government interventions
learning of English. can bring demonstrable benefits that will touch upon the
d. The possibility of weak information literacy among quality of life of the average citizen.
the population has to be considered while evolving
digital society strategies.
e. The findings indicate that affordability or financial Barka model campaign at
capability to own and access computers cannot be a
Recognise the need for social change and
schools; Separate training rooms preparedness; Maximize social inclusion;
for men and women; Reach out Minimise gender disparities while achieving
matter of serious concern in the context of Oman. to the village neighborhoods and
family gatherings through mini-
targets; Use of ICT shall become a way of life.
Consider that Oman is at the threshold of access
f. There is a general lack of sense of purpose for having internet bus; More public access
points – some free, some as IT POLICY
to technology – if public access points and more
promotional schemes to own computers are
appropriate threshold level with respect to access to Awareness training to government officials, Training of trainers approach; Enhance
technology. community leaders (includes Sheikhs and elected ACTORS English and computer skills at schools;
members) and key private entity officials, as a National Government Generate a sense of purpose and
Local Government
i. There is a poor cognitive access situation, with
prelude to mass campaigns- shall aim at motivation through demonstrable projects;
transforming their attitude and their way of dealing Private entities Immediate steps to enhance Arabic web
with the public in the information era; Define their People content – in terms of quantity, relevance
weakness at the levels of linguistic literacy, computer roles in the national IT literacy mission; Multi-stake
holder situation- source support from the private
Community leaders
Volunteers
and reliability; Enhance trust
confidence in local web content and e-
and
literacy and information literacy. sector and the academia; Incentives to community Academia services; Quality standards for government
volunteers websites; At least one IT literate person in
every household in two years.
VI. RECOMMENDATIONS
Fig.5. Strategic framework for a national IT literacy campaign
E mergency management is “the process of gathering a failure. During the early hours of December 26th, 2004, such
resources and acting upon the problems immediately effective communication was not demonstrated on many fronts,
during and after a critical incident” [1]. Emergency incidents both national and international. Indian soldiers on the islands of
may be natural or man-made. Limited information, Nicobar and Andaman sent warnings to their mainland
unpredictable disaster development, short time window, and counterparts after getting hit a full two hours before it struck the
high operation complexity renders the management of Indian coastal region. These warnings went unheeded and no
emergency incidents a challenging task [2, 3]. In order to evacuation orders were issued for Sri Lankan or Indian coastal
mount an effective response to an emergency situation, it is regions resulting in the loss of tens of thousands of lives. The
critical that the key actors communicate effectively and timely head of Thailand’s Meteorological Service refused to issue a
so as to exchange task critical information and form tsunami warning despite the looming danger because he didn’t
collaborative response activities. While communication has want to cause unneeded panic [4]. News of the true damage
been increasing recognized as the key to success emergency done to Aceh Indonesia was not known for hours after the
response, the practices in the field are far from satisfactory and initial impact possibly due to cut phone lines and poor military
failures are frequently documented. Fig. 1 presents a brief communication channels [5].
illustration of emergency communications involving the The goal of this research is to uncover the dynamics in
variety of stakeholders. emergency communication and to recognize the major
challenges and barriers. More importantly we discuss the
available technical instruments and propose their design and
management to meet these challenges for effective inter-agency
communication. The Indian Ocean Tsunami provides an
Manuscript received September 21, 2008. This work was supported in part
by the National Science Foundation under Grant 0809186.
opportunity to observe and analyze how the information and
Rui Chen is with Ball State University, Muncie, IN 47306 USA (phone:716- communication technologies impact on the socio-economic
880-2161; e-mail: coupae@gmail.com). development in the world. Despite of the plethora of prior
John Coles is with State University of New York at Buffalo, NY 14260 USA
literature on the response to Indian Ocean Tsunami, little work
(e-mail: jbcoles@buffalo.edu).
Jinkyu Lee is with Oklahoma State University, OK 74106 USA (e-mail: has been done to analyze the phenomena using a theory driven
jinkyu.lee@okstate.edu). approach. As a consequence, the knowledge thus gained is
H. Raghav Rao is with State University of New York at Buffalo, NY 14260 segregated and a complete understanding is missing. Through
USA (e-mail: mgmtrao@buffalo.edu)
301
the lens of Activity Theory, we identify the key issues email) to exchange task critical information (e.g., chemical
pertaining to the socio-technical systems involved in release) in an attempt to build common operating picture of the
emergency communication. This paper thus contributes to the status quo of the ongoing incident and mitigation. This
improvement in information system to enable and support inter-agency communication is influenced by the rules (e.g.,
emergency communication in large scale disasters. standard operating procedure –SOP), community (e.g., both
This paper is organized as follows. In the subsequent section, professional communities and municipality communities), and
we review Activity Theory that is used in this paper as the division of labor (e.g., operations, logistics, and planning) in
theoretical foundation to explore emergency communication. emergency management realm [16].
Next, we apply Activity Theory in emergency management Instrument
using the case of tsunami disaster. We then elaborate the Email, Phone
information technology design and management to cater to the Real Time
requirements and challenges raised. We conclude this paper Information Sharing
with limitations and future studies.
Subject Activity Object
II. THEORETICAL FOUNDATION Emergency Communication Situational
With the advancement in modern information technologies, Responder Awareness
information systems designated for emergency management
have been practiced and they have proved to play an important
role in facilitating the incident management [6-8]. To design
and implement effective systems for emergency
communication, it is important that we understand the set of Rule Community Division of Labor
design requirements. The development of communication Standard Government, NGO, Task & Task
operating volunteer, business, Assignment
systems requires systematic approaches to elicit and analyze the procedures news media, victims
internal elements, structure, and relationships of core
Fig.2. Application of Activity Theory in Emergency Communication
communication elements. In this paper, we use Activity Theory
to guide the requirement engineering process in
communication system development [9, 10]. An approach III. COMMUNICATION IN EMERGENCY MANAGEMENT
driven by Activity Theory represents a method that has gained In this section, we employ Activity Theory to analyze
increasing attention in recent years [11-13]. Activity Theory emergency communication in an attempt to understand the
provides a lens to analyze the computer-supported activity of a challenges and requirements for emergency communication
group or organization and to study the design of artifacts for system design.
individuals and organizations.
Activity Theory suggests that human activity is directed A. Emergency Communication in General
toward a material or ideal object, mediated by artifacts or In accordance with Activity Theory, the subjects in
instruments, and socially constituted within the surrounding emergency communication are the first and second emergency
environment [14]. Activity can be understood as a systemic responders. They may include fire department, police
structure with various activities that are collated or extended department, emergency medical services (EMS), hazard
away from the core activities [15]. The subject is the active materials workers (Haz-Mat), etc.
element of the process and can be either an individual or a The subjects actively communicate with each other to
group. The object transformed by the activity can be an ideal or achieve situational awareness. Biros et al define situational
material object. The transformation process is enabled and awareness as “the decision-makers’ moment-by-moment ability
supported by instruments (physical or logical). The instrument to monitor and understand the state of a complex system and its
provides the subject with the experience historically collected environment” [17]. During emergency, the completeness and
by his/her community. During the interaction, subjects accuracy of decision makers’ situational awareness is crucial to
internalize and/or externalize their cognitive schemes and their their abilities in comprehending the disaster facts, hazards and
understanding of the relationship between themselves and the risks, and mitigation capabilities, which “piece together an
external objects, instruments, surroundings, etc. Activity accurate of reality” [18].
Theory also considers interaction as one critical aspect and The communication activities function on the basis of
suggests that interactions are the driving force in human technical and non-technical instruments such as existing forms,
interaction and system design. The interactions may also exist guidelines, and alert systems. These instruments are likely to be
inside the subjects, objects, instruments, and their interactions. developed under the supervision of the individual actors and
In Fig. 2, we present the application of Activity Theory in with propertied designs in terms of taxonomy, notation, and
emergency management. That is, the emergency responders ways of implementation. For example, the fire companies in
(e.g., firefighter, police, and emergency medical service neighboring counties may adopt different instruments (e.g.,
personnel) use the communication systems (e.g., phone and alert systems and reporting forms) in managing disaster
302
response. These heterogeneous and incompatible instruments involved as is the news media. The four issues that plague
are typically involved in a large scale incident response where private citizen involvement are as follows. These four issues
multiple subjects are called in. are exacerbated by the involvement of news media.
Emergency communication is subject to influences from 1. Easy numbness to warnings
rules and division of labor. The rule systems embody the norms 2. Dramatization of information given
and conventions adopted in emergency response community 3. Mistrust of centralized government
and their management practices. Example may include 4. Demand for constant information especially immediately
responders’ attitudes toward risk taking and risk sharing. before and after a disaster occurs
Division of labor, on the other hand, refers to the individual News media descends upon any significant event quickly
domain expertise and tasks typically performed by response and in an unfiltered manner, however, they are also the primary
subjects. Based upon the rules in place and the division of source of information that private citizens turn to in emergency
labor, subjects may behave differently in their communication situations. For this reason it is crucial for the news media to be
behaviors. Additionally, response community impacts the viewed as an ally and not an enemy in emergency response
emergency communication patterns. General speaking, information gathering and dissemination. A relationship built
emergency community include government agencies, on mutual trust in critical to the effective utilization of local and
non-governmental organizations (NGOs), independent international media and should be closely guarded [23].
volunteers, local businesses, news media, and victims [19]. Table 1 summarizes a list of typical challenges to emergency
Government agencies are often the actors looked to in times communication. We elaborate an example challenge.
of disaster, thus it is critical for open and functional Table 1. Emergency Communication and Challenges
relationships to be established prior to a disaster in order to
facilitate effective communication. With the variety of Challenges to Effective Communication and Action
Emergency Response
governmental levels and the red tape that must be sorted Tasks Data Time
Scope of
Lack of Information
Decisions
through, it is also critical for the intra-governmental Quality Pressure Trust Complexity
Impact
communication channels to be aided by the early establishment Creation of a Common
X X X X
of relationship and trust [20, 21]. Without these additional Operating Perspective
channels the government paths are dysfunctional and often Formal Decision
X X X
Making Procedures
avoided by other actors when possible. The issue of agency Collection of Relevant
tribalism between local government sanctioned response X X X X
and Effective Data
agencies adds a new dimension of complexity to the situation. Correct Interpretation
X X X
In the United States, an example of this comes up often between of Data
Appropriate Actions
the individual fire companies and the police departments due to Identified and Executed
X X
issues in funding and jurisdiction. Coordinated Response
X X
Some governmental actors can be highly effective in Effort
emergencies by nature of their position or the information and
resources to which they have access. Local elected officials are
The quality of data which a disaster generates can easily be
a great avenue of communication to the general populace and
overwhelming for three major reasons: the varying lack of
are most effective when they have the crucial relationships
quality information, the time consuming task of accurate data
established with potential resources in the region and have the
interpretation, and the differentiation of pertinent compared to
trust of the general populace. Another great resource are the
additional extraneous information [24]. The process of finding
government monitoring agencies such as the Pacific Tsunami
quality data sources often involves trust-bonds created prior to
Warning Center (PTWC) and the US Geological Survey
the disaster [25]. Thus it is crucial for emergency managers to
(USGS), which were the agencies on the front line of tsunami
be open to sources that could provide additional information
detection and information dissemination in the early hours of
but should also not be too accepting of all information so as not
the Indian Ocean Tsunami [4].
to be overwhelmed with the task of accurately interpreting the
Other resources that can often provide crucial data or
pertinent elements of data. Once the seemingly pertinent data is
resources are the local scientific community, private for-profit
extracted, the issue of deciding whether the new information is
organizations, and private not-for profit organizations. Local
immediately relevant is crucial in making informed and
colleges and universities can be often be a great source of
effective decisions. When the data quality is low, emergency
crucial information both prior and in response to emergencies
management is hindered as decisions are often be short-sighted
due to the intricate and invasive approach to problem solving
or ineffective in the larger response as a whole [20].
on a regional and international scale [22]. Private organizations
are great resources for both equipment and manpower B. Case of Tsunami
mobilization. From factory workers to church goers, the We discuss in this section the major communication failures
effective training and utilization of local labor is critical to a in the response to tsunami incident. Our discussion is focused
speedy and effective recovery [4]. on subject and community in Activity Theory as these are
Whenever there is a disaster, local private citizens are
303
deemed as the major failure in emergency communication. communication activities, it is necessary that one determines
Emergency responders, subjects, along Indian coast did not the scope of decision impact (i.e., parties he/she needs to reach)
equip systems capable of modeling and predicting tsunami [24]. By identifying these critical individuals, the reduction in
scales and impacts. Immediately after an earthquake, computer travel time and increase in accuracy of information between
models can calculate how fast the waves will travel, as well as actors will lead to more effective responses [21]. This broader
their amplitude. In an interview, Tsunami expert Tad Murty perspective gives the basis for a much stronger international
mentioned that he had persuaded the Indian government to response instead of a solely national response. In the case of the
build a fully functional tsunami warning system but was told tsunami, this would have significantly strengthened the early
that there were not sufficient financial resources [26]. A system response and preparedness of other nations had India, Thailand,
like this might had saved hundreds of lives. As a matter of fact, Indonesia, and the United States communicated their
26 countries in the Indian Ocean were notified of the tsunami knowledge both with each other and the other nations affected
within 15 minutes after the earthquake took pace. India was in a timelier manner. However, the scope of decision impacts
unfortunately not among them and thus missed the opportunity was hard to decide by the emergency mangers who tried to
to detect the incident. gather and disseminate information during the response to
Emergency responders also experienced difficulties in sense tsunami. As a consequence, responders were not aware of what
making of the incident and how it unfolded. In spite of a lead information should be passed on to what parties. Without a
time of almost six hours, no early warning could be issued formal central framework and method of collecting information
before tidal surges lashed the Indian coasts. The main reason centrally and disseminating, it is difficult for actors to receive
was the absence of systems to translate common knowledge or even where to start when attempting to acquire the
into scientific information. The response communities were not information necessary for an effectively mounted response [19,
familiar with the phenomenon of tsunami and had neither 22].
proper prior experience nor trainings. Using Activity Theory (AT) framework, Table 2 summarizes
Even after the information was collected interpreted, it was the observations of Tsunami response communications along
difficult to pass the information from one point to another. A with the problematic issues raised. In the subsequent section,
great majority of communication infrastructures failed in the we propose the set of solutions.
wake of Tsunami, just at the very time when they are most
needed. At Maldives Island of Sri Lanka, telecommunications Table 2. Activity Theory Informed Analysis on Tsunami Response
infrastructure was based on a microwave terrestrial backbone Communications
network. When Tsunami hit, all public telecommunication AT Construct Problems Consequence
services to the 13 atolls (163 inhabited islands) were Lack of input Failure in detecting the event
interrupted, causing the complete breakdown of the Subject Lack of subject Failure in interpreting the
communications [27]. matter knowledge raw data
Disconnections among response communities also lead to
Lack of
the devastation of tsunami incident. First, communication Failure in information
Community international
breakdown took place inside the domestic response sharing among countries
collaboration
communities. The IAF administration of India, for example,
was informed that the Car Nicobar air base was hit by tsunami Lack of Failure in information
an hour before it reached Tamil Nadu. However the first Instrument communication exchange and sharing
message was not published from the Indian Meteorology infrastructure between any two points
Department to the government until 41 minutes later. This
example reveals the lack of standard operating procedure in the Lack of standard Failure in domestic response
Rule
local government. Standard operating procedure prescribes the operating procedure information supply chain
control and flow of emergency information among the
stakeholders. Without it, emergency information will not flow IV. EMERGENCY COMMUNICATION SYSTEMS
among the involved agencies smoothly and the information The information system discipline has progressed with new
supply chain will break down. systems designs and management practices developed. For
Breakdown also took place among international response example, Hale develops a layered crisis communication
communities. The Pacific Tsunami Warning Centre of the architecture (CCA) that enumerates the communication
United States successfully recorded the earthquake in the functionality requirement for emergency information systems
Indian Ocean and warned that “There was no threat of tsunami [18]. The layers include connectivity, data-validation, filtering,
to the US area of responsibility.” While the warning would be value interpretation, organizational memory, and group process
useful to alert the countries along the Indian coast, the warning layers. Michalowski et al develop a palm-based mobile system
did not fully reach the Asian response community which could Mobile Emergency Triage (MET) [28]. In this section, we
had saved thousands of lives. discuss the technologies that facilitate emergency
Considering the broad areas affected, emergency communication. As in Table 3, the discussions are driven by the
communication activities were difficult to initiate. To initiate analysis of principal constructs of Activity Theory and their
304
interactions. United States, for example, are launching a locally-targeted
emergency alerting service featuring real-time tsunami
Table 3. AT-Driven Communication System Design warnings with evacuation routes. RAINS' Connect & Protect
Designs driven by AT Constructs service captures NOAA/National Weather Service's tsunami
AT Construct Problems Design Components warnings when issued, and immediately sends localized alerts
Lack of input Early warning systems via computers, pagers and cell phones, to local citizens
Knowledge management responsible for public safety. It will dramatically increase the
Subject Lack of subject
systems, intelligence
matter knowledge speed and reach of the warnings within a community.
processing
B. Knowledge management systems
Lack of During emergency response, individuals, teams, and
International emergency
Community international
communication organizations share and apply knowledge as they process
collaboration
information, make decisions, and act on existing knowledge
Lack of Improved infrastructure [30, 31]. Knowledge may be categorized in two distinct forms:
Instrument communication management for mobility, explicit knowledge which is easy to communicate and can be
infrastructure reliability and redundancy codified and tacit knowledge which is inextricably woven with
the experiences and situational contexts [32, 33].
Lack of standard Rule-based interoperable Knowledge management systems such as knowledge
Rule
operating procedure information supply chain repository provide vital support to help the decision makers
analyze and understand the facts. As manifested by the
Designs driven by Interactions between AT Constructs Tsunami case, responders may not be familiar with all types of
Subject –n– Community Grid computing incidents and the approaches in mitigation. The lack of training
Instrument –n– community Open source approach and expertise may account for failures as such. To this end,
knowledge management systems may be adopted to equip the
A. Early Warning Systems responders with the right information when incident first
strikes. There exist a few public knowledge repositories for
Early warning systems are important to disseminate incident
emergency response such as the Responder Knowledge Base
(e.g., tsunami) occurrence to the stakeholders in the disaster
(www.rkb.us). Private knowledge repositories are available in
affected regions. The significance of early warning for
most commercial-off-the-shelf (COTS) emergency
emergency management has been emphasized in major
management systems such as DisasterLAN (disasterlan.com),
international agendas including the Barbados Plan of Action
E-Team (eteam.com), and WebEOC (esi911.com). Systems as
for Small Island Developing States, the Johannesburg Plan of
such retain domain-specific knowledge that complements the
Implementation, and the G8 summit in Gleneagles as well as
knowledge repositories maintained by individual teams.
major environmental agreements such as the UN Framework
Convention on Climate Change and the UN Convention to C. Intelligence Processing
Combat Desertification [29].To improve the alert system, it is During emergency, the completeness and accuracy of
important to leverage the existing systems as well develop new information shared and exchanged among responders is crucial
warning systems. to their abilities in comprehending the disaster facts, hazards
A lesson learned in the response to tsunami is that the and risks, and mitigation capabilities, which “piece together an
existing warning systems function below optimal performance. accurate of reality” [18]. System designs such as information
International Tsunami Information Centre (ITIC), for example, infusion are advised in this regard. As noted above, however,
monitors the tsunami activities in the Pacific and disseminates
the information quality in emergency communication is
tsunami information and warning messages to well over 100
typically low where errors are frequently observed. For
points scattered across the ocean. Unfortunately ITIC was
example, Chen et al suggest that information on emergency
confined to its member countries before the Indian Ocean
Tsunami; this led to non-member countries such as India did incident is likely to be tainted and prone to mistakes [34].
not receive warning information in a timely manner. A series of Eye-witness accounts of the scene are often biased by their
consultation meetings led to a general agreement by comprehension processes, background, recollection, and
implementing agencies and donors that all countries affected by verbalization skills. As a consequence, triangulation-based data
the 24 December 2004 tsunami and other countries at risk to validation [35, 36] is an indispensable component for
tsunami should be involved in this project for the establishment emergency response information systems. The validation
of tsunami early warning system in the Indian Ocean Region usually relies on multiple and independent sources to validate
[29]. an event, fact, or conclusion. Designs using data correlation
New warning systems are also in need to augment the and mining techniques have proved to be adequate solutions
existing capabilities of warning systems and to utilize modern [37, 38]. Reference databases are also advertised for
computing technologies. Oregon and RAINS (Regional employment as they enhance the system competence in logic
Alliances for Infrastructure and Network Security) in the reasoning and inconsistency detection [39].
305
Field interviews with emergency responders find that channel accessibility, and system adaptability. Take
incident information may expand rapidly during the course of connectivity reliability for example. During incidents, “wired”
mitigation; the increasing level of response enactment [40] communication infrastructures are likely to be destroyed by the
accumulates new information on incident, responder, resource, disaster. It is important that the response community plan and
operation, and environment from all stakeholders. This implement technologies to remedy the communication
multi-dimension information cross interrelate with each other breakdowns. In the case of tsunami, the lack of communication
and together match the high “requisite variety” [41] of the connectivity in the most tsunami-affected Indian Ocean
decision problems in extreme events. Literature on Cognitive countries negatively affected some national and local level
Processing Capacity (CPC) suggests that individuals are activities. Options may include (1) mobile systems to deploy
limited in cognitive resources and incapable to allocate them communication channels in the field, including satellite phones,
optimally for problem solving [42, 43]. In the face of portable radio systems, WAN network, and mesh networks; (2)
information of large volume and high complexity, emergency radio-communication systems such as maritime, aeronautical,
managers are prone to be “overloaded.” The mental stresses and radio determination services; and (3) redundancy plans
resulted by time pressure, perceive risks, and concerns on with backup channels to support system breakdown or overload
“public image” further intensify the cognitive overload. To this due to request and command flooding. The existing emergency
end, tactics such as information filtering are recommended by system literature has identified and evaluated a set of solutions
prior emergency studies so as to reduce the amount of irrelevant which include increased technology investment, partnership of
data, organize related messages into cohesive and coherent sets, public/private network, load balancing scheme, priority
and prioritize message sets according to level of importance [8, telecommunication services, and redundancy design [18,
18, 34, 39]. In addition, research on Human Computer 53-56].
Interaction (HCI) suggests that cognitive overload may be Information assurance, on the other hand, refers to
reduced through design schemes in interface structure, information security and privacy. Security is a key element to
information packaging, and information accessibility [44, 45]. emergency communication as long as sensitive information is
present; it continues to be focal concern for channel design
D. Communication Infrastructure Management
considering the increasing awareness of terrorist attacks [57].
Timely communication among responders (i.e., “subjects”) System designs such as encryption, decryption, and intrusion
relies on the establishment of robust and efficient detection may suffice the security requirement. Privacy is a
communication infrastructures. Before the Indian Ocean more recent issue facing emergency management nowadays.
Tsunami, many of the countries in the region did not have a Existing regulations such as HIPPA (Health Insurance
well-organized disaster management system except for tropical Portability Accountability Act of 1996) require that privacy
cyclone-prone countries such as India and Bangladesh. information including personal identity and medical history be
National disaster management offices in most counties were carefully handled [58].
very weak, and there were few established infrastructures
(channels and platforms) for emergency communication. E. International Emergency Communication
Modern emergency management employs a variety of To mitigate large scale incident such as tsunami, emergency
communication channels including LAN, WAN, ad-hoc communication may easily span over multiple countries and
wireless network, mesh network, 800M Hz radio, and satellite districts. The presence of a coordinated communication plan
phone [46-48]. During any typical emergency operation, during emergency is pivotal to allow task critical information
hundreds of task critical updates, briefings, reports, requests, and intelligence travel across the boundaries of response
queries, and orders are circulated inside the incident communities both domestic and international.
management organizations [49]. These channels enable Regional coordination of emergency communication should
real-time information sharing and communications, first be strengthened through activities of working groups on
establishing a “common operational picture” to keep all the mitigation, preparedness and response in order to ensure
decision makers on the same page [50, 51]. A common integration of communication channels such as early warning
operational picture is a single identical display of relevant systems into national and regional mitigation, preparedness and
operational information shared by more than one actor. response capability building efforts within a multi-hazard
Through information transfer, responders working on varying framework [29]. In addition, communication coordination may
aspects of the mitigation tasks are able to synchronize their be established through both country and community level
visions on the focal incident, smooth out potential approaches to ensure those good practices, including hazard
inconsistencies, and synthesize a complete “shared mental and vulnerability assessment, organizational strengthening,
model” [52]. Information sharing mechanisms such as “publish community participation, warning system operation, capacity
/ subscribe” system are recommended for this regard. building, evacuation planning, and the design and construction
Emergency communication infrastructure is subject to of shelters, can be shared and exchanged in a coordinated
threats on performance, information assurance, and manner to avoid duplications, confusions, and overlaps.
communication interoperability. Channel performance Communication among international organizations is subject
measures the communication capacity, connectivity reliability, to interoperability challenge. The interoperability issue is
306
pronounced for large scale emergency management (e.g., offered cost-effective alternatives to help the responders. Grid
tsunami) because the technologies adopted by participating computing and grid storage has recently been employed in
organizations from varying countries and regions are in general emergency response thanks to the rapid growth in network
incompatible for reasons ranging from the ability of local bandwidth. The computing and storage resources of individuals
agencies to fund availability to the lack of unified guidelines and volunteer organizations can be pooled to boost the
for software and hardware [59-62]. To this end, it is extremely information analysis capabilities of the local responders. These
important that the emergency response community as a whole systems have been used recently to study weather, pollution,
coordinate to develop consistent data standards, transmit and remediation, etc [68].
protocols, homogenous devices, compatible application Telecommunication software and systems, instruments,
interfaces, and congruent regulations [16, 63-66]. A number of function to support communication activates when incidents
efforts have been observed national wide with as to establish strike. The development of systems as such may be improved
the communication interoperability solutions. These examples through the aids from the community. Open source approach
include the National Information Exchange Model offers the opportunity for technology innovations and
(www.niem.org), Emergency Data Exchange Language improvements in disaster communication system designs. This
(www.comcare.org/edxl.html), Common Alert Protocol approach is also known as Free and Open Source Software
(www.incident.com/cap), SAFECOM initiative (FOSS) development and community mechanism, where
(www.safecomprogram.gov), and Federal Enterprise collaborative designs are achieved through inputs from the
Architecture (www.whitehouse.gov/omb/egov). Each of them entire community. Example FOSS systems include the Sabana
addresses a portion of the interoperability challenges and more disaster system [69].
efforts are needed before a full-fledged solution may be
achieved. Without coordinated efforts, emergency V. CONCLUSION
communications are likely to break down when they travel The damage caused by the tsunami was unprecedented, but
across the boundaries of communities, both domestic and the lessons learned are crucial to creating communication
international. That is, relief teams from the world will not be frameworks that can withstand such buffeting uncertainty. By
able to communicate using their individual systems during identifying the key characteristics that can cause
response to large disaster such as tsunami, ultimately slowing communication in emergencies to be ineffective or to be
down the response progress. entirely void, future researchers and practitioners will have a
F. Rule-base Information Supply Chain frame of reference from which to approach the issue of
emergency communications [20].
Standard operating procedure (SOP) prescribes the rules to
In this paper, we employ Activity Theory to study
control and direct emergency information among stakeholders.
communication in emergency management and explore the key
Unfortunately many developing countries do not have
dimensions of issues. Taking Indian Ocean Tsunami as an
systematic SOP and they may learn from the developed
example, we identify and discuss the pitfalls in the
countries. Emergency management in U.S. is guided by the
communication practices for incident response. The authors
National Incident Management System - NIMS [67]. NIMS
further explore the issues in information system design and
facilitates the emergency communication in the following
management in an attempt to address the pending challenges in
ways: (1) it constructs a clear organizational hierarchy that
emergency communication. New knowledge gained will allow
directs the information flows; (2) it standardizes the format and
effective pro-active measures to be taken to mitigate and
content of information input and output; and (3) it establishes
respond more effectively to future disasters by recognizing the
the protocols of reporting, meeting, and cross-boundary
key players in an emergency, effectively facilitating
adjustments. And thus, SOP supports the information to flow
international cooperation, turning academic research into
smoothly among the agencies and form an emergency response
practical application, and avoiding pitfalls in areas needing
information supply chain. To this end, rule-based
greater research.
communication systems that share and exchange information
To further improve emergency communications, it is critical
following SOPs are desirable to the emergency responders.
that future studies extend the proposed design and management
G. More Designs driven by Interactions between AT factors frameworks to learn from and connect with the broad spectrum
Activity theory underlines the importance of interactions of information and communication technologies and
between factors as an important driving force for activity development (ICTD) applications.
system development.
Emergency responders, subjects, analyze the incident facts REFERENCE
before they can share it with the other agencies. While
numerous initiatives have been started to improve information [1] S. Y. Shen and M. J. Shaw, "Managing Coordination
processing, local agencies are typically limited in the systems in Emergency Response Systems with Information
that they can offer to build and manage. To this end, the broad Technologies," in Tenth American Conference on
community of public/private firms and the general public has Information Systems, New York, 2004, pp.
2110-2120.
307
[2] R. Chen, R. Sharman, R. H. Rao, and S. Upadhyaya, [18] J. Hale, "A Layered Communication Architecture for
"Coordination in Emergency Response Management," the Support of Crisis Response," Journal of
Communications of the ACM, vol. 51, pp. 66-73, Management Information Systems, vol. 14, 1997.
2008. [19] D. Alexander, "World Disasters Report 2005: Focus
[3] E. L. Quarantelli, "Problematical Aspects of The on Information in Disasters," 2006.
Information/Communication Revolution for Disaster [20] L. K. Comfort, "Crisis Management in Hindsight:
Planning and Research: Ten Non-Technical Issues Cognition, Communication, Coordination, and
and Questions," Disaster Prevention and Control," Public Administration Review, 2007.
Management, vol. 6, pp. 94-106, 1997. [21] S. Corbacioglu and N. Kapucu, "Organizational
[4] J. Telford, J. Cosgrave, and R. Houghton, "Joint Learning and Self-Adaptation in Dynamic Disaster
Evaluation of the International Response to the Indian Environments," Disasters, vol. 30, pp. 212-233, 2006.
Ocean Tsunami: Synthesis Report," Tsunami [22] N. Martin, "The Asian Tsunami: An Urgent Case for
Evaluation Coalition2006. Improved Government Information Systems and
[5] R. Samarajiva, "Policy Commentary: Mobilizing Management," Disaster Prevention and Management,
Information and Communications Technologies for vol. 16, pp. 188-200, 2007.
Effective Disaster Warning: Lessons from the 2004 [23] N. Kapucu, "Collaborative Emergency Management:
Tsunami," News Media Society, 2005. Better Community Organizing, Better Public
[6] B. Van de Walle and M. Turoff, "Emergency Preparedness and Response," Overseas Development
Response Information Systems: Emerging Trends and Institute, 2008.
Technologies," Communications of the ACM, vol. 50, [24] M. W. Seeger, T. L. Sellnow, and R. R. Ulmer,
pp. 29-31, 2007. Communication and Organizational Crisis. Westport,
[7] M. Turoff, "Past and Future Emergency Response CT: Praeger Publishers, 2003.
Information Systems," Communications of the ACM, [25] J. M. Kendra and T. Wachtendorf, "Elements of
vol. 45, pp. 29-32, 2002. Resilience after the World Trade Center Disaster:
[8] T. X. Bui and S. R. Sankaran, "Design Considerations Reconstituting New Year City's Emergency
for A Virtual Information Center for Humanitarian Operations Centre," Disaster, vol. 27, pp. 37-53,
Assistance/Disaster Relief Using Workflow 2007.
Modeling," Decision Support Systems, vol. 31, 2001. [26] S. Jain, A. Malik, and R. D'Souza, "26 Countries
[9] Y. Engestrom, "Activity Theory and Individual and Alerted in 15 Minutes, India not One of Them," in
Social Transformation," in Perspectives on Activity Indian Express Mumbai: Indian Express Newspapers
Theory, R. M. a. R. P. Engeström, Ed. Cambridge, Ltd, 2004.
UK: Cambridge University Press, 1999, pp. 19-38. [27] T. A. o. M. (TAM), "Report on the Failure of the
[10] Y. Engestrom, Learning by Expanding: An Telecommunications Services as a result of the Asian
Activity-Theoretical Approach to Developmental Tsunami on 26 December 2004,"
Research. Helsinki: Orienta-Konsultit, 1987. Telecommunications Authority of Maldives
[11] L. Uden and A. Kumaresan, "Usable Collaborative (TAM)2005.
Email Requirements Using Activity Theory," [28] W. Michalowski, S. Rubin, R. Slowinski, and S. Wilk,
Informatics, vol. 31, pp. 71-83, 2007. "Mobile Clinical Support System for Pediatric
[12] V. Kaptelinin and B. A. Nardi, Acting with Emergencies," Decision Support Systems, vol. 36,
Technology: Activity Theory and Interaction Design. 2003.
Cambridge, MA: MIT Press, 2006. [29] D. C. Pattie, S. Dannenmann, and Y. Hagiwara,
[13] A. Chaudhury, D. N. Mallick, and H. R. Rao, "Web "Evaluation and Strengthening of Early Warning
Channels in E-Commerce," Communications of the Systems in Countries Affected by the 26 December
ACM, vol. 44, p. 99, 2001. Tsunami," in Information Systems for Crisis Response
[14] L. S. Vygotsky, Mind and Society. Cambridge, MA: and Management, 2008.
Harvard University Press, 1978. [30] C. Ryu, Y. J. Kim, A. Cahudhury, and H. R. Rao,
[15] O. W. Bertelsen and S. Bodker, "Activity Theory," in "Knowledge Acquisition via Three Learning
HCI Models Theories, and Frameworks: Toward A Processes in Enterprise Information Portals," MIS
Multidisciplinary Science, J. M. Caroll, Ed. San Quarterly, vol. 29, pp. 245-278, 2005.
Francisco: Morgan Kaufmann, 2003, pp. 291-324. [31] T. S. Raghu, A. Chaudhury, and H. R. Rao, "Business
[16] R. Chen, R. Sharman, N. Chakravarti, H. R. Rao, and Process Change: A Coordination Mechanism
S. Upadhyaya, "Emergency Response Information Approach," Knowledge and Process Management,
System Interoperability: Development of Chemical vol. 5, pp. 87-98, 1998.
Incident Response Data Model," Journal of the [32] I. Nonaka, "A Dynamic Theory of Organizational
Association for Information Systems, vol. 9, 2008. Knowledge Creation," Organization Science, vol. 5,
[17] D. P. Biros, M. Daly, and G. Gunsch, "The Influence pp. 14-37, 1994.
of Task Load and Automation Trust on Deception [33] J. C. Spender, "Strategy Theorizing: Expanding the
Detection," Group Decision and Negotiation, vol. 13, Agenda," in Advances in Strategic Management, P.
2004.
308
Shrivastava, A. Huff, and H. Dutton, Eds. Greenwich, Coordination: Lessons from the World Trade Center
CT: JAI Press, 1992, pp. 3-32. Response," Center for Technology in Government,
[34] R. Chen, R. Sharman, H. R. Rao, and S. Upadhyaya, 2004.
"Design Principles for Critical Incident Response [49] R. Chen, R. Sharman, H. R. Rao, and S. Upadhyaya,
Systems," Information Systems and E-Business "Design Principles of Coordinated Multi-incident
Management, vol. 5, 2007. Emergency Response Systems," in IEEE
[35] J. Douglas, Investigative Social Research Newbury International Conference on Intelligence and Security
Park, CA: SAGE Publications, 1976. Informatics 2005, Atlanta, GA, 2005.
[36] E. J. Webb, D. T. Campbell, R. D. Schwartz, and L. [50] J. J. Carafano, "Preparing Responders to Respond:
Sechrest, Unobtrusive Measures. Chicago, IL: The Challenges to Emergency Preparedness in the
Rand-McNally, 1965. 21st Century," Heritage Lectures, vol. 812, 2003.
[37] H. Chen, D. Zeng, H. Atabakhsh, W. Wyzga, and J. [51] S. F. Midkiff and C. W. Bostian, "Rapidly-Deployable
Schroeder, "COPLINK: Managing Law Enforcement Broadband Wireless Networksfor Disaster and
Data and Knowledge," Communication of the ACM, Emergency Response," in The First IEEE Workshop
vol. 46, pp. 28-34, 2003. on Disaster Recovery Networks (DIREN '02) New
[38] H. Chen, J. Schroeder, R. Hauck, H. Ridgeway, H. York, NY, 2002.
Atabakhsh, H. Gupta, C. Boarman, R. K, and A. [52] T. X. Bui, "Towards a Theory of Shared Mental
Clements, "COPLINK Connect: Information and Model in CSCW," in 1992 Workshop on CSCW, 1992.
Knowledge Management for Law Enforcement," [53] D. Berge, The First 24 Hours: A Comprehensive
Decision Support Systems, vol. 34, pp. 271-285, 2003. Guide to Successful Crisis Communications.
[39] M. Turoff, M. Chumer, B. Van de Walle, and X. Yao, Cambridge, MA: Basil Blackwell, 1990.
"The Design of A Dynamic Emergency Response [54] B. S. Manoj and A. H. Baker, "Communication
Management Information System (DERMIS)," Challenges in Emergency Response,"
Journal of Information Technology Theory and Communications of the ACM, vol. 50, pp. 51-53,
Application, vol. 5, 2004. 2007.
[40] K. E. Weick, "Enacted Sensemaking in Crisis [55] NCP, "Priority Telecommunications Services,"
Situations," Journal of Management Studies, vol. 25, Washington, DC: Department of Homeland Security,
1988. National Communications System, 2006.
[41] K. E. Weick, K. M. Sutcliffe, and D. Obstfeld, [56] M. Turoff, S. R. Hiltz, M. Bieber, B. Whitworth, and
"Organizing for High Reliability: Processes of J. Fjermestad, "Computer Mediated Communications
Collective Mindfulness," in Research in for Group Support: Past and Future," in HCI in the
Organizational Behavior. vol. 21, B. M. S. a. L. L. New Mellennium, J. Carroll, Ed.: Addison Wesley,
Cummings, Ed. Greenwich, CT: JAI Press, 1999, pp. 2001.
81-123. [57] H. Chen, F. Y. Wang, and D. Zeng, "Intelligence and
[42] J. Sweller, "Cognitive Load During Problem Solving: Security Informatics for Homeland Security:
Effects on Learning," Cognitive Science, vol. 12, Information, Communication, and Transportation,"
1988. IEEE Transactions on Intelligent Transportation
[43] B. Britton and A. Tesser, "Effects of Prior Knowledge Systems, vol. 5, pp. 329-341, 2004.
on Use of Cognitive Capacity in Three Complex [58] t. Congress, "Health Insurance Portability and
Cognitive Tasks," Journal of Verbal Learning and Accountability Act of 1996," in Public Law 104-191,
Verbal Behaviour, vol. 21, 1982. T. Congress, Ed., 1996.
[44] Y. J. Kim, R. Kishore, and G. L. Sanders, "From DQ [59] BJA-DOJ, "NIEM Executive Briefing," D. o. Justice,
to EQ: Understanding Data Quality in the Context of Ed., 2007.
E-Business Systems," Decision Support Systems, vol. [60] J. Fedorowicz, U. J. Gelinas, J. L. Gogan, M. Howard,
forthcoming, 2003. M. L. Markus, C. Usoff, and R. Vidgen, "Modeling
[45] R. Chen, "Consumer's Initial Acceptance of Physical Barriers to Interorganizational System
E-Commerce Website: A Contingency Approach," in Implementation Success," International Journal of
AMCIS 2007, Keystone, CO, 2007. Information Technology and Management, vol.
[46] P. Anderson, "Information Technology: Where is It in forthcoming, 2007.
the Coordination of Emergency Services," in Asia [61] J. L. Gogan, C. B. Williams, and J. Fedorowicz, "Fatal
Pacific Police Technology Conference, 1991. Flaws in Information Sharing," International Journal
[47] G. E. G. Beroggi and W. A. Wallace, "Real-Time of Technology, Knowledge and Society, vol. 1, pp.
Decision Support for Emergency Management: An 93-100, 2005.
Integration of Advanced Computer and [62] C. B. Williams, J. L. Gogan, and J. Fedorowicz,
Communications Technology," Journal of "Public Safety and Cross-Boundary Data Sharing:
Contingencies and Crisis Management, vol. 3, p. 18, Lessons from the CapWIN Project," IEEE Computer,
1995. vol. 38, p. 28, 2005.
[48] S. S. Dawes, T. Birkland, G. K. Tayi, and C. A.
Schneider, "Information, Technology, and
309
[63] D. Aylward and E. Jones, "Data Interoperability:
Sharing Information for a Safer America." vol. 2006:
COMCARE Emergency Response Alliance, 2006.
[64] DHS, "Fact Sheet: Achieving First Responder
Communications Interoperability," in Department of
Homeland Security Press Washington, DC, 2005.
[65] D. Frale, "Emergency Interoperability Consortium
Announces Agreement with Department of Homeland
Security to Promote Data Sharing During
Emergencies," in PrimeZone Media Network, 2005.
[66] T. Harrison, J. R. Gil-Garcia, T. A. Pardo, and T.
Fiona, "Learning about Interoperability for
Emergency Response: Geographic Information
Technologies and the World Trade Center Crisis," in
The Thirty-Ninth Annual Hawaii International
Conference on System Sciences, Hawaii, 2006.
[67] DOS, "National Incident Management System,"
Department of Homeland Security, Washington,
DC2004.
[68] IBM, "World Community Grid," 2008.
[69] P. Currion, C. De Silva, and B. Van de Walle, "Open
Source Software for Disaster Management,"
Communications of ACM, vol. 50, pp. 61-65, 2007.
310
Abstract – In the present information society, technical Rehana Ghadially is with Department of Humanities and
education has acquired immense importance as the Social Sciences, Indian Institute of Technology
Information and Communication Technologies (ICTs) hold Bombay, India (e-mail: rehana@hss.iitb.ac.in) .
potential for bridging socio-economic divides and
empowering the marginalized such as women and minority
groups. This paper explores the access and use of
computer/Internet and examines how these affect
empowerment levels among young men and women. This Even as India is a leading destination for
paper considers the psychological, social, educational and outsourced IT and IT enabled work that promises
economic benefits following from computer education and
employment opportunities, there are several divides within
usage of computer and Internet technology. Data was
collected from 155 young girls (N=82) and boys (N=73) from
the country that leave the disadvantaged groups out of the
three computer training centers in Mumbai. Statistical tests development radar. The digital divide operates across
such as t-tests, 2-way ANOVA and chi-squares were geographic location (rural-urban), class (rich-poor), gender
computed to compare male and female subjects on (male-female) and language (English-vernacular) [1].
empowerment and ownership/access and use. The figures for While development effort in India has considered one or
ownership and home Internet connection were low for the the other of these factors, they overlook the marginalized
entire sample. Computer training centre and cyber café are status of minorities such as Muslims. The Muslim
important points of access for females and males respectively. minority is disadvantaged both educationally and
Further, it is found that young women report higher gains
economically, thus driving them off the information
from computer learning and technology use, hence, computer
education can be a key gender equalizer. In light of the highway [2, 3, 4]. In addition, they face societal barriers
above, policy measures to widen access and provide such as limited awareness and discrimination [5]. Cost and
subsidized training are suggested. language barriers compound the constraints further
limiting the extent of computer and Internet use. Women
Index Terms – access, computers/Internet, empowerment, are doubly disadvantaged as they face an additional gender
gender barrier [6]. A coverage of ICT and India reports that ICT
skills and access provide an opportunity to leapfrog and
I. INTRODUCTION bridge the disparities in information, education and income
[7]. Hence, we postulate that equipping the young
Today’s world is shaped by availability of members of the Muslim community, especially its women
information and ability to communicate, both of which are with the technical skills and access to the technology will
enabled through the rapid expansion of Internet and result in a more balanced development of the country.
extensive use of computers. While there have been
developmental efforts to increase the access and use of In a developing country like India, ownership of
Information and Communication Technologies (ICTs) in personal computers and Internet penetration is limited at
general, there is still a marked difference in the technology 2.5 and 4.2 per cent respectively [8]. Most users depend on
access across different groups, resulting in the digital shared access points, predominantly at educational
divide. Access considers equality of access to resources, in institutions and workplaces. Three-fourths of Internet
this case to computer and Internet technology. The uneven users in the country depend on community access points
distribution of ICTs across the world as well as within such as cyber café, telecenter, and information kiosk, with
societies gives rise to a digital divide; that results into two cyber cafes as the predominant public access model in
groups; namely, the information rich, those who have urban areas [9]. It is estimated that there are around 50,000
access to abundant information and the information poor, cyber cafes used by almost 70 percent of Internet users
who lack such an access. [10]. However, the cost of surfing Internet at cyber cafes
restricts the usage to the middle and high income groups
Manuscript received September 22, 2008. [11]. In addition, research indicates that even when access
Farida Khan is with Department of Computer Science and is available, men are advantaged over women in the extent
Engineering, Indian Institute of Technology Bombay, of computer usage [12]. In light of the above discussion on
India (phone: 91-022-25764966, e-mail: disparity in the access and use of the technology, the
farida@iitb.ac.in). present study seeks to explore the different access points
for the two sexes and assess gender differences in
computer ownership, Internet connection at home, points
of access and extent of computer/Internet use.
1
311
Education especially technical education is sidelining the other facets of empowerment. This paper
instrumental in expanding capacities and improving attempts to provide a more complete view of
employment opportunities, thus resulting not only in empowerment by considering the psychological, social,
economic but also personal and social empowerment such educational and economic gains flowing from computer
as enhancing confidence levels and social status [13]. learning, technology access and usage.
Considering that computer skill is of immense importance
in the technology driven society, this paper considers the
gains resulting from having acquired basic computer
education. Unlike the much emphasized economic gains,
Objectives of the study
the present paper adopts a multidimensional perspective
and considers four different facets of empowerment ---
psychological, social, educational and economic. In • To explore gender differences, if any, in
addition, it explores the access and use of computer and computer ownership, Internet connection, points
Internet among men and women and examines how these of access and extent of computer/Internet use.
affect the different dimensions and levels of • To assess the empowerment flowing from
empowerment. computer education for males and females.
• To explore differences, if any, in empowerment
In recent times, research in gender and among males and females with different levels of
development studies has focused on the concept of computer/Internet use.
empowerment [14]. According to their theoretical
background and parameters of assessment, empowerment II. METHODOLOGY
is understood by different researchers in different ways.
Those from the field of sociology focus on empowerment Participants of the study comprised of 155
of the marginalized groups or society as a whole, and trainees (82 females and 73 males) from Mumbai, enrolled
assess it in terms of political activism, advocacy and in a one year diploma in computer applications and multi-
networking. Those from the field of economics focus on lingual desktop publishing offered by the Ministry of
empowerment of the individual, group or economy as a Human Resource Development, under the National
whole and assess it in terms of improved income levels. Council for Promotion of Urdu language (NCPUL)
Experts in gender studies consider women’s empowerment scheme 1 . The course requires the trainees to complete a
per se and elaborate on how altered gender relations would diploma in Urdu language along with the computer course.
benefit women and assess it in terms of increased income While theoretically open to anyone interested in the Urdu
opportunities, greater participation in decision making, language, it is particularly the low income Muslim young
social networking and advocacy. On the other hand, those who are drawn to it because it is highly subsidized. Hence,
from psychology and management studies concentrate on a large section of Muslim boys and girls who are well
psychological gains and characterize empowerment in versed with Urdu language enroll for the course.
terms of personality (self-efficacy, internal locus of Computation from the demographic profile revealed the
control), motivational (feeling enabled) and cognitive following information about the sample. The average age
(meaning, competence, self-determination, impact) factors of the subjects was 21.34 years (males = 20.46 years;
[15, 16, 17]. For the present study, a cognitive perspective females = 22.31 years) and their educational level varied
of psychological empowerment is adopted. Hence, from higher secondary (45.93 per cent) to undergraduates
psychological empowerment is construed as --- (22.67 per cent) and college graduates (31.40 per cent).
meaningfulness (commitment to the task), Competence Majority of the subjects (69.77 per cent) studied in Urdu
(self-confidence), Self-determination (choice) and impact medium schools, while less than one-third (26.74 per cent)
(making a difference). had English as the medium of instruction.
In the present information society, computers and Majority (73.2%) of the sample belonged to low
Internet are the pervasive technologies that hold potential (monthly family income of less than Rupees 5000/-; US 1$
for equalizing the disparities in information, education and = 48.70 Rupees) and low middle income group (monthly
income [18]. With the growth of Information Technology family income of Rupees 5000/- - 10,000/-). 47.67%
enabled services (ITES) in India, a plethora of computer fathers were running their own business, 33.14% engaged
related jobs have opened up. Comparison of work- in service and 19.19% had retired. 87.2% mothers were
participation rate of the ITES with other service sectors homemakers, while 12.79% engaged in home-based work
indicates that the field is more accommodating of women
[19]. Research has highlighted the personal (self
confidence) and social gains (status, increased 1
For a detailed description of the scheme, refer Rehana
connectivity) that come from ICT education [20, 21]. Ghadially and Farida Umrani (2004) IT Education: Initiatives among
However, most of these studies are focused on women; Mumbai Muslims”, i4D Information for Development, February.
hence a gender comparison does not follow. In addition,
the literature is skewed towards the economic benefits thus
312
or service. Participants report low educational levels of Permission was ought from the heads of three
parents as nearly one-fourth (24%) of the fathers and half NCPUL computer training institutes, to conduct the study.
(44%) of the mothers had completed only primary Data was collected across twelve visits by the senior
education. author from a group of 10-15 subjects, on completion of
ten months of computer training.
The tools for data collection included
standardized psychological scales to measure
empowerment, a questionnaire to gauge access and use of
computer/Internet and a demographic profile. Computer Data Analysis
ownership was estimated by one question each for
computer and Internet connection at home. Point of access
Chi-squares and independent t-tests and were computed to
was assessed by one question where subjects selected the compare male and female subjects on ownership/access,
places of access from a list provided to them. Extent of
use and empowerment. Two 2-way ANOVAs were
computer and Internet use was estimated in terms of the
computed to gauge the influence of gender and differential
number of hours subjects spend on a computer/Internet in level of computer and Internet use on the different
a week. In addition, a personal profile was estimated to get
dimensions of empowerment.
a few demographic details.
III. RESULTS
Psychological empowerment was assessed by an
adapted version of Spreitzer’s 12-item scale [17]. It has
four subscales – meaningfulness (value of the task), The study gathered information on computer
competence (efficacy), self determination (choice) and ownership, Internet connection at home and points of
impact (difference made). A sample item from the access. It was found that 14.8 per cent subjects had a
psychological empowerment scale is, ‘Learning computers computer at home. This included 12.2 per cent females
has given me considerable opportunity for independence and 17.8 per cent males (χ2 =1.38, ns). 6.0 per cent of the
and freedom in how I do my daily tasks.’ The reliability sample (5.0 per cent females and 6.8 per cent males, χ2
co-efficients (internal consistency) range between 0.61- =1.58, ns) had Internet connection a home. These findings
0.72. Only overall empowerment score is considered for indicate that although the difference between male and
analysis in the present study. For social, educational and female subjects’ computer ownership and Internet
economic empowerment, the scales were designed by the connection is not significant, there is a trend in favour of
researchers. Social empowerment was measured by a 4- males.
item scale that tapped four aspects --- increased status,
social comparison, being with the times and keeping in The different places from where subjects
touch with friend and/or relatives. A sample item from accessed the technology beyond class hours at the training
social empowerment scale is, ‘Computer education makes centre were explored. The computer institute where the
me feel more up to date and current.’ Educational subjects got their training and cyber cafe emerged as the
empowerment was measured by a 3-item scale that most important points of access followed by home
covered three aspects, information source on (own/relative/friend/neighbor) and workplace. A small
courses/colleges/universities, accessing online resources group of subjects did not have access to computers beyond
and preparation of class reports and presentations. A class hours. A gendered view of these findings indicates
sample item is, ‘Learning computers has opened new ways that females and males differ significantly on points of
to find information about different access (χ2 = 20.19, p<0.001). While the computer training
courses/colleges/universities for me.’ Economic institute emerged as a major access point for females (χ2 =
empowerment was assessed by a five-item scale that 10.88, p<0.001; females 57.3 per cent and males 27.4 per
tapped the following five areas --- opening of new job cent), cyber café is more important for males (χ2 = 4.26,
opportunities, application in family business, starting a p<0.05; females, 12.2 per cent, males, 38.4 per cent). The
computer related entrepreneurial venture, earning from other points of access --- home and workplace --- did not
home and managing family responsibilities with a job. A show any significant gender difference. 8.5 per cent
sample item of economic empowerment scale is ‘Learning females and 4.1 per cent males did not access computers
computers has opened new job opportunities for me.’ The beyond class hours (Table 1). The findings on the points of
reliability co-efficients (Cronbach’s alpha) of the social, access indicate that computer training institutes and cyber
educational and economic empowerment scales were 0.66, cafes are the major access points for females and males
0.72, and 0.77 respectively. Each of the empowerment respectively. Though not significant, the trend indicates
scale asked the subjects to indicate the extent to which that more males have access to home computers and twice
they agree that learning computers has resulted in the as many females as compared to males have no access to a
following outcomes on a 6-point rating scale ranging from computer at all.
1 = disagree to 6 = highly agree. The scales were pilot
tested and in line with participants’ feedback, Hindi
translation was provided along with the English version.
313
χ2
df=2 7.94** 32.15***
**
p<0.01, * p<0.05
TABLE 1
Points of Access to Computers for Males and
Females
Points of Total Males Females Gender A similar classification was done with regard to
Access (T) (M) (F) χ2 Internet use. More than half of the sample did not use
df =1 Internet at all, one-third surfed Internet for less than 5 hours
Computer 43.2 % 27.4 57.3 10.88*** and a small number surfed for more than 5 hours in a week.
institute Female and male subjects differed significantly with regard
Cyber café 24.5 38.4 12.2 4.26* to the extent of Internet use (χ2 = 32.15, df = 2, p<0.001).
The two sexes differed significantly on each of the three
Home 23.2 27.4 19.5 0.44 categories of Internet use --- no use (χ2 = 17.39, df = 1, p <
0.01), low use (χ2 = 11.08, df=1, p < 0.01) and high use (χ2 =
Workplace 2.6 2.7 2.4 0 4.45, df = 1, p < 0.01). 80.5 per cent females reported not
using Internet at all, while the corresponding figure for males
Do not 6.5 4.1 8.5 1.6 was 35.6 per cent. 17.1 per cent females and 52.1 per cent
access males report low Internet use; while 2.4 per cent females and
χ2 20.19*** 12.3 per cent males fall in the high Internet use category
df =4 (Table 2). Thus, Internet use is significantly low for females
as compared to males. Comparing computer and Internet
*** usage of the two sexes indicated that the disparity in
p<0.005, *p<0.05 technology usage is more marked in case of Internet than the
computer, thus pushing the women off the information
In addition to the above, the extent of computer highway.
and Internet use was also measured. On the basis of the
number of hours in a week subjects worked on computers, In addition to the above, the study estimated the
beyond the class hours at the institute, they were classified levels of empowerment of male and female trainees. An
into three groups --- no use (not at all), low use (less than 5 estimate of composite empowerment was calculated by
hours) and high use (5 hours or more). More than one-fourth adding scores on the psychological, social, educational and
of the sample did not use computers at all, two-fifth reported economic empowerment dimensions and dividing it by four.
low use and less than one-third were in the high use To test whether the male and female subjects differ
category. The gender view indicates that female and male significantly on empowerment, independent sample t-tests
subjects differ significantly with regard to the extent of were computed. The two sexes did not differ significantly on
computer use. The number of females and males in no any of the dimensions of empowerment --- composite (t = -
computer use category differed significantly (χ2 = 7.71, df = 1.40, df =153), psychological (t = -0.93, df =153), social (t = -
1, p<0.01). Even when the women use computers beyond 1.10, df =153), educational (t = -1.53, df =153) and economic
class hours, the extent of use is limited as compared to men. (t = -1.27, df =153). Although no significant differences were
However, this difference is not significant. Thus, the obtained on the different dimensions of empowerment, the
findings provide evidence for the disadvantage of women trend in all cases was for women to experience more
participants (Table 2). empowerment than men. Basic computer education results in
relatively higher composite empowerment for females (mean
TABLE 2 = 30.58, SD = 3.82) as compared to males (mean = 29.75, SD
Extent of Computer and Internet Use for Males = 3.55). Similarly, females reported higher psychological
and Females (mean = 58.83, SD = 9.46) and social (mean = 21.44, SD =
Computer Use χ2 Internet Use χ2 2.99) empowerment than males (mean = 57.51, SD = 8.05;
df=1 df=1 mean = 20.90, SD = 3.05, respectively). Following the trend,
T M F T M F females reported relatively higher educational (mean = 16.17,
No 27.1 16.5 36.6 7.71* 59.4 35.6 80.5 17.39** SD = 2.83) and economic (mean = 25.89, SD = 4.33)
use % * empowerment as compared to males (mean = 15.53, SD =
2.27; mean = 25.04, SD = 3.93 respectively).
Low 41.9 47.9 36.6 0.38 33.5 52.1 17.1 11.08**
use
314
The interaction effect of gender and The trend of the scores indicated that females
computer/Internet use on the empowerment was assessed by with computer use reported higher empowerment than males
means of a two-way ANOVA (Tables 3 and 4). The combined with computer use as well as females and males with no
effect of gender and computer use resulted in a significant computer use (Table 5).
difference in the composite [F (1, 151) = 8.02, p<0.01],
psychological [F (1, 151) = 4.14, p<0.05], educational [F (1, 151) =
9.62, p<0.015] and economic [F (1, 151) = 6.12, p<0.01]
empowerment. In other words, females reporting computer
use obtained significantly higher scores on all aspects of
empowerment as compared to males.
TABLE 3
Results of Two-Way ANOVA Assessing the TABLE 5
Effect of Gender and Computer Use on Empowerment
Dependent Source SS df MSS F t-test Scores of Male and Female Ss with Different Levels
Variable of Computer Use on Empowerment
Composite Gender* 101.17 1 101.17 8.02**
empowerment Computer use Emp Computer Use Internet Use
(CE) Error 1905.34 151 12.62 M F t M F t
Psychological Gender* 306.63 1 306.63 4.14* df=111 df=61
empowerment Computer use CE 29.61 31.71 - 29.63 31.70 -2.09*
(PE) Error 11189.03 151 3.22***
Social Gender* 25.04 1 25.04 2.82 PE 57.39 61.07 -2.43* 57.23 61.19 -1.64
empowerment Computer use
(SE) Error 1340.46 151 8.87 SE 20.85 22.04 -2.22* 20.98 21.56 -0.70
Educational Gender* 58.17 1 58.17 9.62**
empowerment Computer use EDE 15.43 17.02 - 15.19 17.00 -2.82**
(EDE) Error 913.115 151 6.05 4.26***
Economic Gender* 101.84 1 101.84 6.12** ECE 24.75 26.69 -2.61** 25.11 27.06 -2.03*
empowerment Computer use
(ECE) Error 2512.89 151 16.64
**p<0.01;*p<0.05 (one-tailed)
**
p<0.01, *p<0.05
A similar computation was done with respect to
TABLE 4 gender and Internet use. The interaction effect of gender and
Results of Two-Way ANOVA Assessing the Effect of Internet use resulted in a significant difference in educational
Gender and Internet Use on Empowerment empowerment [F (1, 151) = 4.40, p<0.05)]. The trend of the
mean scores indicates that females with Internet use reported
Dependent Source SS df MSS F higher empowerment than males with Internet use as well as
Variable females and males with no Internet use (Table 5). The mean
CE Gender* 21.69 1 21.69 1.59 scores of females (mean = 31.70, SD = 2.91) and males
Internet use (mean = 29.63, SD = 3.59) on composite empowerment were
Error 2065.48 151 13.68 significantly different (t = -3.22, p<0.005). The mean scores
PE Gender* 99.42 1 99.42 1.27 on psychological, social, educational and economic
Internet use empowerment of the two sexes reporting computer use were
found to be significantly different (t = -2.43; p<0.05, -2.22;
Error 11791.48 151 78.09
p<0.05, -4.26 p<0.005, -2.61; p<0.01). Females reporting
SE Gender* 0.02 1 0.02 0.002
computer use was significantly higher than males on each of
Internet
Error 1391.486 151 9.21 the dimensions (Table 5). Thus, technology use provides
EDE Gender* 28.90 1 28.90 4.40* maximum benefits for women.
Internet use
Error 990.60 151 6.56 IV. DISCUSSION
ECE Gender* 11.80 1 11.80 0.68
Internet A. Computer Ownership/Internet Connection, Access,
Error 2607.009 151 17.26 1.27 Computer/Internet Use
*
p<0.05
315
The study assessed computer access and usage of the had primary education or were illiterate), parents may not
technology in terms of family ownership of computer, be aware of the potential benefits of Internet.
having an Internet connection at home and the various
points of access to computer/ Internet. Computer Besides private access, the different points of
ownership and home Internet connection for the sample is access were explored. It was found that significantly
considerably lower as compared to ICT penetration in higher number of females access computers at the training
Mumbai where the study was conducted. The city ranks institute. Access at the institute does not require cost
first in PC penetration with 32 per cent household owning investment, but does not provide Internet access either.
a PC [22]. Research indicates that Mumbai and Delhi Further, the women will be left vying for access after
together constitute close to 70 per cent of the total installed completion of the course. Cyber café is significantly more
computers in Indian homes and 46 per cent of the important point of access for males than females. This
computer owners have an Internet connection [12]. Thus, difference may be explained by three reasons --- nature of
even though the figures of computer ownership and use, cost and socio-cultural attitudes. Males seek to use the
Internet connection for the sample under study are better Internet, which the computer training centre does not
than the national average (2.5 and 4.2 % respectively [8]), provide. In urban India, cyber cafes are the predominant
they are much lower than the figures for Mumbai city. The public access model for Internet usage. Even though the
sample consisting of college students and fresh graduates, cost of surfing in the cyber cafes has reduced from Rupees
have no personal disposable income to purchase a 50 to Rupees 20 per hour in the last decade, this is still out
computer. Besides 80 per cent of the subjects in the study of reach of the low income groups. Young girls may
belonged to low income group and hence cannot afford a hesitate to use family resources for surfing the Internet,
personal computer or Internet connection. In addition, due thus curtailing their visit to cyber cafés. Third, socio-
to low education levels of the parents, there is lack of cultural restrictions add to the disadvantage of young
awareness about the potential of the Internet. Hence, even women as families impose sanctions against cyber cafes as
when they purchase a computer to support education of these are characterized as boys’ hangouts and lack female
their children, less than half of the sample has an Internet attendants [26]. Besides, families may be more willing to
connection. Besides, technology access at home is provide financial support to the boys for accessing
adversely affected by space constraints. Households in technology, thus placing them at an advantage as far as
Mumbai utilize an average of 2.9 square meter of floor access at cyber café is concerned.
space per person, one of the lowest in the world [23].
Given the low socio-economic levels of the sample and the Workplace does not emerge as an important
pigeon-hole like houses of Muslim ethnic enclaves of access point as the sample consists predominantly of
Mumbai, the disadvantage for the young members of the college students and fresh graduates. Only six participants
community with regard to computer ownership and (4 men, 2 women) were working in part time jobs such as
Internet connection follows. giving home tuitions, office assistant, etc. Thus, despite
possessing the technical skills, the women from the current
The two sexes did not differ significantly on sample have lower access to the technology as compared
computer ownership and home Internet connection. This is to their male counterparts. Limited Internet connection at
because majority of the sample belong to low income home and cyber café usage seriously disadvantages the
groups and are financially dependent on their families. women as research indicates that Internet provides women
Hence, both boys and girls were affected by economic access to information, greater participation in decision-
constraints. However, the trend indicates that males are making and networking, thus empowering them [18].
slightly advantaged as compared to females. There is lack
of gender segregated data on computer ownership, hence Subjects were asked the number of hours they
stretching a point, a study on difference in family used computers/ Internet in a week, besides the regular
computer usage is considered. A study on computer usage
class hours at the training institute. On the basis of their
in India indicates that 82 per cent of males compared to responses, the usage was classified into three levels --- no-
only 16 per cent of females are the primary users of family use, low-use and high-use a week. A similar classification
computer [12]. Besides, Indian families tend to make
was done with regard to Internet use. Two issues are
higher investments in education of the boys compared to apparent with regard to computer/ Internet use; one, the
girls due to different role expectations [24]. As computer disparity across technology --- computer v/s Internet; and
is identified as an educational tool, it is likely that families
gender --- males v/s females. While a quarter of the
would purchase computers for their sons rather than subjects fell in no-use category for computer use, with
daughters. In addition, males identify the home computer regard to the Internet, this rose to 60 per cent. One-third of
as personally owned; while females identify it as
the subjects reported high computer use but only 7.1 per
belonging to the head of the family [25]. Cost barrier is cent report high Internet use. These figures bear testimony
more prominent in case of Internet connection as
to the fact that computer use and Internet use are distinct
compared to computer ownership, in terms of recurring categories and merging them into a single unit could be
monthly payments. In addition, due to low education misleading. While using computers requires only access to
levels (one-fourth of the fathers and half of the mothers
the machine, for Internet use the cost gets multi-fold due
316
to investments in modems, optical fibers, charges of paragraphs.
Internet service providers and computer maintenance. This
may explain the lack of Internet usage. Although the two sexes do not differ significantly on all
measures of empowerment, the trend in all cases is for
The other reasons for limited Internet use are the females to be more empowered than males. This is
non-availability of subsidized public access points, limited supported by research that indicates that though
English language competency and socio-cultural attitudes. modernization and technological development initially
Though cyber café is the main access point in urban India, disadvantaged women, ICT promises a technological U-
the cost of usage is beyond the reach of low income turn [29]. Findings of this study are strengthened by
groups. Research indicates that one-third of Internet users evidence from a qualitative study of the impact of
in Indian cities have an income above Rupees 25,000/- , computer education that indicates that more women than
while only 4.4 percent of the Internet users have an men experience higher boost in self-confidence and secure
income below Rupees 5000/- [11]. Access at cyber cafes is computer related jobs [30]. It is likely that since South
beyond the reach of low economic groups. As majority of Asian women in general are socio-economically deprived,
the sample are poor, they have little disposable income their expectation from something new will be higher than
that could be used for Internet access. Research indicates males who have better chances of being exposed to these
that there is a strong relationship between the use of the technologies before hand.
Internet and ability to speak English [9, 27]. Surveying a
group of students from Gujarat, it is observed that students Analyses of the participants’ response on
educated in the vernacular medium at school struggled at psychological empowerment indicate that females
college where the medium of instruction is English [28]. expressed slightly higher meaning (mean =14.91, SD
As 70 per cent of the sample in the present study had Urdu =2.76), self-determination (mean = 14.83, SD = 3.18) and
as the medium of instruction in school, their limited competence (mean = 15.11, SD = 2.57) as compared to
English language skills probably affected their Internet males (mean = 14.78, 14.01, 14.48; SD = 2.32, 2.59, 2.58
use. Although considerable web content is now available respectively). In other words, computer education resulted
in a variety of Indian languages; a cursory review of the in creating a sense of new possibilities for involvement,
Urdu websites by the first author in 2007, revealed that independence and self-confidence for women. On the
they focused on literature, poetry and news rather than other hand, males (mean = 14.23, SD = 3.00) reported
education and employment. Hence, tangible gains do not slightly higher impact than females (mean =13.97, SD =
follow for the subjects surfing these websites. 3.28), indicating that computer education provided them
possibilities for enhanced control over their environment.
B. Computer Education, Use and Empowerment Thus, the psychological gains for females and males focus
on micro and macro levels respectively. Women reported
In addition to computer access and use, the study relatively higher gains on all aspects of social
focused on analyzing gender differences in empowerment empowerment --- feeling contemporary, enhanced status,
levels following from technical skills training. It was social comparison, and connectivity --- as compared to
found that basic computer education yields similar males. This is in line with research that women perceive
empowerment gains for males as well as females. There is family gain, communication and social gain as the first,
no research comparing men and women on empowerment second and fifth most important benefit of personal
flowing from computer education hence, the explanations computer use [20]. In addition, the social connectivity
offered are speculative in nature. Lack of gender aspect is supported as women are responsible for
differences on empowerment could be due to four reasons. maintaining family and kin relations [31].
First, the sample consisted of young, college going or
fresh college graduates from a metropolitan locale on the Women report relatively higher educational
verge of entering the job market. The homogenous nature empowerment than men with higher scores on information
of the sample makes for similar sources of empowerment on courses/colleges/universities, accessing online
such as marketable skills. Second, empowerment as journals/books/reports. This is corroborated with the
assessed by the scales focuses on the cognitive level and research that when females use computers, they do so
reflects the subject’s estimate of the possible gains from mainly for educational purposes; unlike males, who use it
computer education. It does not assess the tangible for general purposes or playing games [11]. More than
benefits flowing from technical skills, thus overlooking one-third females have completed college; hence search
constraints in the real world that affect the actual benefits for educational information follows. Males score slightly
of computer education. Third, the item pool of the higher on the use of computers to prepare reports and
different empowerment scales is narrow and specific to presentation, may be because unlike the women, more than
this sample. It is likely that if a more diverse measure was half of the men are in college. Females reported slightly
adopted, the two sexes would indicate differential higher scores on four of the five aspects of economic
empowerment levels. Fourth, men and women score empowerment --- job opportunities, managing family
differentially on the items on the scales, thus canceling out responsibilities with work, earning from home and
the gap. This will be discussed in the following assisting in family business. There is a general perception
317
in urban India that a combination of college degree and avenue to become economically independent. Even though
computer skills enhances the possibility of employment. the advantage of females in social empowerment is not
Since a higher number of women participants had already significant, the trend is in the same direction. Media
completed college, they felt better equipped to find a job. advertisements of personal computers in India, project the
The economic gains for females focus on the potential of women computer user as a modern, westernized
IT for combining family with work, a gain which ensures individual; cues that highlight the social status and
maintenance of traditional values and yet being modernity associated with this technology [36]. A
productive. On the other hand, males reported a higher computer being associated with masculinity, its usage
possibility of setting up personal business. This may be brings the females on an equal social footing, enhancing
because belonging to business families (47%), males may feelings of being contemporary and moving with the
be socialized to start their own business. Besides, families times.
could be more willing to provide them the financial capital
due to future role expectation. The interaction effect of gender and Internet use
was found to result in significant difference on educational
Looking at computer/Internet use and empowerment, with females using Internet obtaining
empowerment, it was found that significantly more highest mean empowerment scores as compared to their
females (more than one-third) fell in the no-use category male counterparts. This is not surprising as research
of computer use, than males (one-fifth). Similarly, indicates that when women used technology, they do so
significantly more females (more than 80 per cent females) for educational purposes unlike males who use it for
than males (one-third) report no Internet use. Although general purposes or playing games [11]. The sample
both computer and Internet use was relatively limited for included higher number of females who had completed
females, they were more disadvantaged with regard to college; hence they are likely to explore opportunities for
Internet use. This is a cause of concern, as most of the continuing education using Internet. Thus, young Muslim
research that make a case of ICT-led empowerment of women derive educational benefits out of using Internet.
women, focus on the Internet technology and its ensuing
benefits like networking, political activism and e- With regard to other aspects of empowerment, it
commerce [14, 32]. Women’s low Internet use in this was found that there was no significant interaction effect
sample is in line with previous findings. In a survey of of gender and Internet use. As far as psychological and
Macau residents, it was found that Internet users were social empowerment is concerned, the trend is in favor of
more likely to be males [33]. In a survey of Indian Internet females. This is in line with previous research that Internet
users, it was found that females start using cyber cafes, a use increases self-confidence, facilitates networking and
common Internet access point in urban India later; use expands income generation capacity [14]. There is limited
them less often and for shorter duration per session [11]. comparative evidence on the differential benefit to males
and females from computer and Internet usage. In light of
Gender and extent of computer/Internet use this, the present research fills an important gap by
significantly influence empowerment levels. Findings providing empirical evidence for the benefits of computer
indicate that with the exception of social empowerment, education and technology usage for empowering the two
the interaction effect of gender and computer use is sexes. However, more research with gender as an analytic
significant for composite and the other three kinds of variable is necessary before a clearer picture emerges.
empowerment --- psychological, education and economic.
The mean empowerment scores of males and females V. CONCLUSION AND POLICY
indicate that females reporting computer use are more RECOMMENDATIONS
empowered than males with computer use as well as
females and males without computer use. Thus, the Five major conclusions can be drawn from the
relevance of computer use for empowerment of females study. First, computer ownership and Internet connection
cannot be overstated. It is in line with the research that is very low for the sample compared to figures for
emphasizes psychosocial, educational and economic Mumbai city. Hence, the minority poor as other urban
benefits of computer access and use for women [34, 35]. A poor are on the wrong side of the digital divide. Second,
study of six projects of the Information for Development computer training centre is the major access point for
Program (infoDev) found that computer skills increased females, while cyber café emerges as the chief access
self-esteem and promoted self-confidence of the young point for males. The primary access model in urban India
women participants [18]. Similarly, a study on novice is cyber café and is underused by women. Related to this,
women computer learners found that computer skills the third conclusion is that when women do have access to
resulted in increased self-esteem and enhanced status [20]. computers and the Internet the extent of usage is limited as
The sample of the present research comprised of young compared to males. Thus, the poor urban minority women
college students; hence, educational benefits of using are off the information highway. Fourth, technology use
computers like preparing class assignments or empowers females more than the males, thus making a
presentations seem relevant. Further, females using strong case for ensuring sustained access and usage. The
computers understand that this technology offers an fifth important conclusion is that computer education is a
318
key gender equalizer as unlike other technologies [1] K. Kenniston, “The Four Digital Divides,” in Bridging
women’s engagement with new technologies open’s new the DigitalDivide: Lessons from India, K. Kenniston
possibilities and empowers them as much as the men. and D.K. Das, Eds. New Delhi, India: Sage, 2003.
[2] Sachar Committee Report, “Social, Economic and
In light of the above conclusions, policy Educational Status of the Muslim Community of India,”
recommendations are suggested. In order to widen access Prime Minister’s High Level Committee, Government
points and encourage technology use amongst those at the of India, 2006.
edge of the information society, community access points [3] A. Razaack and A. Gumber, “Differentials in Human
based on the tele-center model should be set up in urban Development: A Case Study of Empowerment of
locales with public-private partnerships. Government can Muslims”, Programme of Research on Human
provide tax benefits to business houses investing in such Development, New Delhi, India: National Council of
endeavors. Besides, to improve women’s access, cyber Applied Economic Research, 2003.
cafés can reserve special timings, provide separate space
for girls, appoint female support staff and open women [4] A. Shariff and A. Razzack, “Communal Relations and
only cyber cafes. This will help the young women to Social Integration,” in India Social Development
overcome the inhibitions in visiting these places for Report, Council for Social Development. New Delhi,
technology access. In addition, civil society organizations India: Oxford University Press, 2004.
can create awareness about potential benefits of [5] O. Khalidi, “Ethnic Group Recruitment in the Indian
technology use and facilitate higher familial support for Army: The Contrasting Cases of Sikhs, Muslims,
technology use by women. Gurkhas and Others,” Pacific Affairs, vol. 74, no. 4,
pp.
In order to maximize tangible gains from 529-552, 2001.
[6] Z. Hasan,. and R. Menon, “Unequal Citizens: A Study
technical skills the computer training center has to be
diversified. The center can liaison with the local small and of
medium enterprises to provide placement support to its Muslim Women in India”, New Delhi, India: Oxford
University Press, 2004.
trainees. It can serve as an intermediary to secure micro-
credit to its students, with special concessions for girls. [7] R. Heeks, “Social Outsourcing: Creating Livelihoods”,
These finances can be used for purchase of a computer and Information for Development (i4d), September, pp. 17-
19, R. (2006.
setting up computer-related business such as a cyber café.
Further, it can assist the trainees to form a self-help group [8] IMRB “International Market Research Bureau,” 2005.
to pool their resources and secure credit for computer- Available: http://www.imrbint.com
[9] A. M. Haseloff, “Cyber cafés and their Potential as
related income generation. Urban Muslims being largely a
Community Development Tools in India,” The
petty business community, their business spirit may assure
Journal of Community Informatics, vol. 1, no. 3, pp.
sustainability. What is required is the initial thrust from an
53-65, 2005.
external agency.
[10] A. Pasricha, “To Get Online, Most Indians go to
Cyber cafés,” VOA News Service, 2004. Available:
It is likely that the impact of these policies can be http://apiap.blogspot.com/2004_10_01_apiap_archive.htm
delimited due to societal conditions such as discrimination l
and prejudice. Hence, fair disbursement of micro-credit [11] A. Haseloff and R. Ghadially, “Gender, Cyber cafés
and sensitization to overcome discrimination will help to and Internet Access: Bridging the Digital Divide?” in
widen the gains flowing from technology. These measures Urban Women in Contemporary India: A Reader, R.
will enable the Muslim youth to compete on an equal Ghadially, Ed. New Delhi, India: Sage, 2007, pp. 270-
footing and lead to their mainstreaming in India’s 283.
economic growth. [12] A. Venkatesh, “Computers and New Media
Technology in Indian Households: Based on a Study
As far as directions for future research are of Eight Major Cities in India,” Centre for Research
concerned, a more comprehensive study can yield on Information Technology and Organizations, 2000.
representative results. The findings of the study are limited Available:
in generalizability as data was not collected from locations http://www.crito.uci.edu/noah/paper/IndianSurvey1.pdf
other than Mumbai and across socio-economic class. A [13] S. Huyer, “Gender, ICT and Education,” 2003.
comparative sample from other minority groups may Available: http://www.wigsat.org/engenderedICT.pdf
provide better insight into the standing of Muslim youth [14] S. Mitter, “Globalization, ICTs and Economic
on the information highway. Empowerment: A Feminist Critique,”
GenderTechnology and Development, vol. 8, no. 1, pp.
5–29, 2004.
[15] J. A. Conger and R. N. Kanungo, “The Empowerment
REFERENCES Process: Integrating Theory and Practice,” Academy
of
Management Review, vol. 13, pp. 471–482, 1988.
319
[16] D.L. Corsun and C.A. Enz, “Predicting Psychological [28] V. Ramanathan, “Ambiguities about English:
Empowerment among Service Workers: The Effect of Ideologies and Critical Practice in Vernacular-
Support-Based Relationships” Human Relations, vol. medium College Classrooms in Gujarat, India,”
52, no. 2, pp. 205-224, 1999. Journal of Language, Identity, and Education, vol. 4,
[17] G. M Sprietzer, “Psychological Empowerment in the no. 1, pp. 45–65, 2005.
Workplace: Dimensions, Measurement and [29] A. Goyal, “ICTs and the Technological U-turn for
Validation,” Academy of Management Journal, vol. Women,” in Urban Women in Contemporary India: A
38, pp. 1442– 1465, 1995. Reader, R. Ghadially, Ed. New Delhi, India: Sage,
[18] N. Hafkin, “Are ICTs Gender Neutral? A Gendered 2007, pp. 255 – 269.
Analysis of Six Case-studies of Multi-donor ICT [30] F. Umrani, “Computer Adoption among Mumbai
Projects,” Background paper to UN/INSTRAW Virtual Muslim Youth: Empowerment and Impact
Seminar Series on Gender and ICTs, 2002. Assessment”, Unpublished Dissertation, Indian
Available:http://www.un- Institute of Technology Bombay: India, 2007.
instraw.org/en/docs/gender_and_ict/Hafkin.pdf. [31] R. Ghadially, “Women’s Friendship,” in
[19] G. Kelkar, “Women in the Digital Era: Gender and Encyclopedia of Women and Islamic Cultures, S.
Culture,” Paper presented at the International Joseph, A. Najmabadi, J. Peteet, J. Siapno and
Conference on Women in the Digital Era: J.Smith, Eds. Netherlands: Leiden, EWIC, pp. 194-
Opportunities and Challenges. Organized by 195, 2003.
Annamalai University and Software Technology Parks [32] S. Huyer and M. Carr, “Information and
of India, at Chidambaram, India, 2004. Communication Technologies: A Priority for
[20] F. Umrani and R. Ghadially, “Empowering Women Women,” Gender, Technology and Development, vol.
through ICT Education: Facilitating Computer 6, no. 1, pp. 85-100, 2002.
Education,” Gender, Technology and Development, [33] W. H. Cheung, “Internet Adoption in Macao,”
vol. 7, no. 3, pp. 359-377, 2003. Journal of Computer Mediated Communications, vol.
[21] N. Hafkin and N. Taggart, “Gender, Information 7, no. 2, 2001. Available:
Technology, and Developing Countries: An Analytic http://jcmc.indiana.edu/vol7/issue2/macao.html
Study,” Academy for Educational Development, for [34] S. Huyer and T. Sikoska, “Overcoming the Gender
the Office of Women in Development, Bureau for Digital Divide: Understanding ICTs and their
Global programs, Field Support and Research, US Potential for Empowerment of Women,” Virtual
Agency for International Development, 2001. Seminar Series on Gender and ICTs, United Nations
[22] Dataquest, “Momentous years of Indian InfoTech”, International Research and Training Institute for the
Dataquest, vol. 18, no. 1, 25-28. Advancement of Women (INSTRAW), 2003.
[23] A. Bertaud, “Mumbai FSI Conundrum: The Perfect Available:
Storm: The Four Factors Restricting the Construction http://www.uninstraw.org/docs/gender_and_ict/Synth
of New Floor Space in Mumbai,” 2004. Available: esis_Paper.pdf
http://alainbertaud.com/AB_Files/AB_Mumbai_FSI_ [35] N. Primo, “Gender Issues in the Information
conundrum.pdf Society,” United Nations Educational, Scientific and
[24] K. Chanana, “Interrogating Women’s Education: Cultural Organization (UNESCO) Publications for the
Bounded Visions, Expanding Horizons,” New Delhi, World Summit on the Information Society, 2003.
India: Rawat Publications, 2001. Available:
[25] R. Heimrath and A. Goulding, “Internet Perception http://portal.unesco.org/ci/en/file_download.php/2505
and 61f24133814c18284feedc30bb5egender_issues.pdf
Use: A Gender Perspective,” International Journal of
Human-Computer Studies, vol. 55, pp. 919-938, 2001. [36] R. Ghadially and S. Ranganathan, “Gendering
[26] F. Umrani and R. Ghadially, “Economic Impact of IT Computer Marketing: A Study of Print Media
Education among Muslims,” i4D Information for Advertisements in India,” Asian Women, vol. 22, no.
Development, 2006. 1, pp. 69-96, 2006.
[27] V. Vehovar, Z. Batagelj and K. Lozar, “Language as
a Barrier,” 1999. Available:
http://www.isoc.org/inet99/proceedings/3i/3i_3.html
1
320
Abstract—A growing proliferation of ICT4D interventions leveraging new agriculture, improving human capital and sus-
has necessitated the exploration of innovative solutions for the taining the rural environment [1]. The adoption of ICT in rural
provisioning of eServices in rural, marginalized communities. communities is a phenomenon with pronounced variability and
The challenges currently faced in these interventions include:
situating the developed applications within the cultural and one which is still not thoroughly understood [2]. It is therefore
ethnographic context of the target communities, integrating still not very clear why some rural communities adopt ICT and
greater levels of granularity and flexibility within the applications why others are sometimes even resistant to the integration of
for increased context sensitivity, handling the intermittence and ICT into their lives. However, research has found that some
instability of supporting infrastructural services. These are the of the leading reasons for rural communities adopting ICT are
challenges that we address in the context of ICT4D intervention
undertaken in a rural community in the Eastern Cape province more social than economic. These communities adopt ICT for
of South Africa. We explore the design and implementation of a improvement of their well being as far as access to information
Multi-Agent System (MAS) for this community as a platform for is concerned and as far as facilitating communication in the
provisioning of context-sensitive eServices, and highlight some communities, and to that end the widely used applications of
observations with regards to the applicability and adequacy of the Internet in these communities are email applications and
the solution.
search engines [3].
Index Terms—eServices platform, ICT for development, eth- ICT is indeed providing opportunities for dealing with rural
nocentric applications, JADE MAS
poverty and in many ways it is challenging the traditional
paradigms of doing business, delivering services to citizens,
I. I NTRODUCTION and running societal institutions [4]. The dynamics associated
A number of factors are fueling the adoption of ICT as an with ICT in development are two fold: on one side its able
enabler towards community development, poverty alleviation to leap-frog community development based on the ability of
and the elimination of the digital divide. One of these is the the society to synergistically embrace the technology (e.g. the
realization of the benefits that are afforded to a community Asian economies: Hong Kong, Taiwan, Singapore) and on the
from participating in the global information and knowledge other side, ICT can further the marginalization, participatory
society, and the other is the sometimes mis-applied observation exclusion from global economies and retardation of commu-
of the penetration of ICT and technology in the developed nities that are not able to integrate technology into their life-
first world countries. The latter factor often leads to the systems [5].
technological determinism flaw, in which the ICT adopters Technology, in and of itself is not a panacea for the underde-
incorrectly assume that the technology that works in the first velopment woes of communities, it is however a pre-requisite
world countries will necessarily work in third world contexts. for social development in this day and age [5]. Technology is
The third world ICT4D interventions are undertaken in a not a target towards community development and social well-
context of pronounced socio-economic challenges that are being [6], but rather a tool for facilitating the achievement of
often not prevalent in first world contexts. They are also a desirable features in a society: well being, health, peace,
undertaken in contexts of extreme rurality characterised by and communality. To a large extent human activity depends
lack of infrastructure (roads, electricity, telecommunications), on information and therefore a synergestic interaction of
lack of service provisioning (health, education, government technology and information leads to a competitive advantage
services), and high levels of marginalization (information, for societies [5].
economic). One of key contributions to understanding the dynamics
It is therefore generally within this landscape of rurality that associated with the adoption of ICT in rural communities
ICT interventions are explored to facilitate development and and in development contexts is made in the discipline of
enhancement of life in rural communities. This environment ethnocomputing. One of the primary tenets of ethnocomput-
of rurality presents unique challenges and opportunities. Five ing is the realization of the culture specific influences on
of these challenges that have been identified by Drabenstott computing and subsequently on the Internet [7]. The aspects
and that are critical in shaping the rural economic outlook of computing that are universal and that are aligned with
are: tapping digital technology, encouraging entrepreneurs, the general human nature, are easily transplantable from one
2
321
community to another. Aspects that are cultural however, need only at a few of the locations in the community, no centers of
to be considered, adapted and validated for adequacy within governmental services provisioning, a small under-resourced
the environment where they’re being implemented [8]. ICT clinic and poor access to information resources for the schools
for development should therefore take into consideration the and the community members at large.
ethnographic considerations and expressions of a community The initial objective of the project was to implement a
to avoid the technology determinism flaw that has plagued prototype of an eCommerce portal in support of the en-
many ICT4D projects [9]. trepreneurship activities already prevalent in this community.
The ability to adapt and localize ICT solutions into the dif- This has been revised and enlarged to establish a distributed
ferent contexts as far as the cultural expressions, the language, community communication and knowledge platform. Over the
the metaphors, the themes, the interaction modalities, and three years that the project has been running a number of
the associated knowledge are concerned, is becoming a key activities have been undertaken and different infrastructure
requirement for ICT4D interventions. This requirement factors setup:
directly into the sustainability and the successful integration of
the solution into the life-system of the communities. Another 1) Community rapport: One of the key achievements in
requirement is for the solutions to address issues of infrastruc- the project has been establishing a working relationship
ture intermittence and unstable deployment environment that with the community as far as exploring ICT based
are typical in rural marginalized communities. This fact of solutions. The buy-in from the different stakeholders
intermittence necessitates that measures and mechanisms are (the headman, the schools, the ICT ’champions’) in the
integrated at all the levels in the service provisioning stack, community provides the essential support that ensures
to be able to recover from interruptions, and to tolerate faults a synergy out of working together with the community.
in the supporting infrastructure. The other constraint as far as It is on the basis of this relationship that the project is
handling faults and interruptions in the network is concerned transitioning into a Living Lab where the community
is the lack of technical expertise in these communities. As members become the active (and key) participants in
far as possible the services should automatically recover and undertaking the implementation of solutions.
self-heal without the need for a technical expert’s intervention. 2) Training: Prior to the introduction of the project in
In this paper, we discuss a solution that we have im- this region, there was minimal computing literacy in
plemented particularly to address these requirements in ICT the community. We have successfully run a number of
applications. The next section introduces the project that we training sessions, with the schools as the centers for
are undertaking, highlighting the different eServices applica- training. The training that we’re undertaking is based
tions being developed for the community. We then discuss on the openICDL curriculum and one of the training
the design factors that should be taken into consideration objectives is ’confidence’ not ’competence’. What this
for context-sensitivity, followed by the detailed discussion of means is that we aim to get the community confident
the Multi-Agent System (MAS) we have developed. The last enough to learn on their own and thus encouraging
sections provide some observations that have been made with a culture of life-long learning in the community. The
regards to the adequacy and applicability of the platform in schools have become the points of access to the ICT
development contexts. solutions for the community because they’re some of
the few places where there’s electricity. The schools
II. ICT4D I NTERVENTION IN SA are also traditionally the centers for learning within the
The ICT4D landscape in South Africa is one in community, and as such we have managed to tap into
which there’s a growing effort from the government, non- that inertia to establish a training system in which we
governmental organizations and academic entities in under- ’train the trainers’ who then continue training the other
taking community development interventions based on ICT people in the community who are not as yet computer
solutions. It is within this context that a joint ICT4D project literate. We have currently setup ’labs’ in five schools
between Rhodes University and University of Fort Hare was in the region ranging in size from 2 computers to 15
initiated in 2006, with support from government and industry, computers.
towards exploring ICT-based interventions that can be under- 3) WiMAX backbone: We have setup a WiMAX network
taken in rural marginalized communities. between the schools where we have established comput-
ing labs. WiMAX is a suitable solution in this case due
to the landscape of the region, and the distances between
A. Overview and Current Status the different points. It also provides adequate bandwidth
The site of this project is a deep-rural, marginalized re- for the services that are provided within the network.
gion of Dwesa, which is located in the former homeland of 4) VSAT connection backhaul: The link to the Internet
Transkei. Dwesa has an estimated population of about 15000 is via VSAT, which is setup in one of the schools
distributed in 2000 households. Dwesa is predominantly a and shared between the rest of the schools through the
subsistence farming community that characterizes the majority WiMAX network. This also provides a link back into the
of rural communities in South Africa and other developing network for remote management and troubleshooting.
regions. Typical of many rural communities, there is a lack 5) VOIP telephony: An Asterisk PBX has been setup at
of infrastructural services in Dwesa: gravel roads, electricity one of the schools to provide VOIP telephony services
3
322
for the whole community. In one way, this allows the be afforded by porting this service onto the MAS eService
community and the schools to collaborate a lot easily platform.
with each other, and in another way it also allows for the 3) eHealth service: One of the primary objectives of the
provisioning of voice based services for the community eHealth services platform is the preservation and aggregation
which is a relevant alternative in a community where of the local traditional health knowledge in Dwesa. This is the
there are people who are illiterate. knowledge that has been developed over the years and that is
The established infrastructure and activities undertaken thus available in the community. The portal also provides accessi-
far provide the basis on which to explore the provisioninig of bility to other sources of medical knowledge. While the portal
different, relevant network services for this community. is aimed for use by the average members of the community,
it can also specifically provide the extra knowledge resources
for the clinics in Dwesa.
B. Associated eServices 4) Indigenous Knowledge portal: One of the reason for
This section highlights a few of the key eServices that the slow adoption and integration of ICT into the lives of
are being integrated into the network in Dwesa. This is rural communities is the lack of relevant knowledge on the
therefore not an exhaustive list of all the different services Internet. This portal aims to alleviate this by allowing for
deployed. Some of the ones not discussed here include: a the community members to author, create and record the
local Wiki which the community can populate and access, knowledge that is relevant and local to their region, to establish
access to offline versions of Wikipedia and Project Gutenberg, and enforce authorization permissions on ’their’ knowledge,
eJudiciary service portal for the community, a community and to have access to other knowledge that is available of the
help-desk system. These are the services that will be ported portal. This of all the services requires the greatest levels of
onto the JADE MAS eService platform to allow for a greater flexibility and context-sensitivity due to the heterogeneity of
contextualization and sensitivity to the local environment. the underlying knowledge, the diversity of user profiles and
1) eCommerce service1 : The Dwesa community has a high preferences, and the multiplicity of end-user devices that are
economic potential due to the fact that there is a nature utilized.
reserve and arts and craft entrepreneurship activities in the
region. There’s therefore the eco-tourism and cultural-tourism III. D ESIGN C ONSIDERATIONS
potential that can be explored and activated through the The development of a MAS eService platform has to address
eCommerce portal. The current arts and craft entrepreneurship a number of challenges. Of particular significance are the
is also currently operating within the geographical constraints challenges that pertain to rurality, processing of indigenous
of Dwesa and the eCommerce portal opens it up to wider knowledge, keeping abreast of technological developments
international markets. A basic functional eCommerce portal around web 3.0, and also ensuring ethnocentricity and context
has been setup and is operational. sensitivity of the developed solution. The design has to take
This portal however has the limitations that it is not into consideration a new way of structuring and layering the
extensively contextualized to the community. The aspects service platform to achieve the following goals:
under consideration for integration in the subsequent versions 1) Provision of an end-user device agnostic interface to the
include: implementing an ontologies based back-end for a underlying knowledge - provision for handling hetero-
semantic processing of the underlying eCommerce knowledge; geneous device requests
allowing for alternative interfaces into the portal which would 2) Allowance for varied interaction modalities with the
be more appropriate for the community (e.g. voice-based users
interface for some of the illiterate arts and craft entrepreneurs, 3) A context-senstive and ethno-centric knowledge plat-
and a J2ME mobile phone interface which would make the form - through encapsulation of local knowledge and
portal more accessible for the majority of the users); and a the emulation of local knowledge system dynamics.
more integrated handling of different associated multi-media This is towards achieving a seamless mobility between
content on the eCommerce portal. knowledge exchange in the real world and knowledge
2) eGovernment service2 : Government service provision- exchange in the virtual space (on the knowledge plat-
ing is very minimal in Dwesa. This eService portal aims to form)
facilitate the ease of access to government services. To a 4) Handling of multimedia knowledge and media transcod-
large extend this portal interfaces and aggregates the avail- ing depending on the capabilities of requesting devices
able government services to provide a one-stop shop for the 5) A future-proof platform that embraces the developments
community. The activities that can be undertaken through this in knowledge engineering technologies and service ori-
portal include: downloading application forms (grants, birth ented architectures
certificates), accessing information released by the govern-
ment, reporting matters to the police, and engaging in an IV. T HE AGENT-BASED P LATFORM
open discussion forum around government related matters.
The similar improvements to the eCommerce portal will also Application development platforms and architecture are
constantly evolving to accommodate the needs within the
1 http://www.dwesa.com software development domain. Evident characteristics in the
2 http://www.dwesa.com/egov nature of applications that are being developed, is that there’s
4
323
MMIA description
context
login.piask this is the context for handling login
requests
logoff.piask this handles the request for logging
off
know.piask handles all the requests for knowledge
Figure 3. Hyper Text Transfer Protocol Agent on the platform
update.piask handles the updating of knowledge on
the platform and adding new
Besides handling the HTTP requests from web browsers, knowledge
do.piask handles requests to perform platform
HTTPA also handles openVXI VXML browser requests from related operations - send an email,
Asterisk. This provides an alternative access from Asterisk and call an extension, notify a user of a
also a richer set of constructs available within VXML mark-up message, etc.
upload.piask handles uploading of content
language which are not available in AGI scripting. register.piask provides the logic for new user
2) Multi-Modal Interaction Agent (MMIA): The MMIA registrations
interaction agent plays a central role in the implementation delete.piask handles deleting of content
of application service logic within the knowledge platform. Table I
It provides the logical interface between the platform users MMIA CONTEXTS
and the underlying platform knowledge services, and between
the different access layer agents and the knowledge base layer
agents. While these contexts are not an exhaustive specification of
The key modules within the MMIA (Fig 4) collaborate to all the functionality implementable in the platform, they form
provide an end-to-end logical handling of the user requests. the core of the procedural logic for operation on the underlying
The request handler module is the first to receive the request knowledge base repositories, which facilitate the provisioning
from the access agents. This module first unpacks the message of different services.
payload to strip off internal inter-agent platform messages. The MMIA also implements the basic handling of different
These messages encapsulate system information such as User interaction modalities with the platform. The MMIA interacts
Session IDs, request context, and variables required for the directly with the access layer agents to receive the requests
processing of the request. from the users. These multiple channels, existing as the various
The AAA module is primarily responsible for the tasks access layer agents, provide the interaction input and also are
of: authenticating the users on the eServices platform, by available as output channels for communication with the users
interfacing with the Social Networking Agent which is aware 3) Social Networking Agent (SNA): The social networking
of the different platform users’ profiles and authentication agent plays the pivotal role of positioning the eServices (and
credentials; determining and enforcing authorization for vari- knowledge) platform within the social context and environ-
ous requests on the underlying knowledge in the system; and ment by integrating the existing social dynamics within the
finally for maintaining basic audit logs for the requests and community. The role of social networking agents is pivotal
processes that are handled on the platform. in the knowledge platform towards the overall objective of
6
325
requests that are sent from the MMIA to the knowledge base
layer are sent directly to the KBA, which processes the request,
to determine the appropriate agent to forward the request to.
The KBA therefore provides an interface and an aggregation of
the knowledge handling agents. The following are the modules
within the KBA that enable it to achieve its role of managing
the interaction with the rest of the knowledge layer agents (Fig
7):
V. O BSERVATIONS
The following preliminary observations have been made,
Figure 9. Example platform interactions around the applicability and the adequacy of this eService
platform in Dwesa:
The JADE MAS platform inherently provides an Agent 1) The distributed nature of multi-agent systems provides
Communication Language (ACL) messaging mechanisms as an architecture closely aligned to the infrastructural
8
327
situation on the ground in Dwesa (and in most rural communities. The developed solutions have to be implemented
marginalized communities). The resources in Dwesa are with a high level of flexibility to allow for situating within
distributed between different communities who share different cultural and environmental contexts. The services
and cooperate with each other in a symbiotic manner. deployed on these platform also have to be relevant to the local
The eService platform can easily be distributed depend- communities and to integrate the local (indigenous) knowledge
ing on the availability of the resources that the agents and systems dynamics with these communities.
have to access. Thus the agent that provides or handles In addressing these challenges in a ICT4D context is a
connectivity to the Internet, can be located on a gateway community in South Africa, we have developed a JADE based
host in one school, and an agent that interfaces with the MAS that provides an architectural support for the provision-
Asterisks PBX can be co-located with the PBX. ing of eServices for this community. The observations made
2) The eService platform is robust and fault-tolerant. The thus far allude to the applicability of such a platform in the
reality of rural communities is that the availability of provisioning of context sensitive, ethnocentric eServices.
different network nodes is not always guaranteed, some-
times due to the underlying infrastructural constraints R EFERENCES
(e.g. lack of electricity). The fault-tolerance within the [1] Drabenstott, M., “New Policies for a New Rural America”, International
eService platform is established through a mechanism Regional Science Review, vol. 24, no. 1, pp. 3, 2001
in the JADE MAS that allows for migration of agents [2] Mitchell, S. and Clark, D., “Business adoption of information and com-
munications technologies in the two-tier rural economy: some evidence
between different hosts on a network, and the automatic from the South Midlands”, Journal of Rural Studies, vol. 15, no. 4, pp.
initialization and termination of agents available on the 447–455, 1999
platform. The platform is therefore able to self-heal in [3] Malecki, E.J., “Digital development in rural areas: potentials and pit-
falls”, Journal of Rural Studies, vol. 19, no. 2, pp. 201–214, 2003
cases where agents get terminated when a host goes [4] Bhatnagar, S., “Information Technology and Development Foundation
down. There is also the ability for the replication of and Key Issues”, Information and Communication Technology in Rural
the Main Container within the JADE platform, which development: Case Studies from India. World Bank Institute, pp. 1–12,
2000
allows for the continuance of a MAS life in the case of [5] Castells, M., “Information technology, globalization and social develop-
one Main Container being terminated. ment”, United Nations Research Institute for Social Development, 1999
3) The eService platform has been validated for functional [6] Hietanen, O., “The global challenges of e-Development - from digital
divides towards empowerment and a sustainable global information soci-
adequacy through a load testing experiment, in which ety”, in Seminar of global perspectives of development communication,
the platform was observed to provide a linear scaling University of Tampere, 2004
in performance for handling users’ requests. The perfor- [7] Tedre, M., Sutinen, E., Kahkonen, E. and Kommers, P., “Ethnocomput-
ing: ICT in cultural and social context”, Communications of the ACM,
mance degradation, under heavy usage loads, was also vol. 49, no. 1, pp. 126–130, 2006
observed to be gradual. [8] Thinyane, M., Dalvit, L., Terzoli, A. and Clayton, P., “The Internet in
4) The interaction modalities implemented within the eS- rural communities: unrestricted and contextualized”, in ICT Africa -
Nepad Council, Addis Abba - Ethiopia, 2008
ervice platform provide a crucial interface that has been [9] Thinyane, M., Terzoli, A. and Clayton, P., “Transitions towards a
lacking through a number of the solutions that have knowledge society: Aspectual pre-evaluation of a culture-sensitive im-
been implemented in rural communities where there’s an plementation framework”, IFIP – Learning to live in the knowledge
society, July 2008, Vol. 281, pp 271-278
especially higher levels of illiteracy. One such interface [10] Cybenko, G., Gray, R., Khrabrov, A. and Wu, Y., “Information theoretic
is the voice interface to the eService platform, provided principles of agents”, in Proceedings of the CIKM - Workshop on Intelli-
through the available VOIP telephone infrastructure in gent Information Agents, Third International Conference on Information
and Knowledge Management (CIKM 94), 1994
the community. Not only is the voice interface crucial [11] Leymann, F., “Web Services: Distributed Applications without Limits”,
for the illiterate people, it is also import due to the nature Business, Technology and Web, Leipzig, 2003
of the knowledge that is available in these communities [12] Payne, T. R., “Web Services from an Agent Perspective”, IEEE Intelli-
gent Systems, vol. 23, no. 2, pp. 12-14, 2008
and also that the majority of these communities have [13] Leszczyna, R., “Evaluation of agent platforms”, European Commission,
traditionally audio based communication mediums (e.g. Joint Research Centre, Institute for the Protection and Security of the
the African drum). Citizen, Ispra, Italy, Tech. Rep., June, 2004
[14] Bellifemine, F., Poggi, A. and Rimassa, G., “JADE–A FIPA-compliant
These observations position the implemented eService portal agent framework”, Proceedings of PAAM, pp. 97–108, 1999
to provide a solution that meets the needs and the requirements [15] Thinyane, M., Dalvit, L., Slay, H., Mapi, T., Terzoli, A. and Clayton, P.,
“An ontology-based, multi-modal platform for the inclusion of marginal-
that are prevalent in rural, marginalized communities. These ized rural communities into the knowledge society”, Proceedings of the
are preliminary observations with will be validated through ACM SAICSIT conference on IT research in developing countries, pp.
the extensive usability testing and usage profiling. 143–151, 2007
[16] Thinyane, M., Dalvit, L., Terzoli, A. and Clayton, P., “Towards a Model
of an Ontology Based, Multi-Modal and Multimedia Knowledge Portal
VI. C ONCLUSION for Marginalized Rural Communities.”, in Proceedings of IEEE Infor-
mation Communication Technologies International Symposium, Fez -
The ICT4D research and implementation landscape is one Morocco, 3rd - 5th April,
[17] Eberman, B., Carter, J., Meyer, D. and Goddeau, D., “Building
faced with unique challenges which necessitate new and voiceXML browsers with openVXI”, Proceedings of the 11th interna-
innovative solutions. We have highlighted challenges which tional conference on World Wide Web, pp. 713–717, 2002
are associated with: the fact of cultural diversity in which [18] FIPA, ACL, “Message Structure Specification”, Foundation for Intelli-
gent Physical Agents, 2000
the ICT4D interventions have to be undertaken, and the in-
termittence of the infrastructure services in rural marginalized
328
Abstract—The recent increase in natural disasters has a 15.3% internet penetration rate [17], growing from 11.8%
significant impact on the lives and livelihoods of the poor in in 2006 [5] to over 500 million internet users [17]. One
Asia. The spread of information communication should note there is a disparity in access to these
technologies (ICTs) in this region’s rural areas suggests the technologies, as captured in the digital divide debate.
potential of technologies to enhance recovery efforts. While
Further, urban dwellers have more economic, educational
many ICT initiatives have been implemented to aid disaster
management, from providing early warning to immediate and technological familiarity with these technologies
relief, there exists a gap in the theoretical understanding of compared to the rural population [18]. Despite this divide
the role of technologies in disaster recovery and in technology access and use, the use of ICTs in disaster
rehabilitation. We propose a conceptual framework for warning, mitigation and management shows the benefits
understanding the implementation of ICTs in recovery of technology diffusion in rural areas.
operations, drawing attention to vulnerability reducing An examination of the literature suggests that ICTs are
potential of the initiatives. We review theories on ICT use in being used beyond merely facilitating early-warnings
disaster management, and propose the Extended about impending disasters. The role that ICT systems,
Technology-Community-Management model focusing on
which range from mature technologies such as radio,
vulnerability assessment for the design and implementation
of ICT programs for development in rural areas. We television, and land-line telephony, to advanced modern
illustrate this model using case studies from ICT technologies such as Global Positioning Systems (GPS),
deployments in post-disaster Asia, particularly India, Geographic Information System (GIS), Very Small
Indonesia, Sri Lanka and China, and suggest implications Aperture Terminal (VSAT), and cellular phones and
for theory and practice. satellite communications, play at various stages of
disaster management has been recognized by academics,
Index Terms—Asia, disaster management, ICTD, governmental agencies, civil society groups, non-
rehabilitation, vulnerability. governmental organizations (NGOs), and voluntary
welfare organizations in the development arena [19]-[22].
Communication technologies provide access to
I. INTRODUCTION information that is a vital form of aid in itself. “Disaster-
Globally, more than 250 million people are affected by affected people need information as much as water, food,
natural disasters every year, and the intensity and medicine, or shelter” [23], particularly in the post-disaster
frequency of catastrophes have steadily been increasing in context, from immediate relief to long-term rehabilitation
the last decade [1]. Asia is amongst the most affected and efforts. This suggests the potential of ICTs in various
vulnerable regions, with disasters such as earthquakes, operations immediately following a disaster, from
floods, tsunamis, cyclones, and droughts killing enabling prompt information dissemination to relief
thousands of people each year, while millions have lost agencies and affected communities, to livelihood
their homes, properties, livelihoods and families [2]-[4]. rebuilding initiatives in the recovery phase. Researchers
The increase in the proportion of people living below the have focused on the use of these technologies for disaster
poverty line, from 30% to 50% in post-tsunami Indonesia, response in the mitigation and preparedness stages [24]-
suggests the impact of calamities on the poor, especially [29], with less emphasis placed on the scope of ICTs in
those living in less developed Asian countries [5], [6]. the rehabilitation stage with the objective of exploiting
These marginalized groups face the challenge of coping the long-term and continual benefits of technologies,
with, and recovering from the effects of disasters as particularly in rural communities.
overall development and economic growth suffer major There is a need to establish a canon for best practice
setbacks in the aftermath [7]-[10]. [7]. This study aims to bridge the existing theoretical gap
The recent spread of information communication by proposing a conceptual framework to guide the design
technologies (ICTs) in Asia [11], [12] offers more people and implementation of ICT for development (ICTD)
access to technologies that can to aid in disaster programs, specifically in post-disaster management.
management efforts [13]-[15]. There has been a steady In this paper, we critically examine and categorise
rise in the penetration, accessibility, and use of ICTs in existing theories in disaster management and ICT. We
Asia. India, in particular, continues to be one of the integrate the theoretical implications of this framework in
fastest growing major mobile telecom communication relation to the Technology-Community-Management
markets in the world, with annual growth rate of 91% [5]. (TCM) model [18], [30]. We propose an extended model
With China’s 253 million internet users [16], Asia has a
329
and illustrate it using primary data from India and structure of the steps that need to be taken at each level,
Indonesia, and cases from China and Sri Lanka. the stages theories chart an overview of the levels through
which ICT users and disaster operation managers
successively progress. The longitudinal approach was
II. REVIEW OF MODELS AND FRAMEWORKS initially adopted by the eight socio-temporal stages of
We conducted a structural understanding of existing disaster framework [31]-[33] that proposed the
conceptual frameworks on disaster management (DM) classification of the disaster management efforts
and ICT usage in order to assess their theoretical impact according to specific outcome strategies, such as warning,
and practical effectiveness to technology in disaster impact, rescue, and recovery. This argument was further
recovery (see Fig. 1). The ICT-DM framework identifies expanded in models, such as the six-phase disaster cycle
specific factors that are emphasised consistently in these [34], [35], [4], three stages of disaster [36], three time-
studies, to draw two broad categorizations – Procedure- dimensions of rehabilitation efforts [14], and strategic
Based and Community-Oriented theories. Procedure- disaster management cycles [37], that elaborated the
Based theories cover the temporal segmentation of relationships between specific temporal parameters and
disaster management and ICT initiatives, and illustrate objectives guiding these efforts. Relating to practical
the management structures involved. Community- situations, the influence-impact model [8] emphasized the
Oriented theories provide insights into practically importance of accurate, timely, and complete information
relevant issues involving the community, and suggest outflow for basic level ICT impact to occur. Furthermore,
impediments to productive technology influence. These the ICT step change table [25] identified three levels
categories are further divided to guide the systematic within projects, where the informational needs of people,
review of concepts within each category, and to aid such as knowledge of technology, literacy level, and skill
analysis in recognizing the commonalities between capacity of users, determine the hierarchal levels of ICT
theories. The objective of the review is to identify gaps in projects. We argue that while stages theories are useful in
theory and focus, and to propose a theoretical model that highlighting the hierarchy of actions and measures
fills this research gap. necessary prior to, during, and after disaster contexts;
they fail to specify potential barriers to the efforts
implemented at each stage, particularly in ICT
deployment. It is difficult to draw distinct boundaries in
processes where users, especially those affected by
disasters, possess varying, and often overlapping
informational needs, and capacities to process
information. In these cases, the boundaries between
stages can be fuzzy, and stage-related behavior can be
concurrent [38].
Systems theories present factors affecting information
flow in recovery efforts, arguing that ICT providers in
disaster contexts have to be critical of the information
content disseminated to affected people. They draw
Fig. 1. Review of Disaster Management and ICT Models/ Frameworks. attention to the management of recovery projects by
pointing out ways in which governments and NGOs make
critical investments in ICT projects with the support of
A. Procedure-Based Theories
the private sector. This illustrates the essential financial
Procedure-Based theories describe the processes roles external bodies play in ensuring the economic
involved in effective management of disaster and ICT livelihood of disaster affected people. Systems theories
projects by pointing out the key stages in planning such as the CARICOM structure [39] and disaster
recovery efforts, as well as identifying the stakeholders management consortium [40] identified stakeholders
and management implications to technology involved in decision-making and coordination of disaster
implementations. The theories emphasize the importance operations, while the optimal disaster information flow
of informational flow in influencing economic and model [39] and the technical conceptual system model
psychological rehabilitation in recovery. While these [41] map the roles information plays in disaster
theories are useful in understanding enabling factors, they management. We suggest that systems theories can
largely fail to address the importance of community expand beyond coordination of information and
involvement, or identify barriers to success for ICT interactions by drawing attention to the action-oriented
initiatives in post-disaster situations. information exchanges that highlight the combined efforts
We sub-divide Procedure-Based theories – stages and of stakeholders. An action-oriented strategy relates to the
systems theories. While stages theories classify linear management style adopted by technology managers who
processes and define their scope, systems theories look beyond profit making, and encourage community
demonstrate what needs to be done, by whom, and the level participation in the management of ICT
methods to use for optimal information flow. deployments. We argue that the deliberate attention to
Stages theories illustrate the management of disasters in community involvement and ownership in running
a longitudinal approach, where both disaster-based and programmes would expand the boundaries of systems
ICT-specific frameworks craft the different stages as theories to look beyond traditional organizational styles in
mutually exclusive levels of impact. By mapping a governing ICT initiatives.
330
Together, the stages and systems theories suggest that could increase community vulnerabilities in post-disaster
ICTs can be best capitalized in the recovery stage, where situations. Among the initial debate [46]-[49], it is
emphasis is on the psychological and physiological suggested that the design of development programs
rehabilitation of affected communities. Most stages should focus on decreasing vulnerability of communities
theories contend that ICT rehabilitation efforts, in the to disasters and their negative consequences [50].
areas of physical and social rehabilitation, and health and Supplementing this, the capacities and vulnerabilities
educational rehabilitation, should focus on multi-year analysis [46] provided disaster managers a framework for
programs. This emphasizes the need for well-planned understanding and reducing vulnerabilities, and suggested
long-term recovery efforts to ensure that people regain an approach for the design and evaluation of development
their economic and psychological balance. Further, these projects. Similar to the pressure and release model [44],
models suggest that fostering of economic livelihood and access model [44] proposed that reducing
should be based on recognizing principles of social vulnerabilities involves increasing capacities of the
inclusion and gender equity, especially those involving affected people at the individual, group, and societal
fishing and farming communities. On the whole, levels, arguing that well-organized and cohesive
procedure-based stages and systems theories describe the communities can withstand or recover from disasters
linear dimensional design and practical management better than those divided by issues of race, religion, class,
aspects useful in planning and implementation of ICTs in caste, gender, ethnicity, or age. Reference [51] criticized
disaster situations. the sustainable livelihoods approach [52] framework for
failing to acknowledge socio-cultural determinants that
B. Community-Oriented Theories
influence poor people’s access to economic-enhancing
Various studies have argued that particularly at the resources and livelihood assets. This model emphasizes
recovery stage, plans need to be understood, accepted, the impact of socio-cultural factors on technology
and implemented at the local level to be sustainable, and adoption and use, critical in determining the equal access
to achieve intended outcomes on communities in post- of an individual, group, or community to ICT resources.
disaster situations [42], [43]. Community-Oriented We suggest that vulnerability theories need to examine
theories emphasize various aspects of community factors impeding community livelihood, access, and
involvement in disaster management; essentially, developments, and recognize ways to improve the overall
elaborating the existence of susceptibility factors that community participation in vulnerability reduction.
impact recovery efforts aimed at affected people. While The ICT skills and discretionary slack [53] framework,
the theories connect various practical applications a capacity-oriented theory, highlighted the importance of
recognising community needs and wants in disaster capacity building in vulnerability reduction by proposing
recovery, they generally ignore the technical demands of a correlation between ICT skill level, and availability of
these recovery efforts, where over-emphasis had been time. It fails to account for the fact that the factors of skill
placed upon providing a range of applied methods to and time are affected by both psychological and
enhance community well-being instead of elaborating on physiological determinants of users. This would affect the
the know-how of running these recovery initiatives. adoption, use, continuation, and termination of ICT
We split Community-Oriented theories – vulnerability, deployments. For instance, discretionary time available
and capacity theories. Vulnerability theories describe could be affected by conditions such as the amount of
various aspects of vulnerability that both exist and arise in training sessions offered, and cultural, gender-based and
post-disaster contexts. Conversely, capacity theories look social restrictions. The psychological component
at ways to improve the capabilities of people in order to affecting ICT skill level could include individual literacy
optimize ICT usage and benefits. levels, attitudes towards ICT adoption, and overall
Affecting people with varying magnitudes, willingness to learn. Overall, this theory suggests that
vulnerability has been defined as the “characteristics of a information processing capacity and access to
person or group, and their situation that influence their information is dependent on the socio-cultural
capacity to anticipate, cope with, resist, and recover” determinants as well as the mental state of individual to
from the impact of a disaster [44]. With insufficient absorb and use the information.
capacity to cope, the poor are most vulnerable in the We contend that the vulnerability and capacity theories
aftermath of disasters [9], [24]. Vulnerability theories highlight the mutually dependent relationship between
analyze the role ICTs play in recovery efforts aimed at reducing vulnerabilities and increasing capacities in
reducing vulnerabilities of the rural poor, by increasing enhancing overall community involvement in post-
capacities and focusing on livelihood enhancing disaster ICT use. However, we think community
development efforts. The theories discuss aspects of involvement in ICT initiatives need to go beyond mere
informational, socio-cultural, and psychological recovery participation, and explore opportunities for community
that significantly impact economic vulnerability. In this empowerment through ownership, and active contribution
study, we look at five inter-related vulnerability theories; in the management of ICT efforts directed towards
(1) development-disaster framework, (2) capacities and recovering the livelihoods of the rural poor.
vulnerabilities analysis, (3) pressure and release model,
(4) access model, and (5) sustainable livelihoods
approach. III. EXTENDED TECHNOLOGY-COMMUNITY-
The development-disaster framework [45] highlighted MANAGEMENT MODEL
the importance of information in assessing vulnerabilities,
arguing that the lack of, or poor access to, information The examination of disaster management and ICT
literature allowed us to develop a framework for the
331
differential provided better economic resources to rural Indian fishing communities, with reported livelihoods
farmers. increasing from 20% to 50% [54]. Group ownership of
A mobile communications system was established in the devices ensured maintenance of the equipment. The
Banda Aceh, Indonesia, as a potential means of long-term environmental impact of technology usage was
improving health services, where comparatively poor mitigated by inviting local government officials from the
maternal and child mortality rates worsened after the fisheries department to impart responsible fishing
tsunami [60]. The primary objective of mobile phone practices.
usage by rural midwives was to provide low cost means The effects of poor informational infrastructure can be
of improving the handling of complicated obstetrical devastating on the available educational resources in
cases, thus lowering rates of pre-natal and maternal affected areas. In addition to the destruction and damage
mortality. However, a key information gap identified in of scholastic infrastructure, there is a crucial impact on
the rural healthcare system was the lack of maternal human capital, in terms of the loss of teachers and
health data by urban-based obstetricians. To rectify this parents. For children, ICTs are a source of learning
vulnerability, a Short Message Service (SMS) software opportunities that help in building critical technological
system was developed to allow midwives to upload skills. With the primary objective of providing
critical health information to an online database, thus educational opportunities, World Vision equipped two
improving access, as well as enabling the rural midwife to secondary schools in affected-villages of Tamil Nadu,
consult senior nurses and doctors using mobile phones for India, with computers, providing lessons integrated with
particular cases [58]. the school curriculum. In addition to computer-oriented
Offering trainings to aid greater adoption of technology skills, children learnt Tamil, English, and other academic
can make a significant difference to the information subjects [54]. It was essential to provide an outlet for
accessibility of rural community, beyond mere access to graduates to utilize their skills in an economically
the technology. In drought-hit Yellow Sheep River, productive manner. Business training, including
China, the “Town and Talent Project” equipped the introduction to productivity software, was provided to
agricultural village’s e-commerce centre with 25 young adults to improve their job prospects in the
computers. Online courses equipped farmers with increasingly technologically-savvy Indian workforce.
relevant computer skills for better retrieval and usage of
D. Socio-Cultural Vulnerability
agricultural information [59]. A community centre in
Jayagiri, Indonesia promoted community radio and Community pressures are seen most directly in the
multimedia as educational tools to encourage people- existing networks of power within the social system. The
centered development in disaster recovery. The training traditional power of the higher castes (social power), the
series, directed at capacity-building of centre managers middlemen (economic power), and males (gender power)
and volunteers running the community radio and limits the opportunities available to those groups with less
programmes, aimed to improve technological skills for power. Introducing ICTs in such biased environment
creating program content [57]. could work in the favour of the haves, and likely further
Finally, informational capacity building was greatly marginalize the have-nots. Tactfully designed ICTD
enhanced by community management of technology in initiatives have the potential to champion the cause of
recovery efforts. The collective-ownership of facilities development without widening existing social
can promote the dissemination and sharing of information inequalities. Recognizing this, the M. S. Swaminathan
within communities. The "Town and Talent Project" in Research Foundation, an Indian NGO, established
Yellow Sheep River introduced internet-enabled knowledge centers in Dalit (lower caste) villages [61];
computers to enrich educational resources in the local while World Vision trained women Self Help Groups to
school. Collaborating with a local company, the teachers take ownership of the ICTD projects within their
were trained to set up the scholastic network. These communities [62], [63]. The World Vision multimedia
teachers were the central information providers on centres in most villages claimed no socio-cultural
computer usage to both their peers and children [59]. The restrictions on access, with no separate timings for young
Nanasala telecenter project in Sri Lanka was managed by girls, or lower castes individuals, allowing a co-mingling
individual entrepreneurs within the community to provide of different sections of society [54].
internet access to rural villages [55]. The initiative In some situations, the socio-cultural issues
exposed rural people to internet, computer, and telephone surrounding communities may indirectly affect people’s
technologies with free training sessions to gain perceptions on their capacity to use technologies. In rural
knowledge of technology use and online retrieval of classrooms in China, elder teachers were reluctant to take
information on new agriculture practices and techniques, on computer usage for fear of not being able to
and information on a variety of health issues. effectively learn and adopt computer skills,
predominantly possessed by younger users. Training
C. Economic Vulnerability courses helped staff overcome this technophobia, and
Information communication technologies have the grasp basic E-literacy in daily teaching practices [59].
potential to empower communities to regain economic Indonesian midwives in the Banda Aceh mobile-phone
stability in the aftermath of disaster. However, beyond the project initially perceived the ICT-based health practice
economic outcomes, the ability of communities to engage as a monitoring system; rather than a tool to assist their
in relatively expensive technology usage, and to manage work. This negative perception, triggered by existing
their economic expectations, need to be considered. The hierarchical class structure within the health
use of GPS and fish-finders led to income generation in infrastructure, could have encouraged them to revert to
334
the prior, more inefficient, paper-based system for followed by designing initiatives, is an iterative process
reporting maternal health [64]. Another criticism was the involving evaluation and reconfiguring of project design,
unfamiliarity with text-based technology. To overcome and will necessarily be a long-term phenomenon. The
these barriers, joint training seminars were held with durations of reaching this goal will vary depending on
doctors, health coordinators and rural midwives. several intrinsic and extrinsic conditions affecting ICTD
Secondly, a JAVA-based applet was developed that programs at different stages of execution. First, it is
functioned with scroll and select functionality, rather than necessary to determine a methodological approach to
using the traditional text-based, and possibly elaborate on the criteria for assessing the vulnerabilities,
cumbersome, SMS. Constructive responses to the determining the order of priority for tackling the
suggestions and feedback from midwives increased obstacles, and gauging the effectiveness of measures.
credibility of the initiative, which subsequently Secondly, beyond limiting the analysis of psychological
encouraged the use of ICT beyond existing social class- vulnerabilities to disaster affected rural poor, the
based barriers between rural midwives and hospital-based assessment should observe other extrinsic conditions such
doctors. as the mental barriers of ICTD program managers and
policy makers in related domains. Preconceptions of the
marginalized communities’ learning capacities, adoptive
VII. DISCUSSION AND CONCLUSION abilities, enthusiasm levels, and financial capabilities to
An examination of Asian cases reveals that the use and manage technology could impede their
deployment of ICTs in disaster recovery needs to take commitment to implement and sustain ICT initiatives.
into consideration the vulnerabilities impacting social We should note that the identified variables of
sustainability. This study extends the theoretical literature physiological and psychological vulnerability,
on disaster management and ICT, by addressing a gap in informational vulnerability, economic vulnerability, and
research on the best practices for ICT deployment in socio-cultural vulnerability, are not mutually excusive
disaster recovery. The TCM model proposed that ICTD variables. The cases highlight the existence of
program design should include critical variables overlapping vulnerabilities, and those that are sometimes
involving communities, including needs, ownership, and in opposition to one another. ICT programs focusing on
training, in addition to the external impacts of technology reducing economic susceptibility, such as the use of GPS
and management. We extend the TCM model to include system by fishermen, also have a positive impact on the
four key vulnerability dimensions. We recognize the psychological well-being of users. Educational programs
limitations of frameworks with clearly defined for rural midwives involving cell-phone training
boundaries, and discuss the inter-relationships of addresses the psychological fear of technology, while
vulnerabilities and ICTD program deployment simultaneously reducing the informational barriers
characteristics of technology, its management and the present. However, vulnerabilities can contradict each
community involvement. The case studies illustrate real other when the attempt to reduce one barrier magnifies
world complexities, in terms of barriers to the negative impact of another. For example, ICTD
implementation and considerations for long-term program programs that aim to enhance information flow among
success. communities through the usage of mobile phones could,
A critical distinction needs to be made between core in the long-run post-subsidy, negatively affect the
development objectives and vulnerabilities, as these may financial capacities of communities which struggle to
not coincide, and may even be antithetical. Vulnerabilities maintain and continue usage of the adopted technology.
are inherent factors inhibiting communities from We have previously critiqued the stage-based approach
accessing, adopting, and realizing the benefits from of extant theories; yet we acknowledge that the
technology implementations. We believe these barriers establishment of clear stages can provide a guideline for
need to be addressed in addition to core development ICTD program assessment, particularly for practitioners.
objectives, such as those recommended by the United It is possible that different vulnerabilities may take
Nation’s MDG [65]. The choice of development priority at certain stages of implementation, leading to
objectives can influence the selection of beneficiaries, specific TCM configurations. The extended TCM model
their control and use of technology, leading to an does not incorporate hierarchies in management flow,
amplification of existing vulnerabilities in communities. community participation, or technology diffusion,
For instance, the management of rural initiatives, largely suggesting their relationship to vulnerabilities occurs in a
led by village leadership councils, could lead to cooperative manner.
privileged access to technologies, in turn furthering the The communication infrastructure within a community
sociological imbalance within society. Further, the consists of both old and new ICTs, as well as established
emphasis on education in post-disaster societies, often and evolving social networks. This study focused on
placed on basic primary education, could ignore the needs implementations involving a single technology, aided
of adolescents and young adults, raising issues of age largely by the fact that most ICTD interventions are
discrimination to technology access. We believe that designed as such. We propose that researchers examine
project designers need to recognize the existence of dynamic communication and information environments,
vulnerabilities within the social system, and take steps to themselves embedded within fluid social networks, rather
mitigate them, in addition to focusing on overall project than one technology at a time.
objectives. Future research can build on the extended TCM model
Realistically, project managers have to recognize that as a theoretical framework to guide research in the design
the process of identifying existing vulnerabilities, of ICTD programs, and to test the significance of
335
vulnerabilities beyond disaster recovery. We propose that [Accessed: Aug. 15, 2008].
[9] B. K. Paul, “Disaster relief efforts: an update,” Progress in
the vulnerabilities assessment approach can and should be
Development Studies, vol. 6, pp. 211–223, 2006.
applied to varied contexts such as education, healthcare, [10] United Nations Development Programme, ICT and human
agriculture, and conflict management. Subsequent studies development: Towards building a composite index for Asia. New
in the area of disaster recovery could examine how the Delhi: Elsevier, 2004.
[11] International Telecommunication Union, ICT and
TCMV model can examine the impact of individual
telecommunications in the least developed countries. Geneva:
vulnerability dimensions to technology, management, and International Telecommunication Union, 2006.
community variables. Alternatively, research could focus [12] United Nations Development Programme, “Community-based
on specific components within the TCM model, such as networks and innovative technologies: New models to serve and
empower the poor,” United Nations Development Programme,
community training, within the Management-Community
2005. [Online] Available:
intersect, and examine its relation to the four identified http://www.undp.org/poverty/docs/ictd/ICTD-Community-
vulnerabilities. Here, the aim would be to explore Nets.pdf [Accessed: Sep. 10, 2008].
thoroughly the barriers that exist and arise within specific [13] N. Gunawardene and F. Noronha, Communicating disasters: An
Asia Pacific resource book. Bangkok: United Nations
components, and suggest solutions to overcoming them.
Development Programme Regional Centre, 2007.
There is no doubt that calamities have a [14] M. Jansz, “Disaster recovery and ICT in Sri Lanka: The day after,”
disproportionately negative impact on the poor. However, Information for development, vol. 3, pp. 10–22, Jan. 2005.
the rapid proliferation of ICTs, particularly in Asia, has [15] C. Wattegama, “ICT in disaster management,” 2007. [Online]
Available: http://www.apdip.net/apdipenote/16.pdf [Accessed:
the potential to engender positive recovery to the lives
Sep. 15, 2008].
and state of those affected. We believe well-designed [16] China Internet Network Information Centre, “Statistical survey
ICTD programs have the power to increase the capacities report on the internet development in China: Abridged edition,”
of communities, while simultaneously reducing the China Internet Network Information Centre, 2008. [Online]
Available:
vulnerabilities towards technology adoption. Finally, rural
http://www.cnnic.cn/download/2008/CNNIC22threport-en.pdf
communities can witness long-term social sustainable [Accessed: Aug. 09, 2008].
progress. [17] Internet World Statistics, “Asia marketing research, internet usage,
population statistics and information usage,” Internet World
Statistics, 2008. [Online] Available: from
REFERENCES
http://www.internetworldstats.com/asia.htm [Accessed: Sep. 15,
2008].
[1] Department for International Development, “International day for [18] A. Chib and J. Zhao, “Sustainability of ICT interventions: Lessons
disaster reduction,” Department for International Development, from rural projects in China and India,” in
2007. [Online] Available: Communicating for social impact: Engaging communication
http://www.dfid.gov.uk/news/files/disasters-day.asp [Accessed: theory, research, and pedagogy, L. Harter and M. J. Dutta, Eds.,
Aug. 15, 2008]. ICA 2008 Conference Theme Book, Hampton Press, in press.
[2] Centre for Research on the Epidemiology of Disasters, “Annual [19] H. E. Hudson, “Universal access to the new information
disaster statistical review: The numbers and trends 2007,” Centre infrastructure,” in Handbook of New Media: Social Shaping and
for Research on the Epidemiology of Disasters, 2007. [Online] Consequences of ICTs, L. A. Lievrouw and S. M. Livingstone,
Available: Eds., London: SAGE, 2002, pp. 368–384.
http://www.emdat.be/Documents/Publications/Annual%20Disaster [20] C. K. Prahalad, The fortune at the bottom of the pyramid:
%20Statistical%20Review%202007.pdf [Accessed: Aug. 10, Eradicating poverty through profits. Philadelphia: Wharton School
2008]. Publishing, 2004.
[3] Commonwealth Telecommunications Organization, “Using ICT [21] United Nations Conference on Trade and Development,
for effective disaster management forum: A summary report on Information and communication technology development indices.
sessions and panel discussions,” Commonwealth Geneva: United Nations Publications, 2003.
Telecommunications Organization, 2006. [Online] Available: [22] S. Yodmani and D. Hollister, “Disasters and communication
http://www.lirneasia.net/wp- technology: Perspectives from Asia,” in Second Tampere
content/uploads/2006/07/CTO%20Report%20Uning%20ICT%20f Conference on Disaster Communications, 2001. [Online]
or%20DM%20June%202006.pdf [Accessed: Aug. 14, 2008]. Available:
[4] United Nations Economic and Social Commission for Asia and the http://www.reliefweb.int/telecoms/conference/cdc2001/dh.pdf
Pacific, “ICST-enabled natural disaster management in Asia and [Accessed: Aug. 15, 2008].
the Pacific region,” United Nations Economic and Social [23] J. Walter, ed., World disasters report 2005: Focus on information
Commission for Asia and the Pacific, 2005. [Online] Available: in disasters. Geneva: The International Federation of Red Cross
http://stdev.unctad.org/unsystem/cstd/escap8.doc [Accessed: Aug. and Red Crescent Societies, 2005, pp. 11.
15, 2008]. [24] M. I. Ahmed and E. C. Lentz, “Enhancing the livelihoods of the
[5] International Telecommunication Union, “ICT statistics database,” rural poor through ICT: A Knowledge Map,” 2007. [Online]
International Telecommunication Union, 2008. [Online] Available: http://www-
Available: http://www.itu.int/ITU- wds.worldbank.org/external/default/WDSContentServer/WDSP/IB
D/ICTEYE/Indicators/Indicators.aspx [Accessed: Aug. 15, 2008]. /2008/11/26/000333037_20081126002013/Rendered/PDF/466300
[6] OneWorld South Asia, “Making MDGs disaster proof,” OneWorld NWP0Box31esh0Country0Study111.pdf [Accessed: Aug. 01,
South Asia, 2008. [Online] Available: 2008].
http://southasia.oneworld.net/Article/making-mdgs-disaster-proof [25] A. Chetley, “Improving health, connecting people: The role of ICT
[Accessed: Aug. 15, 2008]. in the health sector in developing countries. InfoDev framework
[7] Centre for Research on the Epidemiology of Disasters, “An paper,” 2006. [Online] Available:
analytical review of selected data sets on natural disasters and http://www.asksource.info/pdf/framework2.pdf [Accessed: Jul. 15,
impacts,” Centre for Research on the Epidemiology of Disasters, 2008].
2006. [Online] Available: [26] K. Delgadillo and R. Borja, “Learning lessons from telecentres in
http://www.emdat.be/Documents/Publications/TschoeglDataSetsR Latin America and the Caribbean,” 1999. [Online] Available:
eview.pdf [Accessed: Aug. 15, 2008]. http://www.idrc.ca/telecentre/evaluation/nn/16_Lea.html
[8] A. F. Gorp, “ICT as an enabler for humanitarian relief? An [Accessed: Aug. 30, 2008].
assessment of communication and information flow in inter- [27] T. Eisensee and D. Stromberg, “Disaster relief,” The Quarterly
organizational context,” presented at the annual meeting of the Journal of Economics, vol. 122, pp. 693–728, 2007.
International Communication Association, New York, 2005. [28] A. Hovanesian and A. Cox, “ICT for tsunami recovery: Best
[Online] Available: practices and lessons learned,” presented at the E-Asia
http://test.scripts.psu.edu/users/a/f/afv103/ICA2005.pdf Conference, Putrajaya, Malaysia, 2007.
336
[29] O. Takizawa, “Using ICT in disaster mitigation,” National [48] A. Kreimer and M. Zador, Colloquium on disasters, sustainability
Institute of Information and Communications Technology News, and development: A look at the 90s. Washington, D.C: The World
vol. 344, pp. 1–2, 2004. Bank, 1989.
[30] S. Lee and A. Chib, “Wireless initiatives for connecting rural [49] A. Lavell, “The impact of disasters on development gains: Clarity
areas: Developing a framework,” in Participation and media or controversy Latin American social science faculty and the
production: Critical reflections on content creation, N. Carpentier network for the social study of disaster prevention in Latin
and B. De Cleen, Eds., ICA 2007 Conference Theme Book. America,” presented at the International Decade for Natural
Newcastle, United Kingdom: Cambridge Scholars Publishing, Disaster Reduction Programme Forum, Geneva, 1999.
2008. [50] L. Pyles, “Community organizing for post-disaster social
[31] R. R. Dynes, Organized behavior in disaster. Lexington, development,” International Social Work, vol. 50, pp. 321–333,
Massachusetts: Heath Lexington Books, 1970. 2007.
[32] L. Palen and S. B. Liu, “Citizen communications in crisis: [51] F. Ellis, Rural livelihoods and diversity in developing countries.
Anticipating a future of ICT-supported public participation,” in London: Oxford University Press, 2000.
Proceedings of the Special Interest Group on Computer-Human [52] C. Ashley and D. Carney, Sustainable livelihoods: Lessons from
Interaction (SIGCHI) Conference on Human Factors in early experience. London: Department for International
Computing Systems, San Jose, California, 2007, pp. 727–736. Development, 1999.
[33] J. Powell, An introduction to the natural history of disaster. [53] T. Newholm, K. Keeling, P. Mcgoldrick, L. Macaulay, and J.
Baltimore: University of Maryland, Psychiatric Institution, 1954. Doherty, “The digital divide and the theory of optimal slack,” New
[34] L. Buzna, L. Issacharoff, K. Peters, V. Rosato, and D. Helbing, Media and Society, vol. 110, pp. 295–319, 2008.
“How can future ICT enhance disaster relief and recovery?” [54] Center for Media Studies, Final Report: Evaluating ICT-4D
presented at the Information Communication Technology-Future project of World Vision India. India: World Vision India, 2007.
and Emerging Technologies Workshop, Dresden University of [55] F. J. Proenza, “ICTA Nanasala Program: Field visit report,” 2005.
Technology, Dresden, Germany, 2007. [Online] Available:
[35] Caribbean Information and Communications Technology Virtual http://siteresources.worldbank.org/INTEDEVELOPMENT/Resour
Community, “Facilitating effective disaster management in the ces/Telecenter.doc [Accessed: Aug. 15, 2008].
Caribbean: A response from the Caribbean Information and [56] infoDev, “Sri Lanka-Govi Gnana system project,” 2007. [Online]
Communications Technology Community,” 2004. [Online] Available: http://www.ict4rl.info/CaseStudies/GGS?v=fau
Available: [Accessed: Aug. 15, 2008].
http://www.devnet.org.gy/documents/Caribbeandisasterbrief- [57] United Nations Educational, Scientific and Cultural Organization,
Final.pdf [Accessed: Aug. 15, 2008]. “UNESCO develops community multimedia centre for education
[36] D. Basu, “Disaster relief: A compendium of learnings from in Indonesia,” United Nations Educational, Scientific and Cultural
engagements in Afghanistan, Iraq, Liberia, Iran, Sudan, Organization, 2008. [Online] Available:
Guatemala, Indonesia, Sri Lanka, Pakistan, Lebanon,” NetHope, http://www.unescobkk.org/index.php?id=6800andtx_wecknowled
2006. [Online] Available: gebase_pi1[tt_news]=1215andtx_wecknowledgebase_pi1[backPid
http://nethope.org/doc/NetHope%20Disaster%20Relief%200906.p ]=6771andcHash=96224fb29a [Accessed: Aug. 15, 2008].
df [Accessed: Aug. 15, 2008]. [58] A. Chib, Information and communication technologies for health:
[37] U. Sastrokusumo, “ICT for countermeasures and strategies Aceh Besar midwives mobile-phone project final report, tech.
against emergencies and disaster crisis: A suggestion to build report, World Vision Indonesia Tsunami Response, Singapore,
Asia-Pacific pilot project,” presented at the Second Joint Project 2008.
Team Meeting for Establishing a Disaster Management Support [59] J. Zhao, The Internet and rural development in China: The socio-
System in the Asia-Pacific Region, Bangkok, Thailand, 2006. structural paradigm. Bern: Peter Lang AG, 2008.
[38] L. Palen, S. Vieweg, J. Sutton, S. B. Liu, and A. Hughes, “Crisis [60] A. Chib, M. O. Lwin, J. Ang, H. Lin, and F. Santoso, “Mobiles
informatics: studying crisis in a networked world,” in Proceedings and midwives: Improving healthcare communications via mobile
of the Third International Conference on E-Social Science, Ann phones in Aceh Besar, Indonesia,” Asian Journal of
Arbor, Michigan, 2007. Available: Communication, vol. 18, pp. 348–364, 2008.
http://www.cs.colorado.edu/~palen/palen_papers/palen- [61] United Nations Centre for Regional Development, “Information
crisisinformatics.pdf [Accessed: Aug. 23, 2008]. and communication technologies for human security in local
[39] Caribbean Community and Common Market, “Caribbean ICT development,” Regional Development Dialogue, vol. 23, pp. 65–
disaster management network,” 2004. [Online] Available: 88, 2002.
http://www.devnet.org.gy/documents/Final_ICT_and__DM__Prop [62] A. Chib, “Information communication technologies for
osal_Nov_15.pdf [Accessed: Aug. 15, 2008]. development: Cases from tsunami-affected communities in
[40] B. R. Balachandran and A. Haer, “Corporate involvement in Indonesia and India,” 2006. [Online] Available:
disaster management: A conceptual framework,” presented at the http://www3.ntu.edu.sg/sci/sirc/download/Arul_presentation%20re
Workshop on Disaster Management: Disaster Preparedness port.pdf [Accessed: Aug. 20, 2008].
and Mitigation, New Delhi, India, 2004. [63] Environmental Planning Collaborative, Tsunami evaluation
[41] Y. S. Alghtani and S. Al-Zahrani, “A conceptual framework for coalition: The international community’s funding of the tsunami
information network system to exchange information to control emergency and relief,” 2006. [Online] Available:
accidents in Saudi Arabia,” presented at the Saudi Technical http://www.un.org.in/untrs/reports/local_response_india.pdf
Conference and Exhibition, Saudi Arabia, 2006. [Accessed: Aug. 29, 2008].
[42] Associated Programme on Flood Management, Social aspects and [64] A. Chib, Information and communication technologies for health:
stakeholder involvement in integrated flood management. Geneva: Midwives with mobile-phone project in Aceh Besar baseline
World Meteorological Organization, 2006. report, tech. report, World Vision Asia Tsunami Response,
[43] Earthquakes and Megacities Initiative, “Disaster risk reduction in Singapore, 2007.
metropolitan regions through the cluster cities project,” [65] United Nations, The Millennium Development Goals Report 2005.
Proceedings of the 2005 Americas Cluster Cities Meeting, Bogotá, New York: United Nations, 2005.
Columbia. Available: http://www.emi-
megacities.org/upload/CCP_America_2005_10_Meeting_Report.p
df [Accessed: Aug. 15, 2008].
[44] P. Blaikie, T. Cannon, I. Davis, and B. Wisner, At risk: Natural
hazards, people's vulnerability and disasters. London: Routledge,
1994.
[45] F. C. Cuny, eds., Disasters and Development. New York: Oxford
University Press, 1983.
[46] M. B. Anderson and P. J. Woodrow, Rising from the ashes:
Development strategies at times of disaster. Paris: Westview Press,
1989.
[47] A. Kreimer and M. Munasinghe, Managing natural disasters and
the environment. Washington, D.C: The World Bank, 1991.
337
Abstract— Featherweight multimedia devices combine audio include printed material, radio and television public service
with non-electronic visual displays (e.g., paper). Because of their announcements (PSAs), and verbal instruction. These methods
low cost, customizability, durability, storage capacity, and energy are certainly viable, but suffer from several shortcomings,
efficiency, they are well-suited for education and information such as requiring literacy to comprehend textual printed
dissemination among illiterate and semi-literate people. We
present a taxonomy of featherweight multimedia devices and also
material, the relatively high cost of producing, distributing and
derive design recommendations from our experiences deploying consuming radio and television PSAs, and the arguably higher
featherweight multimedia in the agriculture and health domains cost of verbal instruction by humans who might not even
in India. We found that with some initial guidance, illiterate deliver a consistent message over time. Of these, only verbal
users can quickly learn to use and enjoy the device, especially if instruction provides the possibility of interactive exchange
they are taught by peers. with the consumer.
In this paper, we explore the notion of featherweight
Index Terms—Audio, illiteracy, information dissemination
multimedia” – various combinations of electronic audio
devices with non-electronic visual displays – for interactive
I. INTRODUCTION
multimedia information dissemination (Fig. 1). Featherweight
B. 2nd Prototype: Poster + Off-the-shelf Audio, Mediated C. 3rd Prototype: Talking Book, Non-Mediated
Based on the feedback from the first iteration, we In this third exploration, we look at a more integrated
abandoned the custom audio player and switched to a featherweight multimedia solution: talking books. These
commercially available one (Creative Technologies MuVo devices are relatively cheap, have had some traction in
V100: Fig. 3, right). The commercial device is more durable, developed countries for childhood education, are built robustly
is capable of playing more than 7 audio files (10+ hours), and because they are designed for children, and provide access to
has pause/rewind/fast-forward capability. The total cost of the audio and visuals via a single integrated package.
system was US$38. Being a low-end device, the audio We repurposed a VTech WhizKid [10] device for
player’s screen cannot show the name of more than one audio agriculture information dissemination. The WhizKid can
file at a time. In addition, it does not have playlists and does operate in a standalone fashion, or it can be hooked up to a
not support titles in the local language. The audio player laptop via USB port, in which case the WhizKid tablet and
allows grouping of audio files by folder but the user interface pen are treated as a mouse by the operating system. In the
for switching between folders is difficult to use. latter mode, it is possible to customize the content and
Therefore, we created a numbering scheme where the title interaction completely, so that an inserted page of our own
of each audio clip consisted of two numbers separated by a design can be mapped to audio files we choose. (Should this
hyphen. The first number is a code for the topic, for example solution work, the intent would be to work with the
(1=azolla, 2=vermicompost). The second number is the clip manufacturer to produce agriculture content for the device in
number within the topic. A printed guide was given to the its standalone mode.) The retail cost of the device without the
mediator explaining the code. Navigation is accomplished by laptop is approximately US$25. The WhizKid has another
skipping forwards and backwards through the list of tracks by benefit in that its robust, colorful design is less intimidating
flicking a bi-directional self-centering spring-loaded switch. than a regular PC. During our studies, the laptop was hidden
Since this is cumbersome with a large number of audio clips, so as to not intimidate users. Screen shots from the video were
we only loaded a few topics at a time onto the audio player. used for visual content (Fig. 5), and relevant portions of audio
Posters were designed similar to the first experiment, and were copied from the videos.
342
During the post-experiment feedback session most of the
users were able to answer the factual questions on the
agricultural processes (mean of 75% answers correct for N=4).
This confirmed that most users had been able to retain the key
pieces of information that they had heard.
Crowds immediately gathered when a participant used the
device. Many of the bystanders often prompted the primary
user. Also, there were a number of instances when people who
listened to the audio information enthusiastically participated
in the post-experiment discussion. One user said that she
would be willing to share the device if the device was left at a
common place in the village and remarked that it will be quite
useful on days she missed the Digital Green video-screening
Fig. 5. An audio-augmented page for a talking book
sessions. This suggests that featherweight multimedia can be
We walked around a village on two different days, for a complementary to other forms of information dissemination.
total of six hours, to various households and requested
residents to try out the device in their homes. Besides the
second author, a translator and a mediator were also present.
Each user was given 20 minutes of free use to work with the
device, and was allowed to continue if he/she wished. We
provided no up-front instruction, but if the user failed to
engage with the device at all after five minutes, the mediator
or the translator provided gentle and low-grade assistance.
A feedback questionnaire was administered verbally that
asked for the user’s background information and opinions
about the device. For the second day (N=4), we also
administered a verbal test and feedback session after each use,
to evaluate how much of the presented information the user Fig. 6. A user navigating through the talking book for information.
had learned and to hear comments about the device. D. Discussion
Observations and Feedback: In spite of some initial
These explorations revealed some preliminary insights
hesitation, most of the villagers who tried the device were
relative to our study questions.
comfortable with it and could use it with little or no assistance.
In terms of literacy and necessary user background,
Those participants who took the pre- and post-tests showed
featherweight multimedia appears able to bridge challenges of
increased knowledge of the content presented.
illiteracy. Audio playback seems sufficient for this purpose,
Over six hours spread over two days, we were able to find
and the addition of even static visual images helps anchor
ten people (5 male, 5 female) willing to try the device to learn
discussion in mediated sessions. This is not a surprise, but it is
agricultural content (Fig. 6).
assuring that there were no unexpected surprises with respect
Individuals were very hesitant to press anything at first. The
to adults with little formal education being able to make sense
elapsed time between when an audio clip had finished playing
of these devices. Many of the participants in this phase of the
and the next selection was chosen was initially as long as 10
study had also been previously exposed to agricultural content
seconds. With more use, this reduced to about one second.
on video from Digital Green, yet none explicitly expressed a
Many users repeated aloud certain pieces of audio information
negative comparison with video.
after they heard it spoken from the device.
The simple interface of featherweight multimedia also
Older users were slower to learn the device and slower to seems to reduce barriers for the first-time user. Users related
learn the content – they listened to all the clips at least twice. well to the use of local content and illustrations depicting local
One user kept the pen pressed down while the audio was surroundings and people, enabling viable user engagement. In
playing and never figured out that tapping would suffice. several cases, both in mediated and non-mediated use, it was
On the first day, we had black-and-white printouts of the seen that users repeated pieces of information to re-affirm
video. Color turned out to be essential for some of the audio their knowledge. This process of repeating the narration may
descriptions, and so on the second day, we returned with enhance retention and assimilation of information. For us, it
color-printed sheets. confirms that the users were paying attention in the first place.
Users did not experiment very much. They never pressed Featherweight multimedia also seems sufficiently engaging
the Pause button unless it was pointed out to them. They only so as to enable effective information dissemination, both under
played the audio in sequence, even while listening to the audio mediated and non-mediated circumstances.
captions for the second time to review the material. When In mediated sessions, audiences were not only willing to sit
questioned later, users said that they feared that they might through uninterrupted audio playback of five minutes or more,
spoil the device. in many cases, they requested repeated playback of the audio.
343
In non-mediated situations, users expressed initial findings that audio annotations enhanced comprehension of
trepidation handling the device. One remarked that he was drawings provided further justification to our intent to add
afraid he might spoil the device, and others visibly hesitated in audio to these posters.
early interaction, until their confidence grew. In spite of the
initial reluctance and limited direct guidance, most users were
able to engage meaningfully with the device after just a few
minutes of exposure to the device.
During post-experiment feedback sessions, several people
expressed a desire for shorter audio clips that summarized
longer content. On the other hand, longer audio clips engaged
a greater number of passive users, attracted more passers-by,
and generated more discussion because of the clips’ longer
playing time. Whereas longer clips sustain the attention of a
larger group of people, shorter clips cater to an individual’s
short attention span.
With respect to the need for social support, a human
facilitator is essential, at least in the initial introduction of the
device, regardless of the device’s perceived ease-of-use. In all
of the mediated sessions, the audience fully took in the content Fig. 7. Post-operative do’s and don’ts poster. We added numeric codes to turn
it into a “talking poster” when coupled with an audio player.
without concerns about the device. In contrast, users in the
non-mediated sessions needed prompting and encouragement A. Technology
to use the device at all. Based on our agriculture participants’ difficulties with the
The need for such mediation, however, appears to decrease MP3 player interface, we decided to explore a different device
with time and also with group interaction. When crowds for this setting. We wrote a program that displays a large
gathered in non-mediated sessions, the actions of the primary numeric keypad on a HTC Touch smartphone’s touchscreen
user were influenced by people loitering around. They either (Fig. 8). As numbers are typed in, they appear in the textbox at
voiced common requests or prompted him when the user was the top of the screen.
stuck, serving as encouragement. Such voluntary peer
guidance further alludes to the effectiveness of social support.
T he global higher education (HE) scene has seen an increase The paper is structured as follows: section II places FOSS
in the adoption and use of free/open source software within a development discourse and section III introduces ICT
(FOSS). However this presence tends to be restricted to governance in HE. Section IV illustrates the natural fit of
the bottom of the software stack; mainly the infrastructural FOSS within HE with a corollary treatment of developing
components of university computing that are not visible to the country contexts. Section V chronicles a case study depicting
users [1]. On the application side, the debate does not seem to the failed adoption of FOSS in a Gulf university environment.
have evolved beyond the technicalities of FOSS user- In the final two sections, lessons are drawn to help foster the
adoption of FOSS in developing countries’ university
environments.
Abstract— In this study we examine the history and growth of often were transnational expatriate technologists from the
ICTD since the 1990s. We underline the trends defining this developing world.
research field and examine the progress in research areas that By the mid 1990s, there was significant buzz on the role of
have come to dominate discussion in ICTD through a thorough the information technology boom in dramatic global change
literature review of the last decade of ICTD work. In order to [4, 5], and a first generation of ICTD literature discussing the
answer questions pertaining to the rigor, impact and significance
of ICTD, and to compare the expectations and perceived
specific nature of technology projects in development started a
achievements with respect to different development goals, we worldwide interest in the field [6-10]. Around this period,
interview 50 expert ICTD researchers and practitioners. We there was a dramatic rise in the number of ‘ICTD Projects’ –
analyze these results to understand stakeholders' opinions on the i.e., technology projects specifically aimed at creating
past performance of ICTD, both as an academic field and as an developmental outcomes for their recipients. This trend was
area of development practice, and identify defining ideas on the partly driven by a slew of research papers and policy
potential directions for the future of ICTD. This study is work in documents within international agencies [11-13]. As a result,
progress and we have continuing research in this area, the subset by the turn of the century, there were thousands of telecenters
presented here is rigorous and ready for wider discussion. 1 around the world, funded through various sources [14].
Index Terms— ICTD, developing world
By the early 2000s, engineers became interested in ICTD,
not just as a development agenda, but as an area that posed
I. INTRODUCTION
interesting research problems in their own fields of work [15-
T HE Scandinavians started it. It may be a lost piece of
trivia somewhere, but ICTD as we know it began with the
community computer rooms in Scandinavian villages.
19]. This followed the establishment of numerous academic
venues, both as part of existing established forums, and
independent venues specifically for the study and discussion
Starting with the first village (Fjaltring in Denmark), there
of ICTD.
were several small tele-cottages throughout Sweden, Denmark
and Norway by 1985, and it wasn’t until a few years later that
Following this early foundational scholarly and practitioner
Sri Lanka became the first developing country to get its
community technology access center [1]. work, more interesting follow-up work has been done on
emergent areas of ICTD including digital inequality [20-23],
The term ICTD started appearing in academia and industry on technology and sectoral development and macroeconomic
with minor alphabetical variances such as ICT4D, ICT4B change [24-26], on technology and urban change [10, 27], and
(billions), IT4D and so on by the mid 1990s. The years on the ‘potential’ of technology [12, 28, 29]. By the mid
preceding this were formative in the growth of interest in the 2000s, there was introspective work looking at the
subject around the world. The opening up of Eastern performance of the ICTD projects started in the 1990s in India
European economies coincided with phenomenal periods of and elsewhere, and also at the causes and outcomes of the
growth in China and India. The technology boom in the interest in technology and development [30-33]. More
United States featured a large pool of engineers from various recently, scholars have looked back at the growth of ICTD
parts of the developing world. The spillover economic effects through the years by tracing the various stages that ICTD has
of this international technology workforce ranged from moved through in this period [34].
remittances to home countries to the creation of new small and
medium-scale engineering companies [2, 3]. The public Our survey and research here hope to build on such work.
discourse of technology as being an engine of macroeconomic As stakeholders in an emerging field of study that is still in the
growth grew in strength rapidly as the early impacts of process of defining itself and carving a niche, it is critical that
globalization in the tech industry manifested themselves first
we look back at various points in our progress and review the
in the West, and soon thereafter in many parts of the
direction of our growth. Here, we present a fairly
developing world. The international faces of this phenomenon
comprehensive survey of what has happened in the field, and
use the opinion of significant voices from the field to support
1 the literature review, in the hope of shedding some light on the
This material is based upon work supported in part by the National
Science Foundation under Grant No. 0326582. general direction of ICTD work and its validity. We examine
358
the progress in dominant ICTD research areas, we identify the B. Sampling and Recruitment
areas that received more attention in previous research, and We had a selective sample of respondents from a few major
we compare the expectations with the perceived achievements ICTD-related online forums, in addition to other professors
in each of these areas. We also examine questions concerning and researchers that we contacted to fill out the survey.
the rigor, impact and significance of ICTD, and its
appropriateness in serving various development goals.
C. Sample description
II. APPROACH Our total sample included 50 respondents, who were asked
to select their areas of affiliation. We observe a fairly even
We begin our study by categorizing some broad domains of distribution of these areas, with a slight skew towards
ICTD research and practice, and by performing a high-level education and infrastructure. Overall, a higher number of
overview of the progress in each of these domains. We follow researchers than practitioners were represented in this survey.
this by presenting the results of our interviews with ICTD
experts on their opinions about research and practical work in TABLE I
RESPONDENT DESCRIPTION BY AREAS OF AFFILIATION
each of these areas.
Domain Area Research Practice
In addition to these discussions, we consider a few
recurring themes in our early conversations with interviewed Healthcare 11 3
researchers. The first emergent theme was the question of Education 17 7
‘Hope v/s Hype’ in the context of ICTD. This issue, has been Business 8 4
explored in the past [35-37], but continues to remains a Agriculture 8 3
niggling issue for most practitioners and researchers alike, Comm. & network infrastructure 16 4
who face up to it in their own work. A related theme that User interface design 11 0
emerged in our interviews has been that the role of various Governance 10 1
stakeholders. As ICTD moves from being an experimental Other 9 3
area towards mainstream development research and practice, it
is inevitable that we face hard questions about what the role of
government should be in funding ICTD projects, what kind of The respondents’ past academic disciplines (regardless of
ICTD projects are more likely to succeed and so on. Posing whether they are presently scholars or practitioners) were
these questions to our interviewees, we find an interesting fairly evenly distributed between science/engineering, social
variance of opinion on this issue. Finally we explore issues sciences, and hybrid areas such as design and information
about the interdisciplinary nature of ICTD – how far does it studies.
exist, and has it helped or impeded ICTD.
TABLE II
RESPONDENT BY AREAS OF PRIMARY ACADEMIC SPECIALIZATION
III. METHODOLOGY
The research consists of two parts – first, an extensive Academic Discipline Number of
literature review of the various projects in ICTD over the past Respondents
decade, and second, a survey of 50 researchers and Engineering 16
practitioners in the ICTD space. The two were done Information Studies 7
simultaneously – i.e. while we had some idea of what the Education 3
broad literature review would bring up, we had no idea of how Political Science 4
similar it would be to the results from our interviews. Design 2
City planning 2
Sociology 2
A. Instrument Design
Humanities 2
The interview schedule was created over three iterations on Media 2
the basis of feedback about the interview questions from International business and relations 2
researchers in our direct contact. The third version of the ICTD 2
questionnaire was an online hybrid with a mix of open-ended Development Studies 1
and close-ended questions, of which the former were created Anthropology 1
based on categories arrived upon in the first two iterations of Agriculture 1
the instrument. The final interview was anonymized and took Commerce 1
anywhere between 15-75 minutes to complete, depending on Physics 1
the interviewee. Environmental Science 1
Total 50
359
D. Caveats IV. FINDINGS
The findings are from a fairly small sample because we
wanted experts with several years in the field to comment on A. Healthcare
ICTD issues rather than a broad-based survey of a large The interest in using technology to widen access to
number of participants. Over half of our sample is represented healthcare and sanitation pre-dates the ICTD age. While
by people who have been active in ICTD for over 5 years and earlier initiatives focused on increasing access to specialists
are thus familiar with the landscape and the changes in ICTD and getting basic diagnosis to remote regions within the
and are qualified to comment on past and future issues. That developed world [38], recently, the UN Millennium
said, such a small and selectively sampled survey is prone to development goals (MDGs) have renewed the focus on ICTD
bias. This study is a work in progress, and while the results [39, 40]. There are four broad areas where active ICTD
presented here are rigorous and sturdy enough for discussion, projects tackle healthcare challenges. The first has been
there is always scope for increasing the sample size, and telemedicine [41], and here the focus has been on using long-
finding out through an iterative process what areas might need distance communication links to expand access to remote rural
better coverage in the questions. areas where there are no doctors. For-profit initiatives have
been deployed for specific kinds of diagnoses that can be
Finally, one other important caveat here is the regional
reasonably well managed remotely, such as ophthalmology
concentration of respondents. We find a very heavy skew
and dermatology [42]. 2 The second area of ICTD interest in
towards respondents currently based in the US and India, but
healthcare is information gathering especially for
this is explained by the fact that US is somewhat of a hub for
researchers and practitioners, and that a large fraction of the epidemiological research. There have been a number of
ICTD activity is located in India. This also seems to indicate projects focusing either on general population health
that academics from developing countries are going abroad to surveying, patient health monitoring, or healthcare aid impact
study ICTD in American or European Universities. assessment [43, 44]. A third focus of ICTD was driven by
expatriate doctor communities from developing countries who
TABLE III were interested in contributing back to their home countries
LOCATION OF RESPONDENT AND REGIONAL FOCUS through social initiatives. Some ICTD projects have used web
Country Physical Location where 2.0 technologies to connect doctors in the developing world
location of respondent with counterparts and experts in the developed world for
Respondent primarily active remote consulting on specialty care [45]. The fourth area of
in ICTD work ICTD in healthcare has concentrated on the use of technology
USA 25 4 in building low-cost medical diagnostic devices (such as
India 10 20 ultrasound, X-ray machines) and sensors. 3
Malaysia 2 1
Philippines 2 2
We asked respondents in the study to name the areas of
Barbados 1 1
ICTD and healthcare that they felt were of significance
Brazil 1 1
looking into the future. The top 5 responses, from a subset of
Ghana 0 2
Botswana 0 1 36 who commented on healthcare, are presented in the
Chile 0 1 following table:
Colombia 0 1 TABLE IV
Canada 1 0 AREAS OF FUTURE IMPORTANCE IN HEALTHCARE
Greece 1 0 Top 5 areas ranked by respondents Percent
Netherlands 1 0 Respondents
Nigeria 1 0 Medical records 57.8
South Africa 1 1 Supply-chain management 50.0
Spain 1 0 Tele-diagnosis and treatment 44.7
Sweden 1 0 Collection of epidemiological data 44.7
Switzerland 1 0 User interfaces 28.9
Uganda 1 0
Macedonia 0 1
Nigeria 0 1
No Specific Region -- 12 From the interview results, it stands out that supply chain
Total 50 50 management in healthcare and the design of user interfaces for
2
http://www.clickdiagnostics.com
3
Several of the Microsoft Research Digital Inclusion grantees in 2005
were doing some work in healthcare, at least two were specifically designing
low-cost diagnosis devices for developing countries.
http://research.microsoft.com/ur/us/fundingopps/RFPs/DigitalInclusion_2005
_RFP_Awards.aspx
360
healthcare work were both seen as very important areas for agriculture closely matched past work on agriculture within
future effort, though both of these are practically absent in our ICTD, and most respondents were in fairly good agreement
literature review of past work. Medical records, collection of about the perceived progress in these areas. What is worth
epidemiological data, and tele-diagnosis, all areas with noticing here is that ‘access to expert information’ which had
important past work were also viewed as very relevant by been one of the most important areas of concern in the early
respondents. In addition to these, a number of respondents days of ICTD takes a lower position, whereas market access
mentioned health education and emergency assistance for and information, despite the apparent challenges in making
locating medical facilities are potentially strong areas for the market information usable for actual transactions, remains a
future. There was surprisingly very little mention of low-cost top concern for ICTD experts.
diagnostic devices.
C. Education
B. Agriculture
Governments, philanthropic efforts, and private
From the earliest ICTD implementations in developing
corporations have all found the ICT in education space to be
countries, the apparent incongruity of computers in the rural
attractive right from the earliest days of ICTD
hinterland has been a key concern for researchers. We see
implementations. Many of the early ‘low-cost computing’
therefore that many early projects tried to increase the
projects such as the Hewlett Packard’s 4-4-1 computer, 4 and
relevance of computing in rural areas by providing
the Computador Popular were aimed at classroom use [58].
information on agricultural practices, market prices and
Computer aided learning projects have been among the largest
government schemes through telecenters [46, 47]. However,
and most prominent project categories within the ICTD space
persuading rural adults to be regular telecenter users has been
in the last decade, and today, it is arguably true that more poor
a challenge. As a result, besides agriculture, many telecenters
households have “access to technology” through a child in
have also focused on other services such as online assistance
school than through a kiosk or any other means of computer
to help small enterprises in villages, for instance, artisans that
access [59]. Impact of computers on learning is an area of
sell handicrafts on the Internet [48]. There has also been work
ICTD that has seen interest from mainstream economics as
on turning telecenters into points of purchase in supply chains
well [60]. Interface designers have been actively looking at
for rural produce [49].
innovative learning methods with computers and a lot of
interesting work has been done in shared computing [61],
Other kinds of ICTD projects in the rural agricultural
cellphones for game-based learning [62], and networked
market include the use of cellphones and PDAs in organic
systems for contextual classroom video instruction [63].
certifications for farmers [50], sensor networks in helping
In higher education, distance learning in developing regions
water management for rural areas [51], livestock management
has been an early area of interest within ICTD [64, 65]. While
[52], price information [53] and farming extension and
information sharing [54]. With the growing interest in micro- a significant chunk of the work with children has been India-
entrepreneurship, the use of technology to increase access to centric, we find a lot of work in higher education elsewhere,
microfinance has been very prominent in ICTD research. including the use of SMS for question/answers in university
Many projects have attempted to implement technological lectures [66] in Bangladesh, cellphones for interactive
solutions to assist on-the-ground rural microfinance operations learning [67] in the Philippines, and a one-mouse-per-desk
[55, 56]. Work in social sciences has examined whether the approach for lectures [68] in China.
ability of technological interventions in increasing the
efficiency of rural microfinance operations depends largely on TABLE VI
AREAS OF FUTURE IMPORTANCE IN EDUCATION
the scale of operations of the organization adopting the
Percent
technology intervention [57]. A survey of respondents on their Top 5 areas ranked by respondents
Respondents
opinion about the key areas in agriculture revealed the
following: Remote learning 31.4
Educational games 25.7
TABLE V Low cost computing 22.8
AREAS OF FUTURE IMPORTANCE IN AGRICULTURE Life-long learning 17.1
Online content 14.2
Top 5 areas ranked by respondents Percent
Respondents
Best practices and information sharing 56.7
There was far lesser agreement in the education space on
Market access and information 51.3
Supply chain management 45.9 the key areas of research among ICTD experts. Two
Sensors 21.6 interesting factors emerged in our discussions. First, that low-
Access to expert information 18.9 cost computing, which has attracted a lot of interest in recent
years, is seen comparatively as a less important area of
research than working with existing devices. This mirrors well
Unlike in the case of healthcare, the responses for
4
http://www.hp.com/e-inclusion/en/project/441_brochure.pdf
361
the general frame of discussion within existing ICTD The results in this category could possibly be seen as
literature, which has also largely taken the same position. The underlining the distinction between the interests of researchers
second important issue is that remote learning, in which we and those of the industry. Our sample, despite being open to
have seen some promising projects start in the last few years, researchers and industry, had a much higher research
is expected to continue to be an area of focus in the coming component. As a result, we see that low-cost infrastructure,
years. undeniably a major area of concern for ICTD, trumps the low
cost phones and devices, the area that shows more activity in
our literature review. However, this could also be seen as
D. Communications and Infrastructure indicative of what experts feel about the progress of low-cost
Among the earliest areas to get engineering scholars devices in the real world market. Although we list a large
interested in ICTD was the challenge of building and number of low-cost device projects here, only a small fraction
deploying novel high quality communication network of them have actually made any significant impact in the
solutions to connect low income regions with poor legacy market, and even some of the largest players abruptly left the
infrastructure. Studies suggested that communications market. It is also important that most of the research on low-
backbones could be the first form of infrastructure that would cost devices is on computers, and little or no work exists on
be affordable to deploy in the developing world [15] and that making low-cost mobile handsets, though a number of
this could in turn be a major driver for wider socio-economic researchers indicate that this is likely to change looking ahead.
growth [28]. Several projects looked at various inexpensive One potential reason for why work on cellular infrastructure
connectivity solutions both over the short and the long-haul has remained largely unaddressed in ICTD came from the
[69], and the use of low-cost WiFi for long-distance respondents. Several interviewees stated that since worldwide
connectivity became an important ICTD pursuit, with groups cellular coverage has grown so dramatically over the last few
from several parts of the world looking at the scope of this years, cellular coverage is seen as more of an industry
technology in real world deployments [16, 70-73]. Today, problem rather than an academic research area. Furthermore,
long distance connectivity using WiFi has made significant the closed protocols of dominant cellular technologies are not
progress showing test throughputs of upto 6 Mbps at almost open to experimentation, and thus harder for researchers to
400 km distances 5 and enabling functional field deployments work with.
for telemedicine applications [42, 74]. Research has also
focused on creating appropriate front-end infrastructure to E. Governance
work on low power and unreliable, or intermittently connected
networks [75, 76]. E-Governance was a major buzz area in the early days of the
Finally, the ‘low-cost computer’ has been one of the most ICTD, especially given a general subscription to the view that
important areas of work in technology infrastructure in ICTD, a lot of the problems of development are because of bad
and probably the area that has generated the greatest amount governance [77]. This was especially so in places like India,
of industry interest. We found over 50-projects in the past 10 which had seen sluggish growth in several sectors that were
years in this space including many such as Fulong Mini-PC 6, tightly controlled by the state, but in sectors where the state
and E-DUC 7, Sirius 8, and SofComp 9 that originated from kept itself off, growth came speedily and significantly [26].
research labs in the developing world alongside work by big Consequently, several projects emerged throughout the late
corporations such as Intel’s Classmate 10, and AMD with its 1990s to increase the use of technology in the processes of
PIC 11. governance, first in the developed world and eventually in the
TABLE VII developing world [78]. These included projects that migrated
AREAS OF FUTURE IMPORTANCE IN COMMUNICATIONS a number of state functions online as a way to reduce the
manpower cost of governmental transactions [79], enable e-
Top 5 areas ranked by respondents Percent payments [80] and e-voting [81], and presumably also reduce
Respondents
the scope for corruption. A number of projects also
Wireless/Low cost infrastructure 41.7 fundamentally changed certain state functions such as land
Low cost phones and devices 30.6 record maintenance [82].
Mobile phones & phone coverage 22.2
Community radio & TV 22.2 The status on ICTD in governance thus far has been fairly
VoIP 19.4 mixed. Although on one hand some studies have shown
generally positive feedback [83, 84], others have raised
serious questions about their impacts on the disempowered
[85, 86, 87].
TABLE VIII
AREAS OF FUTURE IMPORTANCE IN GOVERNANCE
5
http://radar.oreilly.com/2007/06/wifi-record-range-now-382-km.html Top 5 areas ranked by respondents Percent
6
http://www.lemote.com/ Respondents
7
http://www.e-duc.com/
8
http://www.fiveriverstech.com/sirius.htm Sharing of public information 47.3
9
http://www.ncoretech.com/mobilis/index.html Digitization of records (land, tax) 31.5
10
http://www.classmatepc.com/ Improved transparency, corruption reduction 26.3
11
http://50x15.amd.com/en-us/
362
E-payment and online retailing 23.6 limitations, voice recognition cannot be an effective
E-voting 7.8 replacement for other technologies – so while speech can
work well for small vocabularies and limited UIs, these tasks
can also be done equally effectively by inexpensive visual
The responses from our experts on governance again interfaces or numerical keypads. The applications where
mirrored fairly closely what we found in the literature review speech could really be helpful, for example in complex
– public information, digitized land and tax records were transactions, is also where the technical challenges are
viewed to be the most relevant research areas. There were also greatest.
a small number of respondents who felt that e-voting and GIS-
related work, both of which have been looked at only to a G. ICTD Stakeholders
limited extent, may be among the major areas for ICTD and One of the interesting questions around ICTD from its earliest
governance in the near future. days has been the role of various stakeholders in supporting
such projects. There were a large number of ICTD projects in
F. Design the 1990s, a period marked by the increasing privatization of
The User Interface (UI) design community has been among development funding [97] and a massive boost in Corporate
the major technical drivers of ICTD research. The basic Social Responsibility (CSR) funding, which in turn diverted
assumption with UI work in ICTD has been that most philanthropic funds from companies interested in human
technologies prevalent in markets today are designed with development towards ICTD projects [98, 99]. Because ICTD
high income, educated users in mind, and that re-designing projects are often at the crossroads of business, technology
computers and peoples’ interactions with computers could and human development, many initiatives have been
make technology more accessible for the poor and illiterate supported by large private corporations such as HP, Intel, or
[88]. Among the earliest ICTD projects – the Simputer [89] Microsoft, especially when there is an intersection of their
had an explicit interface design goal to make the computer a business interests with ICTD [100]. What is more interesting
simpler device to use, therefore with few alphabetical keys, is that ICTD projects have frequently been couched in a
and more audio-based interactions. In most of the low-cost discourse of ‘sustainability’ [101] raising questions on
computing projects that followed thereafter [90], including whether these projects should be market driven, and we have
most prominently the OLPC, appropriate UI design was a very seen some research in recent years suggesting that the state
important component [58]. should reconsider spending on ICTD projects, specifically
telecenters [59]. However, these questions are not easy to
Significant design work has also been done on building answer – the nature of development projects requires that a
visually enhanced interfaces for illiterate users [91] and range of stakeholders work closely together, and ICTD is no
speech based systems for agriculture [92] and healthcare [93]. different.
Other design work has looked across domains to redesign
existing devices for new application areas – such as In looking at how ICTD research can be divided up
innovating with small screens on mobile devices for between various stakeholders, we asked respondents to
systematic data collection for healthcare [94, 95], micro- comment on what they felt were important roles for industry-
finance services for the illiterate [55], and audio-visual based based research to play.
English-language education [96] delivered in local languages.
TABLE X
AREAS OF FUTURE IMPORTANCE FOR BUSINESS STAKEHOLDERS
TABLE IX
AREAS OF FUTURE IMPORTANCE IN DESIGN
Top 5 areas ranked by respondents Percent
Top 5 areas ranked by respondents Percent Respondents
Respondents
Microfinance and microcredit 41.7
Voice recognition and synthesis 57.1 Mobile commerce 41.7
Local language software 54.2 Supply chain management 16.7
Translation 20.0 Online commerce 13.9
Accessibility 17.1 Low cost sales devices 13.9
Illiterate-friendly interfaces 14.2
UI design
Agriculture
Education
Business
Energy
Governance
environment
Communication
governance
alleviation
equality
Education for
Eradication of
Sustainable
Gender
Better
hunger
environment
governance
Communication
alleviation
equality
Education for
Eradication of
Sustainable
Gender
Poverty
technologies
Better
hunger
D
pertaining primarily to environmental issues and grassroots
social development. This stereotype misses the reality that
sustainable development and the information society are uring the last decades four themes have emerged as a
operationally interconnected. The two phases of the World
response to the collective concerns of world citizens: peace,
Summit on the Information Society (WSIS) in Geneva (December
2003) and Tunis (November 2005) provide an excellent freedom, development and environment. As a consequence
opportunity to integrate sustainable development principles and this collective concern and the four themes were manifested in
practices into the institutions and policy frameworks that are a political process towards sustainability labeled “sustainable
shaping the information society. While the World Summit on development”. The concept of sustainability lacks a widely
Sustainable Development (WSSD) and WSIS Phase I brought accepted definition. Sustainable development is a wide
these issues to international attention, Southern voices and concept and has over the years been introduced through many
visions are still notably lacking from the debate different definitions. The term “sustainable development” was
As we move into the age of information it is critically popularized by the World Commission on Environment and
important for us to consider the implications of ICT in
Development (WCED), in its 1987 report entitled, “Our
Sustainable Development and vice-versa. This research
undertaking looks at precisely this intersection with its primary Common future”. This report by the Brundtland commission
focus on sustainable communities. Sustainable Communities (named after it’s chair Dr. Gro Harlem Brundtland) provided
development is an evolving discourse. The locally-owned or the most commonly cited definition of sustainable
adapted knowledge of a community is essential for integrated development. The Commission wrote: “Humanity has the
sustainable development, and is becoming a key priority for ability to make development sustainable – to ensure that it
development practitioners. The popularity of the concept of meets the needs of the present without compromising the
communities is growing, and so are ideas on how to support these ability of future generation to meet their own needs”. This
communities with technologies and how these communities might definition has broad appeal and little specificity, but some
use ICTs to support themselves. These issues are pertinent in
combination of development and environment is found in
assessing the real contribution of ICTs to sustainable
development, and therefore merit a closer examination. The
most attempts to describe it. Since the publication of the
paper presents the findings from the study of 9 communities in report, there has been a mass of literature generated in various
the Niger Delta region of Nigeria. The aim of the study was to fields which has resulted in more specific application of the
explore how ICTs may contribute to the social sustainability of concept, such as sustainable agriculture, sustainable
communities in the Niger Delta Region of Nigeria. The study livelihoods, and sustainable transport. Over the years, the
grounds the understanding of ICT usage among indigenous interest and scope of the sustainable development debate has
communities and consumers in the reality of their everyday lives, grown substantially and this has led to an increasing diversity
in order to promote actions for sustainability. of interpretations. Mostly, sustainable development is
modeled on three pillars used to facilitate the comprehension
Index Terms— Communicative Ecology, Information
of the term: the triangle of environmental (conservation),
Communication Technology, Social Change, Sustainable Communities
economic (growth), and social (equity) dimensions. According
to Selman (1996:31), while such universal aspirations may be
helpful, they are insufficient, at the local level, to identify
local needs, choose meaningful targets, and most important,
Manuscript received February 20, 2009. This work is part of a doctoral “harness the energies of local people and organizations and
research. It is funded by a research Grant provided by Akwa Ibom State
Government, Nigeria. aid development of the local society and economy to change
Uduak Okon is a Doctoral Student at Royal Holloway, University of London in ways which are conducive to sustainability”
and is a member of the ICT4D Collective. (Phone: 01784248786; Mobile: There are newer approaches that argue that the point of
07776474379 ; Email: u.akpan-okon@rhul.ac.uk; okonud@yahoo.com).
departure most be the community. Redclift and Sage (1995)
368
argue that if sustainable development is to mean something, communication ecology with a focus on a (new) definition of
then it must be capable of translating into local action. sustainability. It adds to the ongoing dialogue among
Liemgruber and Imhof (1998) go even further by suggesting academics, development practitioners, business leaders and
that the true scale for sustainability is the local level, where government officials, aimed at pinpointing the ways in which
people interact and communicate and where each individual is ICTs can help make the transition to sustainable development
affected by everybody’s actions. Woodhouse (2000) offers a easier, quicker and economically viable. Its focus is on ICTs
similar argument; he states that the concept of sustainability is as means and tools that enable desired changes, since it is
best understood and evaluated on the basis of a sustainable these changes, not ICTs that lead to collective action and
community. He further states that genuine development sustainable development. The presented research is at the base
enhances the sustainability of the community and this does not of ICT for development in developing Countries. A thorough
necessarily involve economic growth. As the debate on understanding of the interaction ICT with the communities’
sustainability matures it is becoming increasing accepted that communication ecology is of primary importance, before the
development cannot be sustainable unless it builds on cultural deployment of any ICT solution. Thereby it will be of use for
traditions (Nurse, 2004) Traditions, norms and customs are all further research and/or development of ICT as part of a
cultural capital that needs to be preserved and passed to future solution for indigenous communities.
generations. Sustainable development cannot be a product
packaged by the international community and delivered to a
local community; it works best when it draws on existing II. CONCEPTUAL FRAMEWORK
community resources and capacity building efforts (Cooper A. Information Communication Technologies (ICTs)
and Vargas, 2004). It relies heavily on community The definition of ICT for this research uses Hamelink’s
involvement and commitment, since sustainable development definition of ICTs: “Information and Communication
is a way of life. Ideas and projects that are developed locally Technologies (ICTs) encompass all those technologies that
often have the greatest staying power because the community enable the handling of information and facilitate different
develops a sense of ownership. Cooper and Vargas also state forms of communication among human actors, between
that sustainable development is both a bottom-up and top- human being and electronic systems and among electronic
down process. Global commitments made by the nations of systems” (Hamelink, 1997:3) and Duncombe and Heeks
the world are necessary but not sufficient to ensure (1999) simplified definition describing ICTs as an 'electronic
sustainability. Hence, the slogan from earlier Earth Days, means of capturing, processing, storing and disseminating
“Think Globally, act locally”. Localizing global aspirations is information'. This includes the “old” ICTs of radio, television
a key challenge for implementing sustainable development and telephone, and the “new” ICTs of computers and mobile
strategies and even more so for developing countries where technology and the Internet.
most people live in rural and sometimes remote communities. Information and Communication Technologies are rapidly
There is a need to ask local questions and focus on meeting consolidating global communication networks and
local needs. This paper makes a distinction between external international trade with implications for people in developing
physical dimensions of sustainability and the internal socio- countries. Despite this there is a worrying lack of empirical
cultural dimension. The latter dimension is mainly concerned evidence or analysis of the actual experiences and effects of
with individual actions, social structures and the social capital ICTs upon poor people’s economic and social development.
of communities. Contrary to what the proponents that The constraints of existing information systems on poor
prioritise environmental concerns believe, within developing women and men and their intersection with ICTs are also little
countries like Nigeria, pressing problems of social exclusion, understood. Little attempt is made by those promoting ICTs
poverty and unemployment are reducing the attention paid to for development to assess their impact on the cultural identity,
environmental problems. Because of these problems, societies the values, and the state of social equity of the less developed
are less willing to accept the structural changes associated economies. Also lacking are analysis of the social and cultural
with shifts towards a more environmentally sound patterns of factors which determine the effective application and use of
consumption. Thus, for countries like Nigeria, a socio-cultural ICTs by developing countries. Social exclusion in the
perspective is integral to all discussions about susutainable developing world cannot simply be resolved by technology if
development. consideration of the factors that can ensure respect for cultural
The aim of this research is to explore from a socio-cultural values, justice, equity and equality in the distribution of
perspective how ICTs may contribute to and support the wealth - including information - is absent. Rodgers et al,
sustainable development of communities in the Niger Delta (1994) argues that by definition, the nature of these elements
Region of Nigeria. This paper outlines the conceptual imposes certain limitations on how ICTs can be appropriated
framework for the study of ICTs and Sustainable communities for development. Even if the introduction of ICTs is feasible
and presents a narrative of the findings of the empirical over a relatively short period of time as some predict, a
research conducted in Nigeria on the Communicative ecology dialogue between those promoting the technologies and the
of Niger Delta communities. This research contributes to the potential beneficiaries must be the foundation of any
understanding of the interaction of ICT with the communities’
369
development action. In the current information age, the According to the organizers of the World Summit on the
capacity of a society to effectively position itself as a Information Society (WSIS), “We are in the midst of a
consumer and producer of knowledge is crucial to its social revolution, perhaps the greatest that humanity has ever
development. ICTs are increasingly playing a crucial role in experienced. To benefit the world community, the successful
most communities’ capacities to produce, access, adapt and and continued growth of this new dynamic requires global
apply information and thus offer enormous opportunities for discussion and harmonization in appropriate areas.”
facilitating the transfer and acquisition knowledge. They Unfortunately, until recently, few discussions have focused on
present (atleast theoretically) a promising potential to lead harmonizing the visions of the emerging information society
developing countries into the ‘highways’ of development. with the principles and priorities articulated by the United
(Friedman 2006; Castells 2000). Despite the potential role and Nations’ Millennium Development Goals (MDGs) and the
contribution of ICTs to development, there is still a growing World Summit on Sustainable Development (WSSD).
voice in the development field raising serious concerns about Meaningful discussions about national policy coherence
the socio-cultural dimension of their application and use between these processes have been limited by restricted
(Panos 1998; Wade 2002; Gumucio 2001). Proponents of thinking about the information society (IS) and sustainable
ICTs (World Bank 2002, UNDP 2001; Pohjola, 2002; Braga, development (SD). Each has emerged from a different
1998) take an optimistic view and highlight the positive community with a different vocabulary and process for
effects of the Internet and other forms of ICTs to create new determining national priorities. Information society specialists
economic, social and political opportunities for developing have primarily been drawn from the fields of
countries and the poor, but it also highlighted the need to ask telecommunications and economic development. While,
both with regards to individuals and society, especially in sustainable development has been delegated primarily to
third world countries, about the nature and extent of the environmental issues despite the best efforts of its
influence and changes, and about the factors which facilitate practitioners to articulate a holistic vision of integrated
or impede these processes, and more importantly how these economic, social and environmental decision-making.
influences and change may promote actions for sustainability. The lack of interaction between these two policy
It has been found more useful to approach the question in communities is currently serving to reinforce stereotypes of
terms of different predispositions and experiences, and with both fields. Sustainable development is seen as pertaining
different historical, ethnic, linguistic, social, economic or primarily to environmental issues and grassroots social
religious backgrounds in different situations, making different development, while the information society is perceived as
use of ICT media (Avgerou, 2001; Walsham, 1993). Taachi et being more relevant to the economic development potential of
al (2003) further argue that it is important to study the urban elites. These stereotypes miss the reality that sustainable
communication needs, communication patterns and network, development and the information society are operationally
and the increasingly important impact which developments in interconnected. Both terms are increasingly used by civil
communication technology might have on these. O’Farrell society and academics to refer to a desired global future that is
(2001) and Heek’s (1999) share the belief that before one can casting its shadow upon our current time and decisions. For
advocate for the development of ICTs among the poor, it is sustainable development to be effective and efficient, it must
important to understand the existing information systems of harness the institutions and tools of the information society.
the poor, how they interact with more formal information and And for the information society to sustain itself, it must pay
the best way to strengthen them before intervening with new careful attention to the stocks and flows of resources (material
information sources and means of access sources. Questions and human) and energy that underpin it.
such as: who benefits and who loses from the introduction of There is now widespread recognition that sustainable
these technologies; how can ICTs be made useful and development is the responsible way forward and it is
meaningful to the developing countries' poor majority who are enhanced by the recognition that information and
struggling to meet their basic needs; what are the social and communication technologies (ICTs) are key drivers of socio-
cultural opportunities and risks they present; and how can economic change and so has to be brought into focus as part
developing countries meaningfully adopt these technologies of the equation. For example, Radermacher (1998) points out
while lessening their undesirable social and cultural that "Modern information and communication technologies
consequences, are some of the questions that emerge when drive the worldwide economic system and the process of
looking at the potential development impacts of ICTs. These globalisation. In this process we see enormous growth
issues are pertinent in assessing the real contribution of ICTs worldwide with opportunities for overcoming poverty and
to sustainable Communities development, and therefore merit promoting human rights but also with major threats to
a closer examination. sustainability and to social justice".
C. Information needs
A number of factors affect how different members of the
community may understand or use information, such as
Okada
gender, economic status, literacy, etc. Each community has
it’s own way of communicating and finding out about what is
going on in their area and outside. The information needs
Fig. 4 Information Flows identified by students and workers were: Latest discovery in
science and technology, Events around the world,
Community leaders filter certain developmental information Entertainment information i.e. music, creative arts, Local and
that comes in and transmit information they consider useful to state news, Health news and information; Social events; Sport
community members. This goes to show the kind of influence news; Business information i.e stock news; Job opportunities
that the leaders have on the community’s access to and Information on community development. Traders and
376
market women are interested in product prices. The elders and desired. Internet use is very limited as stated earlier, because
leadership were more interested in information that has to do cyber cafés are located outside the community and travel costs
with community development and politics. is included with airtime costs. Most people have never used
the internet and can’t use computers but have the perception
VIII. ICT USE AND IMPACT that it is very costly and don’t even try.
The investigations revealed that a combination of old ICTs
(Radio and Television) and new ones like mobile phones, ICT Media Use in Anua
No. of People
and social benefits arising from having access to and using 15
Radio
Mobile Phones
Internet
ICTs. Non-internet Non-internet computing
10 computing
Internet Television
A. ICT Awareness 5
Computers are viewed as a symbol of modernity. A lot of
the older people were quite intimidated by the prospect of
0
1
using one. People under the age of forty said ICTs are Media Used
important for their individual lives and their communities.
Some young people saw it as a tool that helps learning; some ICT Media Use in Mbiaya Uruan
They also thought it can be useful in building capacities and 15 Mobile Phones
Mobile Phones Internet
personal growth and can also provide an opportunity for 10 Non-Internet Computing
wealth creation. Some cited the fact that many people now Non-Internet
Computing
Television
earn an income from running calling centers, so could also be 5
Internet
income generating. These calling centers are usually little
tents that are constructed by the road side where people make
0
1
local and national calls from mobile phones for as little as N25 Media Used
(1p) per minute, and international calls for N75 (3p) per
minute. These calling places are all over the place on almost Fig. 3 ICT use
all street corners. Profits are marginal and competition fierce,
but people eke out constant income from offering these C. ICT Impact
services. This concept explores the interaction between Community
Information Systems and ICT Tools. The study showed that
B. Individual’s exploitation of ICT ICT further reinforces the social structures of these
There were no fixed land lines in any of the communities communities. The greatest area where ICT is affecting
studied, so all calls are made with mobile phones. community systems is in strengthening social networks, social
The graphs below show ICT usage in two of the interactions have been greatly enhanced by using mobile
communities studied. Mobile phones and Radio were the most phones, it is getting more people involved in the happenings
affordable and most used. In more rural communities like in the community. ICT use is also strengthening the sense of
Mbiaya Uruan radio is a major source of Information. There identity people have with their communities, there are
were many people interviewed who owned televisions and indigenes that work and live outside of their communities and
some times because of power outage, would get to watch it radion and mobile phones are keeping them connected to
only once a week, and in cases where the outage is due to happenings in their communities, because as revealed from the
faults, they would go for months without watching TV. Radio findings of the preliminary study Niger Delta citizens feel a
has become a primary source for getting information on local, sense of identity with their communities of origin, and are
state, national and international news and events. Radio is committed to the development of their indigenous
considered cheaper to run because it is battery operated, no communities. In this way ICT is helping to strengthen these
power supply needed, there is a huge supply of radios and low bonds.
level batteries imported from china that is readily available in ICT use is also impacting the development priorities of
most corner shops. Phones are also affordable, portable and community members, because it connects them to the rest of
mobile, although not as cheap as radio but definitely highly the world. This connection has positive benefit for the
communities. They are becoming more aware of global issues
377
like HIV/AIDS, environmental degradation and human rights • Preservation of Indigenous cultures
and have mobilized themselves to address some of these • Job creation and skills acquisition
issues. The empirical data provided possible directions for
ICTs have also had an impact on the security of citizens in exploring how ICTs may be employed to contribute to
these communities. The Nigerian Government is very sustainable development for the Niger Delta people. ICT can
ineffective in protecting their citizens from crime. So citizens play a role in facilitating sustainability by:
have been forced to find ways to fill this gap by defending and 1. Strengthening community systems that enable cohesion
protecting their own communities and ICT is playing a crucial and collective action
role in that. Mobile phones are being used effectively by 2. Providing appropriate information on sustainable
vigilantes for security. development practices through:
The following is a summary of areas where ICT is • Public awareness campaigns and social reorientation on
impacting communities in the Niger Delta development issues;
• Community mobilization • Education;
• Community watch • Provide access to information,
• Circulates information faster 3. Electronically documenting indigenous knowledge and
• Connects people together, Fosters good relationships practices for now and future generations.
• Income generation
X. CONCLUSION
The study shows that ICT is being integrated into the social The study provided a better understanding of how the use
fabric of the community, especially mobile phones. of technology affects relationships within the community and
Computers and the Internet are not as yet having any real how individuals use technology to develop their relations with
impact in the communities. There is still a long way to go others in community.
before ICT can begin to meet the information needs of these The study also showed the importance of mobile phones
communities, because they do not have input on content, this and radio in changing the lives of the people. Mobile phones
is designed by others who may not be sensitive to needs of the is by far the most common communication technology to
people. These initial positive development of ICT use shows effect tangible positive change in these communities (market
that there is great potential for ICT, if appropriately deployed and trading information, emergency and security
to meet community information needs. communications, strengthening kinship relations and social
interactions) and is the backbone of ICTs.
At present there is a vast unmet demand for radio
IX. IMPLICATIONS OF ICT IMPACT ON broadcasting with local content in these communities. A
SUSTAINABILITY combination of mobile technology and rural broadcasting will
Incorporating sustainable development at the local level is enable information and communication services reach more
one the biggest challenges facing the movement towards people than any other medium.
sustainability, how to take global principles and make them Overall, the study has provided a better understanding of
concrete locally. Communities which are culturally rooted, the use of technology in community life and presented
locally produced and technologically adapted are being opportunities of how ICTs may be deployed to help these
rapidly eroded. Vast literature frame the sustainability communities towards sustainability.
discourse largely in terms of an environmental agenda pre-
occupied with ‘green’ issues. Literature has largely neglected REFERENCES
the community development as sustainable and livable places [1] Avgerou, C. (2001), The Significance of Context in Information Systems
which adapt their unique cultural identities and specific and Organisational Change. Information Systems Journal 11:43-63.
historical heritage to contemporary needs. Analyzing the [2] Barton, C. & Bear, M. (1999) Information and Communication
Technologies: Are they the key to viable business development services
cultural environment and information ecosystem is a method for micro and small enterprises? Report for USAID as part of the
of understanding the complexity of a given community, its Microenterprises Best Practices Project. March 1999 by Development
culture, its fragility, resilience and it’s reaction to change. Alternatives Inc, MD, USA
[3] Beck, U. (1992) Risk Society: Towards a New Modernity. New Delhi:
The factors relevant to sustainability of communities Sage. (Translated from the German Risikogesellschaft published in 1986
identified from the research are: [4] Bell, S. (2003). Measuring Sustainability: Learning by Doing. London:
• Awareness about sustainable development; Earthscan Publications Ltd.
[5] Braga, C. P. (1998). Inclusion or Exclusion, Information for
• Information availability and thus transparency; Development (InfoDev), The World Bank,
• Public participation in governance; http://www.unesco.org/courier/1998_12/uk/dossier/txt21.htm
[6] Castells, M. (Eds.) (2000) The information Age: Economy, Society and
• Empowerment of citizens, especially women;
Culture, Oxford: Blackwell, 3 volumes
• Fostering of cultural diversity; [7] Cohen, D., and Prusak, L. (2001). In good company: How social capital
• Building capacity makes organizations work. Boston: Harvard Business School Press.
[8] Chambers, R. (2003) Whose Reality Counts? Putting the First Last.
• Social cohesion Intermediate Technology Publications, London.
378
[9] Commoner, B. (1993). Population, Development and the Environment: [34] Redclift M. and Sage, C (Eds) (1995) Strategies for Sustainable
Trends and Key Issues in the Developed Countries. In Population, Development – Local Agendas for the South. John Wiley & Sons,
Environment, and Development. New York: United Nations. Chichester
[10] Cooper PJ. and Vargas CM, (2004) Implementing Sustainable [35] Rodgers G. et al. Social Exclusion: Rhetoric, Reality, Responses,
Development: From Global Policy To Local Action. Rowman and International Institute for Labour Studies, Geneva, A contribution to the
Littlefield, Lanham, MD World Summit for Social Development, 1995.
[11] Duncombe R., and Heeks R. (1999) Information, ICTs and Small [36] Selman, P. 1996. Local Sustainability: Managing and Planning
Enterprise: Findings from Botswana. Manchester: University of Ecologically Sound Places. New York: St.Martin’s Press.
Manchester, Institute for Development Policy and Management, [37] Slater, D., Tacchi, J and Lewis, P. 2002. Ethnographic Monitoring and
Working Paper 7.<http://idpm.man.ac.uk/idpm/diwpf7.htm> Evaluation of Community Multimedia Centres: A study of Kothmale
[12] Felleman, J. (1997). Deep Information: The Role of Information Policy Community Radio and Internet Project, Sri Lanka. London: DFID.
in Environmental Sustainability. Greenwich, CT: Ablex Publishing [38] Slim, H. & Thompson, P. (1993) Listening for a Change: Oral History
Corporation and Development. London. Panos Publications.
[13] Flamm K. (1987) Targeting the Computer, New York, the Brookings [39] Stacey, M. 1969. The Myth of Community Studies, British Journal of
Institute Sociology, 29, 134-147
[14] Friedman, T.L. (Eds.) (2006) The World is Flat: The Globalized World [40] Taachi, J. and Slater, D 2003. Modernity under construction:
in the Twenty-First Century, UK: Penguin Comparative ethnographies of internet. Amsterdams Sociologisch
[15] Giddens, A (1984) The Constitution of Society: An Outline of the Theory Tijdschrift 30, No. 1-2 Special Issue Digital contact. pp 205-222
of Structuration. Polity, Cambridge. [41] Tacchi, J , Slater, D., and Lewis, P. 2002. Ethnographic Monitoring and
[16] Gumucio DA. (2001), Making Waves: Stories of Participatory Evaluation of Community Multimedia Centres: A study of Kothmale
Communication for Social Change, Rockefeller Foundation. Community Radio and Internet Project, Sri Lanka. London: DFID.
[17] Hamelink C.J (1997) New Information and communication [42] Tacchi, J, Slater,D and Hearn, G. 2003 Ethnographic Action Research: a
Technologies: Social development and Social change, Discussion paper User’s Handbook (New Delhi:UNESCO,
No.6, Geneva: UNRISD. http://cirac.qut.edu.au/ictpr/downloads/handbook.pdf
[18] Heeks R (1999) Information and Communication Technologies, Poverty [43] Tönnies, F. 1957. [Gemeinschaft und Gellenschaft] Community and
and Development. Development Informatics: Working paper Series. Society. East Lansing, MI, Michigan State University Press.
Paper No 5, June 1995 Institute of Development Policy and [44] UNDP (2001), Making new technologies work for human development,
Management. Published on the www: http://www.man.ac.uk/idpm/idpm Human Development Report 2001, UNDP, Oxford.
[19] Heeks, R. (2001) What Did Giddens and Latour Ever Do For Us? [45] Walsham, G. (1993) The Emergence of Interpretivism in IS Research
Academic Writings on Information Systems and Development', Information Systems Research 6(4): 376-394.
Information Technology in Developing Countries , 0, Vol.11(1) [46] Wade, R. (2002) Bridging the Digital Divide: New Route to
[20] Hilty, L. M., Seifert, E. K., Treiber R. (2005) Information Systems for Development or New Form of Dependency? In Global Governance,
Sustainable Development, Swiss Federal Labs for Materials Testing & Vol.8, No. 4 Oct. Dec.2002
Research, Switzerland; Wuppertal Institute, Germany; Hochschule [47] Woodhouse, P (2000) Environmental Degradation and Sustainability. In
Niederrhein, Germany. Allen, R &Thomas, A (Eds) Poverty and Development – Into the 21st
[21] Hoffman, K. (1985) Microelectronics, International Competition and Century. Oxford University Press, Oxford
Development Strategies: The unavoidable issues- Editor’s introduction. [48] World Commission on Environment and Development. (1987). Our
World Development 13(3): 263-272 Common Future. The Bruntland Commission, UNESCO.
[22] Huysman, M., and Wulf, V. eds. (2004). Social capital and information [49] WSIS (World Summit on the Information Society) Civil Society Plenary
technology. Cambridge, MA: MIT Press. (2003) Civil Society Declaration to the World Summit on the
[23] Kling, R. (2000) Learning about information technologies and social Information Society
change: the contribution of social informatics, The Information Society , <http://www.smsitunis2005.org/plateforme/pdf/civil-society-declaration-
Vol. 16 No.3 en.pdf>.
[24] Korten D.C (1995) When Corporations Rule the World, Kumarian Press.
[25] Liemgruber W. and Imhof G. (1998) Remote Alpine Valleys and the
Problem of Sustainability. In Andersson, L & Blom, T (Eds)
Sustainability and Development - On the Future of Small Society in a
Dynamic Economy. Universit of karlstard research report 98:8
[26] Madon, S. (1997) The Information-Based Global Economy and Socio-
Economic Development: The Case of Bangalore, The Information
Society, 13 (3)
[27] Manning E.W (1990 Conservation strategies: Providing the vision for
sustainable development/ E.W. Manning Environment Canada, [Ottawa,
Ont.] : Alternatives 16, 24–29
[28] Nurse K, 2004, Culture as the Fourth Pillar of Sustainable
Development.
www.fao.org/sard/common/ecg/2700/en/Cultureas4thPillarSD.pdf.
Assessed 15/01/09
[29] O’ Farrell, C. (2001) Information Flows in Rural and Urban
Communities: Access, Processes and People. IRDD, The University of
Reading .
[30] Pamlin, D, ed. (2002). Sustainability at the Speed of Light. Solna,
Sweden: WWF Sweden, [Online] Available:
http://www.panda.org/downloads/general/ict_sustainability.pdf.
[31] Panos (1998), The Internet and Poverty. Panos Media Briefing, No. 28,
Panos Institute, London.
[32] Pohjola, M. (June 2002). The New Economy: Facts, Impacts and
Policies, in Information Economics and Policy.
[33] Radermacher F.J (1998) Intelligenz - Kognition - Bewußtsein:
Systemtheoretische Überlegungen, technische Möglichkeiten,
philosophische Fragen. In: Interdisziplinäre Beiträge zur
Kommunikation und zum Mensch-Technik-Verhältnis (C. Stadelhofer,
ed.), Band 6, S. 146-193, Kleine Verlag GmbH, Bielefeld (1998)
379
I
using various health program specific reporting formats and
N this article we describe and discuss a project to develop systems are all being captured in one database application, the
an integrated Health Information System (HIS) in Sierra District Health Information Software (DHIS version 2, from
Leone, which has been going on since early 2007. The article now referred to as DHIS). Unified processing and analysis of
focuses on the key challenges facing the project; these previously disparate information flows have then made
1) Building agreement on the need for an integrated it possible to better assess the quality of data and
approach to HIS in order to solve the prevailing fragmented discrepancies between the different data reporting formats. As
situation of multiple vertical reporting systems which have no a result overlaps, gaps and inconsistent definitions of data
coordination and shared data standards – and building a variables between the different reporting formats have been
consortium to carry out the task. identified. As a result of this, a revision of the data collection
2) Handling the extremely poor infrastructure in Sierra tools within an integrated framework is being planned.
Leone by establishing a “new” solar powered computer Furthermore, the district based database application supports
infrastructure. unified data management, data disaggregating (e.g. make it
3) Developing a free and open source database application possible to “follow” the data from the national level down to
which may handle the problem of interoperability with the health facility where it is collected), data analysis,
existing systems and include and manage data from the programmatic reporting, changes in health units organisation,
various health programs and data sources. and local system integration.
4) Capacity development; how to establish a training
scheme and a support structure sufficient for rendering the A number of paper based reporting formats and routines are
new computerised system sustainable. currently in place in Sierra Leone Some of these are also
computerized, such as HIV/AIDS reporting, which consists of
Ravaged by a prolonged civil war, which was declared over paper forms reported from the facilities, and a database
on 18 January 2002, Sierra Leone consistently scores application called Country Response Information System
extremely low on human development indicators. The public (CRIS, from UNAIDS) located in the districts, where the
HIV/AIDS reporting forms are captured. One of the goals is to
.
380
establish interoperability and data exchange between DHIS be more or less integrated, or aligned. Communication and
and CRIS and other software. In order to achieve this interaction between the components of the network may be
integration, many other critical issues such as infrastructure regarded as going through gateways. Integration can then be
that the system will rely on and human capacity need to be perceived as software exchanging data through gateways, but
addressed. The national power supply system is extremely this can also mean integration between paper based and
poor, as many other national infrastructures. The main power computer based components. An integrated HIS allows data to
source of the existing computerized HIS are generators. As the be collected and analysed at one point and to be shared among
government is not able to provide fuel regularly to run the modules or parts of the system. This makes the data
generators, this reliance on diesel contributes to undermining analysis more meaningful as all data is analysed together. It
the system. Human capacity both in terms of data analysis and also entails knowledge integration as well as solving
information use and IT skills is another key issue that has to challenges of political and economical nature [5], [3]. This
be tackled. Many information/IT officers (called M&E process has shown to be non-trivial [6], especially in
officers; Monitoring and evaluation) at district level have developing countries [3]. The fragmented HIS leads to poor
received on-the-job IT training, but because of power quality health information which can be rendered useless.
shortage, computer troubles (viruses, damaged and old Integrated HIS is therefore relevant, but it does not solve all
computers) and absence of follow-up, the training endeavors the problems of HIS in developing countries.
were not capitalized on and they are still unskilled or
uncomfortable in IT. III. METHODOLOGY
The research enveloping the project in Sierra Leone has
The rest of the article will present the case of Sierra Leone been carried out along the lines of Action Research (AR). AR
by focusing on the challenges described above. Some is a form of participative research where the researcher takes
background to HIS and integration, as well as the methods part in the change processes in an organization, actively trying
applied, will be described first. to improve some stated problem [7], [8]. While it was initially
not used much in IS research, and had a lot in common with
II. BACKGROUND: HIS AND INTEGRATION anthropology and social studies, it has over the last two
Fragmentation and poor quality and use of data are major decades increasingly become accepted as a way to generate
problems with health information systems in African very relevant research findings on technology in its human
countries. Integration of information systems is often context [9].
perceived as a technical task involving primarily incompatible
software and infrastructures. While this is an important part of While we acknowledge the critique often raised against
the picture, fragmentation and poor coordination between using AR in the field of IS (or any field, for that matter), we
organizations, and, as in the case of health, between providers also claim that we, through our active participation, gained
of different services, together with political and social aspects knowledge we would not easily come by taking a more
more generally, are as important [3]. This fragmentation leads passive role. Baskerville and Wood-Harper provides an
to gaps and overlaps, and incompatible definitions in the data overview over common critiques of AR. However, they
that is being collected, registered and reported. Experiences conclude that the traps found are not unique for AR as a
from other African countries, as well as from Sierra Leone, specific method, but are likely to trouble a researcher using
show that without shared data standards, data exchange cannot any method in social science [9]:
take place. Integration is therefore first of all about data
standardization. However, while the practical system focus of “A number of problems confront the action researcher such
integration is on standardization, the overall integration will as lack of impartiality, lack of discipline, confusion with
need to involve and enroll as many as possible of the various consulting and its context-bound nature. However, these
actors in the health system in the process. problems confront researchers using alternative methods as
well. The difficulty with action research may be a matter of
Over the last decade the Health Information System degree, and the easy loss of scientific rigour” (ibid, 144)
Programme (HISP, www.hisp.info) has addressed these
problems. The DHIS software was successfully used for The authors have been an active part in the consortium
integrating data and health services in first South Africa, and described in this paper. Through this role, we have been able
later in many other countries [4]. Integration of data and to engage in a close collaboration with users at different
interoperability between information systems as well as levels. Our findings originate in analysis of our work,
increased use of information are key issues in the HISP including software development and customization, database
approach. development, implementation of pilot sites in Sierra Leone,
training of all monitoring and evaluation officers in the
The health information infrastructure may be regarded as country over 6 weeks, on-site training in the pilot districts, and
heterogeneous networks of actors (Latour 1987), which may discussions with district personnel to clarify data flows,
381
reporting, work practices, and the like. The scientific rigour, as interviews and observationary techniques, and carefully
to quote Baskerville and Wood-Harper, has been upheld by recording the findings.
applying qualitative research methods in the daily work, such
Exce Exce
Data Exce
CRIS
District
Fragmented
Data management SU Data
ICS Exce
Exce
Data
Summary
reports
Fig. 1. The previous Health Information System in. Sierra Leone was characterized by fragmentation and overlap between data collected by the different actors
National level
1. Export data
District level Captured by
Dissemination DHIS
in districts; Other
Gateway: Epi-Info
Reporting, DHIS CRIS Data extract ICS
Services
statistics
indicators & integrate
M&E, GIS/ Maps Data Feedback 2. Import data Data
capture Captured by capture
ICS Epi-info etc
ΔQ /Q
eY =
ΔY /Y
Where: Q = quantity demanded; Y = income and, ey =
income elasticity.
The first derivative of the regression equation for income
yields the coefficient for Y, which is the income elasticity, that
is, the percentage change in mobile expenditure at a 1%
change in income. 23 The estimations results show that income
variable is statically significant and positive for all 17
countries.
23
See Wooldrige (2006) pages 46 and 705, Hej et all (2004) page
296.
399
price elasticity will be relatively low (much less than -1). But,
as prices fall, in a much cheaper service environment, she
argues that people will start using the phone for many non-
essential purposes; ranging from relationship maintenance and
simple conveniences. Added to this, pricing innovations that
enable poorer people to purchase small denomination pre-paid
calling cards to make phone calls it is intuitive that the number
of calls made will increase significantly. Milne (2006) argues
that when this happens, the price elasticity will rapidly rise
towards -1 or even more than -1.
One of the biggest problems in estimating price elasticity of
demand (PED) for telecom services in developing country
situations is the lack of accurate usage data. Unlike in
developed countries where usage is easily obtainable from
monthly bills for post-paid connections, the vast majority in
developing countries, in the case of the current study, between
78% and 99% have no billing records (for prepaid shares see
Table 9). Many use public phones, phones at work or other
people’s phones which equally does not generate records.
ΔQ /Q
eY =
ΔP / P
Where: Q= quantity demanded; P = price of the good and ep
= price elasticity.
The price elasticity of demand is in almost all cases, except
for inferior goods, negative, if the price increases the demand
decreases. To simplify the notation, demand elasticities are in
general expressed in absolute value and therefore expressed as
positive (Frank & Bernanke, 2003). Figure 3: If calls were cheaper what would you do? (source: RIA
2007/2008 household survey).
Milne (2006) provides a very intuitive argument for the
changing structure of price elasticity in a hypothetical case.
She argues that when phones become available for the first
time in a low-income community and usage charges or tariffs
are relatively high (in terms of income) people will make
limited number of essential calls. In such a scenario, even if
the prices falls by a small amount the number of such calls; for
instance in an emergency or substituting a telephone call for a
bus ride to a city, will not increase by much. In other words
400
could be reached with US$1, US$2 and US$5 Average
Revenue per Users (ARPU) business models.
VII. CONCLUSION
This paper shows that countries differ in their levels of ICT
adoption and usage and also in the factors that influence
adoption and usage. Income and education vastly enhances
mobile adoption but gender, age and membership in social
networks have little impact. Income is the main explanatory
variable for usage. In terms of mobile expenditure the study
also finds linkages to fixed-line, work and public phone
usages. These linkages need however be explored in more
detail in future. Mobile expenditure proofs to be inelastic with
regard to income, ie as income increases mobile expenditure
increases to a lesser extent indicating its importance in
individual budgets.
Key policy interventions would be regulatory measure to
increase access and usage of mobile services by promoting
network investment. The current super profits being enjoyed
by operators across the continent as a result of prices that are
amongst the highest in the world need to be moderated. In
most markets there is not effective competition requiring
regulatory interventions to ensure cost based wholesale and
retail pricing.
Increasing Coverage: Licence fees for infrastructure
investment should cover the administrative and usage costs of
national resources only. The current high cost of licences in
most countries used to generated funds for state coffers
Figure 4: Average reaction to cost of using your phone came
translate in high prices for consumers as operators recover
down by half or going up double (source: RIA 2007/2008
household survey). their licence costs. Countries need to look at incentives to
encourage networks rollout not disincentives.
Another difficulty is that cross section data, like the one that Wider Access: Access prices can be reduced by exempting
is being analysed here, is not suitable to calculate price telecommunication equipment, in particular mobile handsets
elasticities since it only captures data for one snapshot in time. and services from import duties and additional taxes.
Changes in consumer behaviour due to price changes can only Allow more Usage: Policy measures to increase the
be analysed using time series data. Cross section data only competition within the industry are the best mechanism to
allows a rough glimpse at price elaticities by using contingent reduce usage costs. In a competitive environment the operator
valuation methods. Respondents with mobile phone or active can choose to compete on price or on service quality.
SIM card were asked how the would react to price decreases Regulatory measure can improve competition on price by
and increases. The answers to these questions are being creating price transparency.
displayed in Figure 3 and Figure 4. Figure 3 shows that the Price Transparency: Contract and prepaid tariffs vary across
vast majority of respondents would make more calls if prices operators to an extent that a product to product comparison is
would come down. Senegal and Ethiopia are the two countries impossible for consumers. Non- transparent pricing, including
with the highest share of respondents that would use the saved rapidly changing promotions and misleading media
money for something else. campaigns, make it difficult for end-users to make informed
Figure 4 shows how respondents state that they would react decisions. This prevents operators from having to compete on
to a doubling and halving of call charges. What can be seen is prices and leads to higher average prices. Establishing price
the response is asymmetrical, ie respondents would react to transparency on the basis of published monthly usage baskets
price increases stronger than they would do to price decreases. will make the cost to the end-user transparent and force
However, approximating price elasticity from these two operators to compete on price or service quality.
contingent valuations results is strictly speaking not possible. Cost-based Interconnection rates: Above cost
Respondents state their preferences and do not reveal it interconnection charges are often used by dominant operators
(observed behaviour reacting to price changes). to restrict small operator from gaining market share. High
Table presents data on the untapped market, those 16 years interconnection charges make off-net calls expensive and
or older that do not have a mobile or active SIM card at punishes consumer or small operator. Cost based
present. It shows the results for the number of people that interconnection rates (or benchmarked on cost base) increase
the competition between operators and hence lead to lower
401
usage cost. [16] Garbacz, C., & Thompson, H. G., Jr. (2007). Demand for
telecommunication services in developing countries.
Licences for Low ARPU business models: Regulators Telecommunications Policy 31 (2007) 276–289.
should award additional licences to applicants that are able to [17] Gasmi, F., Laffont, J. J., & Sharkey, W. (2000). Competition, universal
offer low ARPU business models, preferably below US$2. service and telecommunications policy in developing countries.
Information Economics and Policy, 12, 221–248.
The correct policy and regulatory measures can boost [18] Gefen, D. & Straub, D. (1997). Gender difference in the perception and
mobile access and usage. Avoiding sector and service specific use of e-mail: an extension to the technology acceptance model. MIS
levies and additional taxes will result in lower recovery costs Quarterly, 21, 389– 400.
and therefore lower prices. Lower prices would allow millions [19] Gourieroux, C. (1986). Econométrie des variables qualitatives,
Economica.
of people currently unable to afford to access mobile services [20] Haque, O., M. (2007). Income Elasticity and Economic Development,
to do so and those currently restricting their usage as a result Springer, ISBN 0-387-24292-9.
of the high cost of services to use services more extensively [21] Hausman, J., Tardiff, T., & Belinfante, A. (1993). The effects of the
breakup of AT&T on telephone penetration in the United States.
and effectively. This is likely to produce greater profits for American Economic Review (American Economic Association Papers
companies overall and Governments will as a consequence and Proceedings), 83, 178–184.
have even more money in their coffers due to taxes on greater [22] Heij, C., de Boer, P., Franses, P., Kloek, T. and van Dijk, H. (2004).
Economteric Methods with Applications in Business and Economics,
operator revenues. Oxford University Press, ISBN 0-19-926801-0.
[23] Jalava, J. and M. Pohjola (2002). Economic growth in New Economy:
VIII. REFERENCES Evidence from advanced economies, Information Economics and Policy,
Vol.14, pp. 189-210
[1] Adamowicz W. & Louviere J. & Williams (1994). Combining Revealed [24] Jorgenson D, M.S.Ho and K .Stiroh (2005). Productivity: Information
and Stated Preference Methods for Valuing Environmental Amenities. Technology and the American Growth Ressurgence. London;
Journal of Environmental Economics and Management, Vol. 26, pp.271- Cambridge: The MIT Press.
292. [25] Kabbar, E.F. and Crump B.J. (2006). The factors that influence
[2] Bagchi, K. and G. Udo (2007). Empirically testing factors that drive ICT Adoption of ICT by recent refugee immigrants to New Zealand,
adoption in Africa and OECD set of Nations, Issues in Information Informing Science Journal, Vol.9, pp.111-120.
Systems, Vol. 8, No.2, pp. 45-52. [26] Kaserman, D. L., Mayo, J. W., & Flynn, J. E. (1990). Cross
[3] Baliamoune-Lutz, M. (2003). An analysis of the determinants and subsidization in telecommunications: Beyond the universal service fairy
effects of ICT diffusion in developing countries, Information tale. Journal of Regulatory Economics, 2, 231–249.
Technology for development, Vol.10, pp.151-169. [27] Kelly, T. (2004). Changing ICT ranking of African Nations. The
[4] Bann C. (1999). A Contingent Valuation of the Mangroves of Benut. Southern African journal of Information and Communication, Issue No.
Forestry, Department, Peninsular Malaysia, Danish Cooperation for 5, 2004.
Environment and Development and the Johor State. [28] Kwapong, O.A.T.F (2007). Problems of policy formulation and
www.soc.uoc.gr/calendar/2000EAERE/papers/PDF/A6-Bann.pdf implementation: The case of ICT use in rural women’s empowerment in
Last Appeal: 10.07.07 Ghana, International Journal of Education and development using
[5] Bateman I., Carson R., Day B., Hanemann M., Hanley N., Hett T., Information and Communication technology, Vol.3, No. 2, pp. 68-88.
Jones-Lee M., Loomes G., Mourato S., Ozdemiroglu E., Pearce D., [29] Long, S. (1997). Regression Models for Categorical and Limited
Sugden R., Swanson J. (2002). Economic Valuation with Stated Dependent Variables, Advanced Quantitative Techniques in the Social
Preference Techniques – A Manual. Edward Elgar, Cheltham, UK, Sience Series, Sage Publications, ISBN 0-8039-7374-8.
ISBN: 1-84064-919-4. [30] Long, S. & Freese, J. (2006). Regression Models for Categorical
[6] Clarke, G., & Wallsten, S. (2002). Universal(ly bad) service: Providing Dependent Variables Using Stata, Stata Press, ISBN1-59718-011-4.
infrastructure services to rural and poor urban consumers. World Bank [31] Maddala, G. S. (1983). Limited-dependent and Qualitative Variables in
Policy Research Working Paper, 2868, 1–54. Econometrics, Cambridge University Press.
[7] Crandall, R. & Waverman, L. (2000). Who pays for universal service? [32] McCormick, D. & J. Onjala (2007). Methodology for Value Chain
When telephone subsidies become transparent. Washington, DC: Analysis in ICT Industry Frameworks for the Study of Africa, Institute
Brookings Institution Press. for Development Studies University of Nairobi
[8] Eriksson, R., Kaserman, D., & Mayo, J. (1998). Targeted and untargeted [33] Milne, C. (2006). “Telecoms demand: measures for improving demand
subsidy schemes: Evidence from postdivestiture efforts to promote in developing countries – a toolkit for action”. Discussion Paper WDR
universal telephone service. Journal of Law and Economics, 41, 477– 0603, Main Report (World Dialogue on Regulation).
502. [34] Mitchell R. & Carson R. (1984). A Contingent Valuation Estimate of
[9] Esselaar, S., and Stork, C. (2005). Mobile Cellular Telephone: Fixed- National Freshwater Benefits – Technical Report to the US
line Substitution in Sub-Saharan Africa, The Southern African Journal Environmental Protection Agency. Resources for Future, Washington
of Information and Communication, Issue 6, pages 64-73, ISSN 1607- D.C.
2235. [35] Niininen, P. (2001). Computers and economic growth in Finland, in
[10] Fabiani, S., F. Schivardi and S. Trento (2005), ICT adoption in Italian Information Technology, Productivity, and Economic Growth:
manufacturing: firm-level evidence, Industrial and Cooperate Change, International Evidence and implications for economic Development,
pp.1-25. Matti Pohjola, ed., oxford University Press.
[11] Frank R. & Bernanke B. (2003), “Principles of Economic”. McGraw- [36] OECD (2003). ICT and Economic growth. Evidence from OECD
Hill, Boston countries, industries and firms. Paris (France).
[12] Garbacz, C., & Thompson, H. G., Jr. (1997). Assessing the impact of [37] Oliner, S.D. and D.E. Sichel (2000). The resurgence of growth in the
FCC lifeline and link-up programs on telephone penetration. Journal of late 1990s: Is Information technology the Story?, Journal of Economic
Regulatory Economics, 11, 67–78. Perspectives, Vol.14, pp. 3-22.
[13] Garbacz, C., & Thompson, H. G., Jr. (2002). Estimating telephone [38] Ono, H. (2003). Gender and the Internet. Social Sciences Quarterly, 84,
demand with state decennial census data from 1970–1990. Journal of 111–121.
Regulatory Economics, 21, 317–329. [39] Orbicom (2003). L’observatoire de la fracture numérique... et au-delà.
[14] Garbacz, C., & Thompson, H. G., Jr. (2003). Estimating telephone Conseil national de la recherche du Canada, Québec (Canada).
demand with state decennial census data from 1970–1990: Update with [40] Panzar, J. C. (2000). A methodology for measuring the costs of universal
2000 data. Journal of Regulatory Economics, 24, 373–378. service obligations. Information Economics and Policy, 12, 211–220.
[15] Garbacz, C., & Thompson, H. G., Jr. (2005). Universal [41] Parsons, S. G. (1998). Cross-subsidization in telecommunications.
telecommunications service: A world perspective. Information Journal of Regulatory Economics, 13, 157–182.
Economics and Policy, 17, 495–512. [42] Perl, L. J. (1983). Residential demand for telephone service. White
Plains, NY: National Economic Research Associates.
402
[43] Piccoli, G., Ahmed, R. & Ives, B. (2001). Web based virtual learning
environments: a research framework and a preliminary assessment of
effectiveness in basic IT skills training. MIS Quarterly, 25, 401–426.
[44] Pohjola, M. (2001). Information technology and economic growth: A
cross-country Analysis, in: Information Technology, Productivity, and
Economic Growth: International Evidence and implications for
economic Development, Matti Pohjola, ed., oxford University Press.
[45] Roller, L. H. & Waverman, L. (2001). Telecommunications
Infrastructure and Economic Development: A simultaneous approach.
The American Economic Review, Vol. 91, No 4, 909-923.
[46] Rosston, G. L., & Wimmer, B. R. (2000). The state of universal service.
Information Economics and Policy, 12, 261–283.
[47] Row R. & Schulze W. & Breffle W. (1996). A Test for Payment Card
Biases. Journal of Environmental Economics and Management, Vol. 31,
pp. 178-185.
[48] Sichel, D. (1997). The Computer revolution- An Economic Perspective,
Washington D.C.: Brookings Institutions Press.
[49] Taylor, L. D. (1994). Telecommunications demand in theory and
practice. Dordrecht: Kluwer Academic Press.
[50] Taylor, L. D. (2002). Customer demand analysis. In M. Cave, S.
Majumdar, & I. Vogelsang (Eds.), Handbook of telecommunications
economics (pp. 97–142). Amsterdam: Elsevier Science.
[51] Taylor, L.D. and Kridel, D.J. (1990). “Residential Demand for Access to
the Telephone Network,” in Telecommunications Demand Modeling,
ed. by A. de Fontenay, M.H. Shugard, and D.S. Sibley, North Holland
Publishing Co., Amsterdam.
[52] Tegegne G.E., (1999). “Willingness to pay for environmental protection:
an application of Contingent Valuation Method (CMV) in Sekota
District, Nothern Ethiopia”. Ethiopian Journal of Agricultural
Economics Vol 3: 123-130
[53] UNCTAD (2006). “Rapport 2006 sur l’Economie de l’information”,
UNCTAD/SDTE/ECB/2006.
[54] Valletti, T. M. (2000). Introduction: Symposium on universal service
obligation and competition. Information Economics and Policy, 12, 205–
210.
[55] Venkatachalam L. (2004). The Contingent Valuation Method - A
Review. Environmental Impact Assessment Review, Vol. 24, pp. 89–
124.
[56] Venkatesh, V. & Morris, M.G. (2000). Why don’t men ever stop to ask
for directions? Gender, social influence, and their role in technology
acceptance and usage behavior. MIS Quarterly, 24, 115–139.
[57] Wallsten, S. (2001). Telecommunications investment and traffic in
developing countries: The effects of international settlement rate
reforms. Journal of Regulatory Economics, 20, 307–323.
[58] Waverman, L., Meschi, M. & Fuss, M. (2005). The impact of telecoms
on economic growth in developing countries. In Africa: The Impact of
Mobile Phones: Moving the Debate Forward, The Vodafone Policy
paper series, No.2.
[59] Wooldrige, J. (2006). Introductory Econometrics A modern Approach,
4th edition, South-Western Cengage Leraning, Canada, ISBN 13 : 978-
0-324-58548-3.
403
IX. APPENDIX
TABLE 5: PROBIT MODEL SPECIFICATION
Variable Variable Type of Variable Expected Comment
Name sign
Dependent Mobile Dichotomous variable: Na
Variable (M) Individual has a mobile or active SIM
card = 1, 0 otherwise
Independent Income Continuous variable: Individual income positive Income is assumed to be
Variables (I) in US$ based on end of year nominal the main barrier to
exchange rates (source IMF) adoption
Gender (G) Dichotomous variable: female = 1, male Not The assumption is that
=0 significant there is no gender
difference when it comes
to mobile adoption
Age (A) In years Negative Technology being a
domain for the younger
generation
Tertiary (T) Dichotomous variable: highest education Positive Education should enable
being tertiary = 1, otherwise 0 individuals to use mobile
Secondary Dichotomous variable: highest education phones, while also being
(S) being secondary = 1, otherwise 0 an indicator for income,
Primary (P) Dichotomous variable: highest education model would drop
being primary, remedial or traditional variables if multi-
=1, otherwise 0; collinearity persists
Vocational Dichotomous variable: highest education
(V) being vocational =1, otherwise 0
Rural (R) Dichotomous variable: rural = 1, urban Negative Mobile network coverage
=0 is less in rural areas
compared to urban (major
urban and other urban)
ones and less adoption is
expected therefore
Social Dichotomous variable: social network Positive Membership in social
Network (as church groups, sports clubs etc.) networks might increase
(SN) membership = 1, otherwise 0 communication need
Constant Captures various factors that either are Negative Captures no education, but
constant for a country such as the price also urban
for mobile or fixed telephony or that
serve as reference for dichotomous
variables such as the educational
variables and rural-urban location.
TABLE 6: NATIONALLY REPRESENTATIVE PROBIT RESULTS FORE EACH COUNTRY USING SAMPLING WEIGHTS (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Country Pseudo Prob Positive Coefficients + = significant at 0.1 level, ++ = significant at 0.05 level,
R2 > +++ = significant at 0.01 level
chi2 Negative Coefficients - = significant at 0.1 level, -- = significant at 0.05 level, ---
= significant at 0.01 level
I A G T S P V R SN Constant
Benin 0.396 0.00 +++ ++ No +++ +++ +++ +++ --- no ---
Botswana 0.16 0.00 + No No ++ +++ No Skip -- +++ No
Burkina Faso 0.253 0.00 +++ --- --- +++ +++ +++ Skip --- No No
Cameroon 0.258 0.00 +++ ++ No +++ +++ +++ ++ --- ++ ---
Côte d'Ivoire 0.292 0.00 +++ -- No +++ +++ No +++ --- +++ No
Ethiopia 0.452 0.00 +++ No No +++ ++ No +++ --- No ---
Ghana 0.158 0.00 No - No +++ +++ +++ ++ --- ++ No
404
TABLE 7
WILLINGNESS AND ABILITY TO PAY
Average
willingness and Average National
ability to pay for a expected cost of New users at New users at
mobile handset in a mobile 20 US$ for an 15 US$ for New users at 10
US$ handset US$ handset an handset US$ for an handset
Benin 7.45 11.44 124,972 487,176 677,715
Botswana 19.14 27.38 119,014 196,496 228,203
Burkina Faso 8.92 12.84 427,032 1,242,397 1,451,446
Cameroon 15.11 22.16 864,053 1,728,316 1,865,876
Côte d’Ivoire 29.70 30.06 3,057,420 3,539,351 3,914,283
Ethiopia 6.06 64.19 1,436,628 1,637,668 2,644,673
Ghana 14.02 23.15 1,283,271 1,469,652 1,841,837
Kenya 16.98 26.68 2,857,406 4,160,498 5,658,430
Mozambique 4.00 23.2 56,457 79,895 287,147
Namibia 24.64 25.12 162,992 192,395 232,584
Nigeria* 5.65 12.57 356,907 1,004,573 2,527,884
Senegal 19.55 2543.0% 1,336,691 2,169,548 2,301,775
South Africa 17.44 32.41 2,549,812 3,251,782 3,991,768
Tanzania 10.89 17.3 1,422,927 2,102,510 3,272,065
Zambia* 17.42 22.43 682,864 1,061,607 1,598,555
Benin 7.45 11.44 124,972 487,176 677,715
TABLE 8: MOBILE EXPENDITURE - ROBUST REGRESSION RESULTS FOR EACH COUNTRY USING SAMPLING WEIGHTS (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Country R2 Prob Only significant coefficients are being displayed: * = significant at 0.1 level, ** = significant
>F at 0.05 level, *** = significant at 0.01 level
Y A G WP PP F R SN
Benin 0.399 0.000 0.35*** 1.07*** -0.27***
Botswana 0.298 0.000 0.46*** -0.26**
Burkina Faso 0.206 0.000 0.33*** 0.28* 0.35***
Cameroon 0.22 0.000 0.27*** -0.32** -0.22* 0.54* -0.4*
Côte d'Ivoire 0.196 0.000 0.36*** -0.02* 0.45** 0.37* 0.34*
Ethiopia 0.353 0.000 0.29*** -0.01** -0.47*** 0.23*
Ghana 0.145 0.000 0.47*** -0.01**
Kenya 0.251 0.000 0.33*** -0.02*** 0.44*** 0.47***
Mozambique 0.189 0.000 0.11** -0.03*** -0.72*** 1.04***
Namibia 0.449 0.000 0.43*** -0.23* 0.37**
Nigeria 0.458 0.000 0.71*** 0.24*** -0.41* 0.24***
Rwanda 0.339 0.000 0.55*** 0.66***
Senegal 0.266 0.000 0.43*** 0.247*
South Africa 0.477 0.000 0.47*** -0.02*** 0.2** -0.2*** 0.3***
Tanzania 0.127 0.000 0.33*** 0.84**
Uganda 0.292 0.000 0.39 *** 0.48** -0.24* 0.43**
Zambia 0.378 0.000 0.52*** -0.01** 0.69***
405
TABLE 10: MOBILE WILLINGNESS AND ABILITY TO PAY (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Average Monthly
Number of 16+ without a mobile phone or active SIM monthly untapped
16+ without mobile willing and able to spend monthly: WTP in market in
phone or active SIM US$ US$
card 1US$ or more 2 US$ or more 5 US$ or more million
Benin 69.8% 3,162,099 2,797,101 1,402,507 161,217 2.94 8.26
Botswana 40.5% 446,140 311,446 199,511 66,192 4.28 1.47
Burkina
4,929,897 4,371,694 1,875,892 430,952 3.13
Faso 72.8% 13.71
Cameroon 63.5% 5,177,393 3,452,460 1,855,275 550,724 3.75 13.14
Côte d'Ivoire 58.2% 7,033,592 4,485,498 3,645,855 1,677,528 6.86 31.44
Ethiopia 96.8% 42,497,353 10,231,145 3,104,395 74,428 1.53 25.68
Ghana 40.2% 5,036,815 2,849,435 1,953,135 984,279 9.34 38.40
Kenya 48.0% 9,941,748 5,866,299 5,235,785 1,245,083 3.30 25.69
Mozambique 74.3% 14,078,222 1,407,840 1,199,765 376,037 2.96 6.70
Namibia 50.7% 644,056 275,364 247,254 71,171 4.88 1.35
Nigeria* 22.7% 18,541,687 7,989,151 6,234,941 5,128,000 6.09 65.25
Rwanda 90.1% 4,735,492
Senegal 60.2% 3,779,221 3,428,481 1,294,681 502,730 3.28 11.33
South Africa 37.9% 12,331,758 7,604,512 5,551,777 2,209,625 4.34 36.27
Tanzania 78.5% 15,066,652 5,560,959 4,750,935 1,064,087 2.61 21.42
Uganda 79.3% 11,174,801
Zambia* 54% 2,944,732 1,933,833 1,310,576 329,922 3.45 8.2
406
Abstract—Non-governmental organizations (NGOs) working possible. In this regard, information technologies frequently
in disadvantaged communities have a variety of data-collection play an important role in the storage, analysis, and display of
and analysis needs, for example, for performing surveys or painstakingly collected data. Microfinance institutions
monitoring programs. Because much of this data collection benefit from back-end databases that store client data [22];
occurs in environments with insufficient IT support and healthcare institutions have need to maintain medical
infrastructure, and among populations not always comfortable
records; and NGOs in general can benefit from longitudinal
with technology, paper forms rather than electronic methods
remain the predominant means for data collection.
data collected over the life of their programs.
We consider the design of machine-readable paper forms for Organizations have also considered the use of information
NGOs. We first examine the unique needs of NGOs that technology for the task of collecting the data itself. Micro-
interact with underprivileged populations through interviews finance accounting information systems [28] and healthcare
with eleven organizations and an in-depth investigation of one information on PDAs [2] are two significant examples. The
NGO’s specific form-filling requirements. These explorations expectation is that these efforts minimize transcription of
led to a focus on numeric forms – forms with questions data from paper forms and allow for more rapid analysis.
requiring responses largely constrained to numbers. There are, however, many who question the value of
We then present an experiment which evaluates how a
electronic means for data collection in poor environments.
variety of formats for numeric data would fare with users from
backgrounds similar to those who might fill out such forms.
One cost-benefit analysis of mobile devices used by
Our goal was to balance the tradeoff between ease-of-use microfinance institutions to interact with their clients
among our intended population and machine readability. suggests that the benefits of electronic technology for data
Combining the results of the experiment with an analysis of collection do not always outweigh the costs [9]. Others point
machine-readability from a technical perspective, we propose out that electronic mechanisms are distrusted by populations
the best numeric input methods for different NGO form filling who are used to physical evidence of transactions [23].
requirements. Paper, on the other hand, is a ubiquitous, low-cost, and
Index Terms— machine-readable forms, paper forms, well-understood medium. Even in the developed world,
input methods, ICT for development paper forms remain widely used for the purposes of
gathering information, despite ongoing advances in digital
I. INTRODUCTION technology. However, while it may be easy to collect data
Monitoring, evaluation, measurement, and self- on paper, transferring that data into a format suitable for
assessment are among several critical tasks for non-profit subsequent computer based storage and analysis remains a
organizations working in global development. Knowing the difficult problem that is currently typically resolved only by
nature of one’s impact, ideally with accompanying tedious manual data transcription.
quantitative information, allows for self-correction, reports In this paper, we first study how paper forms are used for
to sponsors and potential donors, and external influence. data collection in the context of non-profit non-
Most program assessment requires data collection in some governmental organizations (NGOs) that seek to monitor
form as a first step. Data collection can be tedious and their own programs. We performed interviews with eleven
expensive (in labor, time, and financial cost), and it is thus development-focused organizations involved in healthcare,
desirable to extract as much value from the effort as microfinance, education, and agriculture. We then probed
deeper with one of the organizations to better understand
Manuscript received September 22, 2008. Revised February 20, 2009. their data-collection pipeline. This investigation resulted in a
G. Singh was with Microsoft Research India, Bangalore, Karnataka,
India (email: gursharan@alumni.upenn.edu). proposal for data collection that uses paper for the “front-
L. Findlater is with University of British Columbia, Vancouver, BC, end” collection tasks, and subsequently uses a combination
Canada (email: lkf@cs.ubc.ca). of scanner and PC to digitize the data. Rather than attempt to
K. Toyama is with Microsoft Research India, Bangalore, Karnataka,
India (email: kentoy@microsoft.com).
handle the myriad of possible paper forms right away, we
S. Helmer is with University of British Columbia, Vancouver, BC, instead focused on an important subset – forms used for
Canada (email: shelmer@cs.ubc.ca). collecting numeric input, which was found to be frequently
R. Gandhi is with Microsoft Research India, Bangalore, Karnataka,
India (email: riking@microsoft.com).
used by NGOs. Although many additional factors besides
R. Balakrishnan is with University of Toronto, Ontario, Canada (email: input type will determine the value of such a system in a real
ravin@dgp.toronto.edu). environment, our study provides a first step towards
building an effective system. We conducted a study of how
407
people with varying educational backgrounds responded to C. Data Collection for the Developing World
form-filling tasks using several different “input methods” In spite of the trend to place digital technology directly in
for recording numbers. Finally we analyzed the literature on front of underserved populations, a few researchers have
machine-readable forms, and gauged how each of the input focused on using paper instead. Parikh et al. [24]
methods would fare for subsequent automatic digitization of systematize paper-based data collection by using paper user
the data. interfaces and automated forms processing for microfinance
institutions. Mackay et al. [20, 21] propose using paper on
II. RELATED WORK top of tablet PC displays. Parikh et al.’s [23] CAM system
We review four categories of research on which our work demonstrates use of a camera-equipped mobile phone to
builds. First, a number of electronic devices and software read bar codes printed on paper forms. Bar codes cause the
tools have been designed to complement paper in phone to issue audio annotations, which instruct the user to
information work. Second, some user studies have compared transfer data on the paper form into the mobile phone.
digital and paper-based methods of data entry. Third, some DeRenzi et al. [6] use a PDA based system to significantly
research has focused on using paper as an interface for increase adherence to medical protocols for pediatric health
systems that interact with underserved populations. And, care. These systems all incur a device cost for the person
finally, existing work on machine-readable forms will be doing the data collection.
reviewed in a separate section. Ellison et al. [7] stress the importance of participatory
monitoring and evaluation for a small scale NGO. Chand et
A. Complementing Paper
al.’s [5] Jadoo, is a paper-only exploration of how structured
The importance of paper as a medium to capture data is content on paper can be made easy to deal with even for
recognized, even in technology-heavy developed-world people with little formal education. Our early investigations
environments: “Rather than pursue the ideal of the agree with this work on the value of paper, and its likely
paperless systems, we should work toward a future in which continued use by NGOs.
paper and electronic document tools work in concert and Our research builds on the above work and focuses on the
organizational processes make optimal use of both” [25]. novel problem of designing paper forms which should, on
And, indeed, many tools have addressed this need. the one hand, allow easy machine readability, and on the
Early research by Johnson et al. [13] showed how paper other hand, be easy to fill in for people with diverse
documents could be stored, indexed, distributed and educational backgrounds.
processed by capturing them as images with scanners.
Wellner’s [29] classic Digital Desk system used a camera III. NGOS AND FORMS
and a projector to record input and project documents.
Over the last several years, we have both formally and
Guimbretière’s PADD [12] attempts to bridge the gap
informally interviewed eleven organizations – five NGOs,
between the paper and the digital world. Built on top of
four microfinance institutions, and two clinics – in India
PADD, Liao et al.’s [17] Papiercraft implemented a gesture
with respect to their data-collection needs and experiences
interface for natural editing. In a lighter-weight approach,
with both digital and paper-based forms. These
the Anoto pen [1] uses paper printed with unique marks and
organizations varied significantly in size and in their use of
a small camera in the pen to synchronize changes on the
forms (from requiring a handful of forms filled in per
paper to a digital version of the document. Arai et al. [3]
month, to hundreds of thousands per year), but all had in
show how a camera-assisted pen can capture words on paper
common that they had data-collection needs and continued
by hyper-linking them to other content. Koike et al. [15]
to utilize paper forms for this purpose, despite isolated trials
used projected Venn diagrams to record input.
with electronic data collection.
Although these systems combine the strengths of paper
Below, we outline what we learned from these
and electronics, their benefits come with additional costs,
organizations, and go into some detail about the data-
which is a concern for NGOs.
collection system for one organization that typifies the NGO
B. Digital versus Paper experience with forms.
Some researchers have compared paper and digital A. Overall NGO Experiences with Forms
systems for data entry, and overall tend slightly in favor of
The need for data collection is widespread among NGOs,
paper. Galliher et al. [8] found that people are more likely to
and means for doing it efficiently and subsequently being
complete paper forms than digital ones. They cite technical
able to analyze the results easily are, in the words of one
difficulties, as well as stolen or lost devices. On the other
NGO, “the need of the hour” [14]. Despite the prevalence of
hand, errors of omission were more common on paper.
paper forms, however, NGOs do not seem completely
Shelby-James et al. [26] disprove an often-made claim that
satisfied with their systems (or lack thereof) for
electronic data capture is more accurate than paper-based
incorporating collected data in their routines.
methods. They found error rates with handheld computers
were over sixty times that for paper-based data entry.
408
The goals of most data collection fall into one of several of data collection. In another case, a bicycle courier was
categories. For surveys, the intent is simply to understand paid 15 cents per village to collect sheaves of paper records
the state of a population as it exists. For monitoring and and bring them back to the NGO office.
evaluation of projects, the objective is accountability, both
in terms of whether NGO staff and beneficiaries are Infrastructure and maintenance: Among the organizations
performing tasks they have committed to, or whether the that had either experience with or expressed interest in
intended outcomes of a project are being accomplished. For information technology (IT) systems for data collection and
baseline surveys, the collected data helps identify various processing, all cited challenges with power and maintenance
parameters and tune efforts before work on a particular issues. Fig. 1 shows a person filling out a form with light
project is started by an NGO, for example, the majority from a flashlight, a frequently encountered situation.
occupation of the population in the area.
Forms can be filled in by different people with different
backgrounds, but for the most part, there seem to be two
classes of people filling in forms: those who are paid by the
NGO to perform data collection and those who are direct
beneficiaries of the NGO’s programs. The former category
consists of people who are almost always literate, and
typically have completed secondary school; some have
advanced degrees, as well. The latter range wider in their
backgrounds, but frequently have not had much formal
education. Anyone asked to fill in a form will likely be
literate, but there are cases where the form-fillers can just
barely sign their name and write numerical digits. Fig. 1. A participant fills out a form during a power outage.
Furthermore, their handwriting may not be very good.
Forms involve recording of numerical data (e.g., age,
price, income, dates, durations), textual data (e.g., name, Training, data accuracy, and consistency: Data-collection
free-text comments), and binary or multiple-choice tasks often require some training, and systems involving
questions. It is worth noting that a significant fraction of the digital devices only add to this need. Training is a time- and
questions asked on forms appear to be reducible to numeric management-intensive activity that NGOs invest heavily in.
information, with multiple-choice answers. The key Ensuring data accuracy and consistency of routine
exceptions are names and free-form comments. Also, in surveys is another headache for many NGOs. Staff, who
some cases, photographs were part of the data collected. themselves are not paid particularly well, may have little
Overall, the repeated concerns expressed by NGOs motivation to perform data-collection tasks with care. In
regarding data collection were: cost, time, training, data many cases, the data-collection staff itself needs to be
accuracy and consistency, storage, and means of data monitored, to ensure that data is being correctly collected
analysis. Among those NGOs who had experimented with according to specification.
electronic systems, these issues were again highlighted, in Even dedicated staff, however, often embed form-filling
addition to difficulties with infrastructure and maintenance. tasks into what can otherwise be busy daily schedules. Thus,
We discuss each of these concerns briefly below. data accuracy and consistency can suffer.
Cost: NGOs are run on tight budgets, and many expressed Data storage and analysis: Almost all of the NGOs we
concerns even about the cost of paper and printing. (Some of spoke with understood the value of digitizing data once it
the NGOs we spoke with explicitly requested a ream of was collected. Many employed data-entry staff whose sole
“white paper” as a meaningful gift in exchange for their job was to take paper forms and convert them into digital
time.) Digital equipment, of course, can be prohibitively format; others outsourced these tasks to transcription
expensive, and even those that can afford the capital costs services; in some organizations volunteers handled this task.
have difficulty with costs of maintaining technology. One These methods are all costly – for some organizations, the
interesting point here is that centralized technology in an costliest part of the data pipeline – or irregular, and
NGO office is easier to justify than devices that must scale depressingly often, we were shown stacks of paper forms
in proportion to the number of respondents (or equivalently, that had yet to be processed; in some cases, they had been
data-collection staff). gathering dust for years.
The cost of staff hours for data collection also adds up, Analysis of data most frequently involved the use of
although these costs match labor costs of the local area, and spreadsheet software as well as tables in word-processing
are typically very low by developed-country standards for software filled in by hand. Two microfinance institutions
hourly wages. Among the NGOs we consulted, data used custom software, which produced fixed reports once
collecting staff were paid as little as US$4 for a 10-hour day the data was entered.
409
Other issues: One thoughtful NGO head mentioned that videos, their interests, suggestions, and so on. The junior
there would be some value in systems that helped staff then summarizes feedback from every session onto a
organizations develop good questionnaires for data form, shown in Fig. 2. These per-session forms are analyzed
collection. He felt that a lot of data collected by NGOs was by the senior staff weekly, and finally aggregated onto a
done without a clear understanding of what information was monthly form. These forms are currently filled in plain text
most useful for evaluation purposes [14]. in Kannada (the local language in the region). The filled
forms are then sent to the regional office. The data is entered
B. One Particular NGO’s Experience
into a database by a human transcriber and is then analyzed
GREEN Foundation is a small NGO based in Bangalore, to spot trends and results in the villages of operation.
India, whose mission is to spread sustainable agricultural We observed that the form-filling sessions in villages tend
practices among rural farmers. With an annual budget of to become interactive “classroom sessions” (Fig. 3) leading
about US$100,000, they have 15 field staff tasked with to vital exchange of information between the NGO staff and
covering an operational area of approximately 100 villages. the participants. Forms filled by the farmers convey what
Recently, they began a project that involves group they want to be taught, and forms filled by on-field staff
sessions with farmers, where facilitators mediate discussion help document their staff progress.
based on video content [11]. As part of the monitoring and Overall, our key findings were that: (1) data collection
evaluation of this project, both facilitators and farmers are and form-filling are important activities for many NGOs; (2)
requested to fill in paper forms. The program has been cost and ease-of-use are major concerns, often preventing
running for over a year, in each of 12 pilot villages, with technology-heavy systems; and (3) digitized data is desired,
typically three sessions per week. The first author visited but digitizing data is the bottleneck for data-collection
these sites several times over a period of two months to efforts. These findings confirm findings from earlier work
better understand how the forms were used. and additionally identify an important problem faced by
Their case is particularly interesting for us, because it many NGOs.
provides information about the two broad classes of form-
fillers: (1) two groups of paid staff of the NGO, all of whom
are literate (senior staff with ten or more years of formal
education, and junior staff, with at least eight years of
education), and (2) smallholder farmers, who earned no
more than $2 a day, and rarely had more than six years of
education; many were all but illiterate.
Fig. 3. The verandah of a house used for meetings and data collection.
will be implemented using multiple-choice or numeric un-coded Bubbles). Coded marking methods, in contrast,
responses as much as possible, because these responses are only provided a single row for all digits (see Table 1a). The
comparably easy to digitize. Free-form text will be given advantage of the un-coded methods is that they allow for a
space within blank rectangles on the form, which can be complete set of numbers: for example, with 4 rows users can
scanned and stored as images (and not converted to enter any number from 1 to 1000. The coded marking
electronic text). reduces the set of numbers that can be input: for example,
The proposed system keeps costs low. No per-staff device “58” and “85” would appear the same when entered as
is required, and all of the equipment is available as mass- coded input and numbers with repeating digits, such as “22”,
market off-the-shelf hardware, which helps to keep costs cannot be input at all. However, the advantage of coded
low and alleviates technical maintenance needs. Since marking is that it greatly reduces the visual complexity of
NGOs that collect data would like to have it in digital form the form and saves physical space, which in turn saves
eventually, the willingness to invest and maintain at least paper. Coded marking could be useful for situations where
one PC is assumed (many will already have a PC). Printers the numeric data is nominal rather than scalar. If the form is
and scanners add costs that are small compared with the PC used to record attendance at village meetings, for example,
itself, and paper is the only additional ongoing cost. The each individual may have an identification code assigned to
system would maintain all of the advantages of paper, them. Since these are nominal values they could be assigned
namely its low cost, a well-understood “interface”, the lack with the goal of entering them as coded input.
of need for power or maintenance, and robustness in the TABLE I
field. Finally, the proposed system addresses the data-entry NUMERIC INPUT METHODS CONSIDERED.
bottleneck by providing an automated means of digitizing
much, if not all, of the data. a. Coded Bubbles
This proposal still leaves us with the following challenge:
on the one hand, existing techniques for machine reading of
hand-marked forms is reliable only when the forms are b. Un-coded Bubbles
designed and filled out in a particular way (e.g., “bubbles”
on standardized test forms) or when digits are neatly written
(e.g., post-office automated digit-reading systems for sorting Coded
c.
mail); on the other hand, the groups that we anticipate will Circles
fill in these forms are less familiar with standardized forms,
and many have borderline penmanship.
This is a non-trivial challenge that requires both d. Un-coded Circles
engineering and interface design. In the remainder of this
paper, we consider how best to handle numeric input on
paper forms with the goal of achieving a reasonable tradeoff Coded
e.
Checkbox
between ease of user comprehension, user accuracy, and
machine readability.
Un-coded
f.
V. NUMERIC INPUT METHODS Checkbox
applications (e.g., [10]). Moreover, there are numerous Kannada, an Indian language of interest to the NGOs we
commercial software packages available that require little surveyed, the recognition rate can be less than 90% [16] and
more than a PC and a document scanner. The process itself as mentioned earlier may not be supported by commercial
involves two stages. The first is form registration, which software. It should also be stressed that accuracy rates in
aligns the document so that values can be extracted. The ICR are stated for sanitized data sampled from a completely
second, and more challenging task is extracting the values different distribution of people. It is certain that illiteracy,
from particular locations in the document. These values are poor handwriting, and form degradation will reduce the
extracted using either optical mark recognition (OMR) or accuracy and this adds uncertainty as to whether the
intelligent character recognition (ICR), depending on the technology is applicable in this setting.
input method.
Processing forms in the developing world, however, adds VII. EXPERIMENT
numerous challenges. The cost and ease of deployment may Our goal in designing paper forms is to achieve a
increase since many commercial applications may lack reasonable balance between ease of user input and machine
support for local languages. Another challenge, based on our readability. We thus also conducted a controlled study to test
NGO interviews, is that since forms may be filled out in less user performance, accuracy and preference with respect to
than ideal conditions, they may be wrinkled or dirty, which the 10 different input methods.
can affect both the form registration and value extraction.
Commercial software packages also generally assume that A. Participants
the end users (the individuals filling out the forms) are We recruited 40 participants who ranged in age from 17-
literate and have a certain level of education. These users are 50 years (M = 26.2). Their formal education varied from
expected to have some level of competency in filling out, four years of schooling to undergraduate university level.
which is not always the case in the developing world. All of Most of them spoke at least three of the following
these challenges need to be weighed against the ease of use languages: Hindi, Telugu, Kannada, Sinhala, Tamil,
for different input methods. Bengali, Konkani and English. Every participant could write
The un-coded and coded checkbox and bubble input in at least one language, although many had not held a
methods could be processed using OMR. In general, this writing implement since finishing school. A very basic
method of input is nearly 100% accurate, assuming the user vision test (reading a series of numbers aloud) was used to
has followed the instructions [4]. This rate is reported for screen participants before starting the experiment. The
systems requiring specialized scanners and forms, which is participants came from disparate professional backgrounds:
not applicable for most NGO use; systems likely to be used auto-rickshaw (three-wheeled mini-cab) drivers, farmers,
by NGOs will have less accuracy. In addition, in the restaurant workers, security guards, housekeeping personnel,
developing world, users will likely be less familiar with such cab drivers, army soldiers, and machine operators.
forms, thus resulting in more user errors for OMR Participants’ literacy levels, ages, genders, and occupations
processing (for more detail on errors see Results section, were recorded. Participants were compensated with a small
below). Moreover, if the forms are degraded because they gift for their time.
have not been filled out or stored in antiseptic conditions,
B. Task
less specialized systems may be more likely to mistake
smudges, wrinkles and dirt for marks. The circling and The task was to enter 20 numbers on a paper form, with
ticking input methods we tested would also fall under OMR, five each of the following: 1-digit numbers, 2-digit numbers,
but these types of marks are not generally supported in 4-digit numbers and 8-digit numbers. Numbers to be entered
commercial software. The reason for this is possibly that the were displayed on the forms themselves. A sample form for
recognition accuracy is poor given the variability in the the digit per box input method is shown in Fig. 4. The
location and form of the marks, especially when contrasted numbers included in the tasks were randomly generated with
to checkboxes or bubbles. no digit being duplicated in a number (for the sake of the
The digits in one box and digit per box input methods, on coded input methods) and every participant saw the same set
the other hand, require ICR processing, which is less of numbers (though not in the same order). No participant
accurate. ICR technology leans upon advances in machine entered the same set of numbers for more than one type of
learning that have allowed systems to learn how to input method.
discriminate characters from samples of handwriting, and is C. Experimental Design
still an active area of research [18]. Currently, even the most
A single factor within-participant design was used: each
accurate ICR systems are only capable of 98% accuracy
participant completed the experimental task with 10 forms,
[19], and this is the case for digits written by a more literate
one for each input method. Presentation order was
population than our target users. In the case of forms
counterbalanced using a Latin square design and participants
requiring recognition of characters of partly non-alphabetic
were randomly assigned to an order.
systems, the accuracy is much worse. For example, for
412
VIII. RESULTS
Participants were divided into two educational groups:
those with up to middle-school education (7 years or less, M
Fig. 4. Sample digit per box form. The numbers above each set of boxes are
the numbers to be entered for the experimental task. = 6.1), and those with more (M = 11.9). A 2x10x10
(educational group x input method x presentation order)
D. Measures repeated measures (RM) ANOVA on the main dependent
We measured speed and accuracy, and asked participants variable of speed showed no significant main or interaction
to provide preference feedback. Speed was measured as the effects of presentation order, so we simplify our analysis by
time to complete the entire task for each input method and only examining the effects of education group and input
was measured manually with a stopwatch. Errors were method (2x10 RM ANOVAs). All pair-wise comparisons
counted after the forms were complete. A mark on a number were protected against Type I error using a Bonferroni
that logically should not be there, or would obviously make adjustment. Where df is not an integer, a Greenhouse-
a machine register the input incorrectly was counted as an Geisser adjustment for non-spherical data has been applied.
error. Subjective feedback was collected for each input One outlier in each of the low and high education groups
method using 5-point Likert scales on difficulty, confusion, were excluded because they were more than two standard
and perceived speed. deviations away from the mean on the dependent variable of
speed. Thus, we report on data from 38 participants (10 in
E. Procedure
the low education block and 28 in the high education block).
Experimental sessions took 45–80 minutes per
participant, as they were all given as much time needed to A. Speed
complete or quit the task. Most of the participants were from Average time to complete the forms with each input
villages in north and south Karnataka and we chose a quiet, method is shown in Fig. 6. Participants in the lower
isolated environment in which to conduct the sessions in education group took 31.8 minutes on average to complete
each of these villages (Fig. 5). The remaining sessions were all forms, while the higher education group took only 21.1
conducted in an office environment, where participants were minutes on average. Note that we ran the RM ANOVA on a
called in. We ensured that there was sufficient lighting in log transform of the speed data since the original speed data
both locations. violated the homogeneity of variance assumption
Participants were first given a background questionnaire (significant Levene’s test).
to collect demographic information. Then, for each of the 10 Both the education level of participants and the input
input methods, participants were given time to examine the method significantly affected the time it took to enter
form and ask questions about the task. Once they were numbers into the forms (main effect of education block: F1,36
comfortable with the nature of the task, they were asked to = 23.3, p < .001, η2 = .393; main effect of input method:
begin. Times were recorded for each of the 4 subtasks, F4.99,180 = 58.8, p < .001, η2 = .620).
namely, 1-digit numbers, 2-digit numbers, 4-digit numbers More interestingly, some input methods were relatively
and 8-digit numbers. After each input method, a better for the higher education group than for the lower
questionnaire was used to collect subjective feedback. education group (interaction effect between input method
413
and education: F4.99,180 = 2.37, p = .041, η2 = .062). To problematic for machine readability, as discussed in Section
understand which input methods were better within each VI. We counted three types of errors: (1) when the number
group of participants, we performed pair-wise comparisons of digits in an entered number was large (four or eight)
and summarize the significant results (p < .05) as follows. participants often found it hard to match the positions of
digits and rows in un-coded versions (Fig. 8a) and marked
two numbers in the same row; (2) ill-formed numbers and
numbers flowing out of boxes were also problematic (Fig.
8b-c); (3) sometimes it was difficult to distinguish which
number had been marked on the control (Fig. 8d).
g. Darkening on the number in bubble h. Overlapping circles Checkmark extending out of box
wise comparisons showed that both the handwritten types One participant commented that she depended on her
(digit per box and digits in one box) were perceived to be husband for filling out textual information since she did not
faster than coded bubbles (p < .05). Digits in one box was know how to write text. However, she was comfortable with
perceived to be the fastest, since participants also felt it was numbers, and felt that if forms only required numeric entry,
faster than two of the un-coded input methods (bubbles and she could do it herself.
checkbox). Participants in the lower education group rated
E. Summary of Results
the tasks as slower than those in the higher education group,
a result that matched actual speed (main effect of education In terms of speed, degree of formal education impacted
block on perceived speed: F1,36 = 7.29, p = .010, η2 = .168). the effectiveness of individual input methods. In particular,
No significant interaction effect was observed between input both un-coded and coded bubbles were the slowest input
method and education group. methods for the higher education group but were no slower
than any other input method for the lower education group.
D. Qualitative Observations We had anticipated that the handwritten methods (digits in
We made a number of observations that, though not one box and digit per box) would be relatively faster for the
directly the focus of the study, illustrate practical issues that higher education participants than the lower education ones,
may be encountered with paper-based forms in this context. but both groups of participants, it turned out, were faster
Several of the less educated participants expressed great with these than with most other input methods. Coded input
enthusiasm with the study. They savored holding a pencil methods were also generally faster than un-coded methods
since they had some formal education but their jobs never and resulted in fewer errors. As expected, higher education
required them to write or read anything. Other participants participants completed the task more quickly than lower
were more apprehensive about holding a pencil because they education participants.
did not know how to write, but they agreed after being Note that since the lower education group had one third as
encouraged by earlier participants. many participants as the higher education group, there was
A few participants without computer backgrounds felt a less statistical power to detect differences among the input
natural inclination to add leading zeros where there were methods; this could explain why some pair-wise
unused “slots”. For example, in a preliminary investigation comparisons were significant for the higher education block
prior to the formal tests, they marked the month of February but not for the lower education block.
in the date field as “02”, filling both the input columns,
instead of only “2” and leaving a column blank. IX. DISCUSSION
Some circumstances of rural life affect the form-filling Combining our understanding of the state of the art for
task in unexpected ways. In one case, we experienced a machine-readability of numeric forms with the user study,
power outage during a preliminary field experiment (again we find that we can make the following recommendations
prior to our formal trials). One of the participants, who was for typical NGO conditions.
over 50 years of age, found it particularly difficult to At the highest level of granularity, if all of speed,
continue, although his vision was fine in bright light. machine accuracy and user accuracy are desirable
characteristics, coded checkboxes seem to be the best fit
415
irrespective of education level of users. If the users are analysis of challenges faced by NGOs around data collection
literate, digits in one box might be a better choice as it was and form filling, (2) a proposal for a semi-automated system
the fastest method for the more highly educated group, with that uses paper as the interface for data collection, and a
low error rate (zero for our participants), and if carefully scanner and PC for digitization, (3) an empirical study with
written these digits have reasonable machine readability 40 participants of how 10 different number-based input
characteristics. methods fare among people with both lower and higher
The relative performance differences between the higher educational backgrounds, and (4) an analysis of how easily
education and lower education groups suggest that there the forms filled using each of these input methods can be
may be a tradeoff between the benefits of structure and automatically digitized.
simplicity for different levels of education. The non- Our main findings are that users tend to prefer and
handwritten methods may be useful for low literacy perform best with techniques that require handwritten
participants since they do not require the same writing numbers. However, these are not the most easily recognized
proficiency as the techniques that require handwritten by a machine. If multiple factors such as speed and accuracy
numerals (digits in one box and digit per box). Our lower of entry, and machine readability are taken into account,
education participants had on average 6 years of schooling. coded checkboxes are likely the best option, although
In our experience from interviewing NGOs, we found that choosing a coded method limits the type of numbers that can
some users of their forms may have even less education, be captured (e.g., numbers with repeating digits are not
which could magnify the relative performance differences easily handled by the coded techniques we considered).
we found based on education. This highlights the need to Another key finding is that the educational background of
consider the educational background of the target user the user can significantly impact performance with different
population. input methods, and as such the demographics of the target
Although coded methods were in general quicker to fill user population should be carefully considered when making
than their un-coded counterparts, and also resulted in fewer decisions as to the type of input method to use.
errors by participants, they are not necessarily suitable for In our study, participants used the same input method
all types of forms. In particular, coded methods cannot across each form. Given the tradeoffs we found with respect
handle numbers with repeating digits. As such, if a form has to coded and un-coded methods, it would be interesting to
some fields that require entry of such numbers, it might be probe further into how people might perform with forms that
best for the sake of consistency to stick to an un-coded have a mixture of input methods, each optimized for the
method throughout the entire form. types of questions being asked on the form. It is unclear if
Interestingly, participants rated handwritten numbers best having more than one input method on a form would be
overall. Both digits in one box and digit per box methods overly confusing to users, particularly those with limited
were reported to be less confusing and easy to fill. In levels of literacy.
contrast, the bubble methods were not particularly favored, Finally, there are many other challenges that will need to
and were the slowest. be addressed in building an effective system to collect and
Taken as a whole, our results indicate that the proposed process data using paper forms as a front-end, and
system of using structured paper forms with subsequent automated input recognition as a back-end. Different types
automated scanning is likely viable. However, the results of data can pose different problems. Our work with only
also caution that there are nuanced tradeoffs that need to be numeric data is a first step in this direction. It would be
made when choosing the type of input method to use. The interesting to actually build a system that handles the
fastest and most preferred techniques (handwritten processing of forms in real use, and to study its performance
numerals) are generally the least accurately recognized by a in the field. We intend to explore this in the near future with
machine, although the digit per box method is arguably a a partner NGO.
reasonable compromise in this regard since the box provides
some structure that eases the task of the recognition REFERENCES
algorithm. 1. Anoto, http://www.anoto.com
2. AED Satellife, http://www.healthnet.org
X. CONCLUSION AND FUTURE WORK 3. Arai, T., Aust D., and Hudson S.E. (1997). PaperLink: a
We have explored the problem space of data collection by technique for hyperlinking from real paper to electronic
NGOs in underserved communities where resources and content. Proc. CHI ’97, p. 327-334.
technological literacy are typically scarce. In particular, we 4. Bergeron, Bryan P. (1998, Aug.). Optical mark
focused on how a particular class of paper forms – those that recognition. Postgraduate Medicine. 104(2).
require only numeric data entry –might be best structured to 5. Chand, A., Dey, A.K. (2006). Jadoo: a paper user
support subsequent automated data transfer to enable interface for users unfamiliar with computers. Proc. CHI
2006. p. 1625-1630.
computerized storage and processing. The main
6. DeRenzi, B., Lesh, N., Parikh, T., Sims, C., Mitchell,
contributions of our work are: (1) an interview-based
M., Maokola, W., Chemba, M., Hamisi, Y.,
416
About a year later, an interesting corollary to Geetha’s Kandukondain Kandukondain [76], took a gentle step in a
statement came from a taxi driver interviewee, in the different direction. The female protagonist in the film, a
neighboring state of Tamil Nadu, a father of two girls in their Brahmin girl, Sowmya is reduced to penury after being
20s. hoodwinked out of her ancestral village home by evil
relatives. She proceeds to move from the village to the city,
“Both my daughters work in Chennai in computers. In the where she rectifies her family’s situation by learning to use
early days, we would never let our (referring to the computers and getting a job as a software engineer. The
Thevar caste) women travel to the city to work, but if they actress not only lifts her entire family out of poverty, but
work for computers that is good. There are good facilities emerges as the most valuable technologist in her firm and
with only ladies housing, and many other families from almost flies out to the US to work as an engineer before better
our village have sent their daughters to work in Chennai sense prevails and she marries her wooing admirer. What is
now.” uniquely compelling about the characterization of Sowmya is
Selvaraghavan, Taxi Driver that she remains ‘traditional’ to the end, but straddles the
Coimbatore District, Tamil Nadu modernity space effortlessly. She dresses conservatively,
except for dream sequences with song and dance, serves her
There is already a growing corpus of work on female family dutifully, and acquiesces to an arranged marriage.
technology workers in Indian call centers which probes this When she does start her software job, her workplace does not
issue more closely, [29], and the local movie theatre as a demand flashy hackers, instead, her interactions with her
starting point into some of the changing social conceptions of engineer boss highlight the importance of urban traditional
women’s work. middle-class work ethos. Despite her success, she does not
turn into a slick skirt-wearing executive, and in doing so, kills
If we think back to Indian cinema and think back to the two birds with one stone. She remains the agreeable screen
typical occupations of women on screen … well, reasonable heroine and makes it to poster child for ‘appropriate ICTD’ –
421
a woman who wins her way out of poverty, and manages stay daughters married off. In the 2000s, that role has been taken
culturally rather untainted. over by the US-returned (or not) software engineer. 3
While there had probably never been an Indian film with a The classic ‘catch groom’ of south Indian cinema was
female character playing an engineer save for the oddball Arvind Swamy, who in the blockbuster hit Roja [121] played
automobile mechanic’s daughter, the female software engineer a computer engineer whose arranged marriage is the early plot
and outsourcing center workers hit the stands in a huge way of the film. Over time, two broad strands of software
around this time [107-113], including films with a reversal of engineer-related marriage scenarios have emerged – the first
roles – where the female lead plays an accomplished in which the engineer is the middle-class hero [122-124] and
technologist of some form, and the male lead is portrayed as
the second is the counterpoint – in which parents are shown
professionally subservient [114, 115]. One film stands out as
hankering after a groom who is a software engineer or NRI,
an interesting continuation to Kandukondain. In the Telugu
and the hero in this case is usually a son-of-the-soil type [76,
film Anand [116], the protagonist Roopa takes Sowmya’s
position a step further. Here, she not only supports herself 125, 126], and while the endgame of these films is often an
through a software job, but sheds much of Sowmya’s demure ode to the anti-hero, the focus on the software engineer as
qualities in reinforcing her independent identity and essential to middle class aspiration is nonetheless highlighted.
challenging traditional behavior roles expected from her by What is also interesting is to look at the contrast between the
suitor and in-laws alike. engineer/technocrat protagonist in cinema around industrial
themes with the engineer in a technology film. The factory
An even more interesting addition to the technology and films frequently had a foil – typically industrialist antagonist
gender empowerment idea comes in the film Swagatam [117]. pitted against the engineer [127], and worker, rather is a the
In it, the male protagonist is a demanding customer at an hero [128]. A “good” engineer typically aligns with the
arranged marriage matchmaking bureau. The manager of this worker instead of the industrialist, breaking class barriers [94,
bureau (coincidentally the female lead) has an online 129]. The flatness of the technology sector is reinforced
candidate repository, and tells the male lead to use the through two factors – first, the “proletariat” is missing in the
computer to filter through his requirements. She kindly software company, unlike for instance the range of films about
reminds the hero that a woman’s greatest trauma is being both factories with workers and service sector offices where
rejected at the arranged marriage meeting, and that technology the ‘peon’ symbolized the underclass [130-132] and second,
should be used effectively to circumvent this problem, and the boss himself (not necessarily herself!) being sharing class
thereby empower women. Given that marriage is practically
with the employees [133, 134]. My reason for excluding
an ever-present theme in Indian cinema, it is not surprising
gender is that women and workplace already come loaded
that several other films have used technology as a go-between
with a large literature as emphasized in the previous section,
for arranged or other marriages [113, 118].
thus the grounds for evaluating “flatness” do not exist in the
Although the computer as a cupid connector has been a same vein – women never played engineers or evil
fairly strong theme through most Indian cinema, the female industrialists (heiresses, though) so the terms for historical
software engineer has been less so. Regional cinema such as comparison do not apply. A woman’s professional ascendance
Bhojpuri, Bangla, and Oriya largely cater to rural audiences in the a technology job is in itself shown as having a range of
(or what in trade parlance is referred to as ‘B circuits’), and power dynamics and it would therefore be confounding to
still feature a lot fewer women in professional positions. In analyse the gender depiction based on flatness alone.
Hindi cinema, on the other hand, female characters who used
computers in Hindi films, have tended to be more westernized In some films, the less than desirable character turns to
or upper class, such as the student who does computer science computers as a means of social acceptability and in others
in college as a qualification, before her eventual marriage, like [135, 136], the drive is primarily economic, often explicitly as
Madhuri Dixit in Hum Aapke Hain Koun [119] or the a means of getting jobs in the US [134, 137]. We find among
‘smartypants’ Preity Zinta in Koi Mil Gaya [120], characters male software engineers a distinction between the north and
significantly different than the Sowmyas or Roopas, where the south Indian stars. In Hindi films, the engineer stereotype is
appropriation of technology into the middle class domain has typically applied to the ‘cool youngster’ hacker characters,
been less of a concern. In our discussion of aspiration, we usually teen idols, [138-141], very comparable to the
return to this issue. Hollywood depictions of computer teenager users from the
B. Computers as Aspiration, Engineer as Hero 1980s and 1990s whereas most of the south Indian depictions
There are few windows more insightful into the aspirational are not necessarily hackers who do cool things with
environment in India than the marriage market. Till date, the
highest catches in the dowry market in rural India go to 3
In the last year South Indian actresses, Renuka Menon, Gayathri
holders of government jobs – which offer among other things Raghuram, and Kanika Subramanian have all left the film industry to marry
– stability, an invaluable element in the rain-dependent software engineers in the US. In the past, Swarnamalya, Ravali, Jyothirmayi,
Maheshwari the common stereotype/joke was of actresses leaving the film
agrarian landscape. But films have long portrayed doctors and industry to marry doctors abroad, which is interesting and a potential topic for
‘big officers’ as good catches for families trying to get a larger study on ‘legitimacy’ of actresses.
422
computers, but rather people who have technology-related with honorifics such as ‘Dear Leader’ or ‘Elder Brother,’
jobs. This distinction is doubly interesting since it not only several of them end up in politics, and the opening few weeks
underlines the differences between the viewing cultures of of a film by a mass star can almost certainly expect to run to
popular Hindi and regional cinemas, a difference that has been full houses, not to mention the celebrations and prayers that
articulated in Ashish Rajadhyaksha’s examination of the accompany any release.
‘westernizing trend’ of Hindi cinema [62], an upper class
urbane narrative, in short. In contrast, south Indian cinema, Several stars, especially in the south, have landed in the
while the fantasy sequences such as songs etc. have roughly ‘mass star’ image trap including Mohanlal and Mammootty
the same ‘modernity’ of the Hindi counterparts, the narrative from Kerala Rajkumar in Karnataka, Balakrishna, Nagarjuna,
themes are much closer to the wider middle-class audience and Chiranjeevi in Andhra Pradesh, and Vijaykanth and
appeal. Thus, the persistence of software engineer characters Rajnikanth in Tamil Nadu, icons with possibly the most
is seen not just in the ‘urban’ south Indian cinema (meaning frenzied fan following of all. The use of technology by these
urban in both setting and primarily appeal) but also in the mass stars is particularly interesting, because their audiences
‘mass’ cinema. are often not the middle class aspirants that watch many of the
urban-based films listed above. Interestingly then, we find the
ideas of technology use in such mass star movies somewhat
C. ‘Mass Films’ and the Computer as Benevolent Superstar
simplistic, but in line with the overall theme of what a mass
star does – save the world.
“Computers can be used to fight evil. We can do anything
with a computer”
Typically a mass star is a perfect son-of-the soil, thus almost
Shivraj, 5th grader
always speaking in the vernacular through a film (unlike in
Devanahalli, Karnataka
Hindi films, for instance, where English is often interspersed
with the Hindi dialogue) excelling and promoting local arts
“Computer can save us. When neighbouring country is
attacking, this is known to our scientists by tracking it on and so on. In a sense, the mass stars’ use of computers is
the computers.” comparable to their use of English. Mass stars generally use
Udhaykumar, 5th grader, English only occasionally, laconically, and with much
Coimbatore, Tamil Nadu emphasis, usually when challenged by an English speaking
adversary or intransigent love interest. In the same vein, a
In this section, we turn to the difference between the ‘mass star’ may use a computer to underline just that he can,
aspriational cinema, which we cover in the previous two even if it is entirely irrelevant to the plot. As a son of the soil,
sections, and the ‘mass films’ which not only highlight an he is able not just to play the game by the rules of the
interestingly different view of technology and modernity, but hinterland, but can match up to modernity [142].
also bear uncanny resemblance to some of the Hollywood
characterizations of the past. By ‘mass films’, we refer here to In Sivaji - the Boss 4 [143], by some estimates the top
the films that are primarily intended towards urban lower- grossing India films of all time, Rajnikanth, arguably India’s
class and rural viewers. The term, derogatory as it may seem, most maniacally followed star, plays a software engineer-cum-
is quite commonly used by filmmakers and analysts alike to social worker bent on rectifying the ills of the Tamil world. In
refer to ‘people’s’ cinema, and several major film stars have his quest, Sivaji uses his voice-recognition-enabled laptop to
come to be associated with this movement. While these can control his vigilante operations and outmaneuver his rivals.
also refer to low-budget movies made for rural audiences, a Similarly the use of computers to maintain and control
fairly large share of ‘mass films’ are actually big budget databases of villains has been used effectively by megastars
productions usually featuring a movie star with a significant, Vijaykanth [144], Chiranjeevi [145], and Ajithkumar [146,
often frenzied fan following. Such actors typically associated 147]. The idea of computers as a tool to assist heroes in
with ‘mass’ cinema tend to have reliably larger-than-life cleaning out society has been used to much effect in recent
characterizations and public images [69]. The dialogue in times, mostly in films featuring online websites for
‘mass films’ is typically political and populist, almost ‘submitting un-rectified crimes’ for appropriate appraisal and
propagandist, and can probably be traced back to the cinema corrective action by the hero on the other end of the internet
of M.G.Ramachandran, who never drank or smoked on connection [148, 149]
screen, committed no act of villainy, remained utterly
attractive to women, helpful to every needy person, and The quote from Shivraj at the start of this was not an
entirely unbeatable in any physical or intellectual contest. His isolated one. We were surprised at how often in our interviews
‘mass star’ successors are somewhat less pious in contrast, but in rural India, the same answer was repeated to us over and
fairly comparable in their abilities to commit superhuman acts, over “Computers can do anything” sometimes ranging from
ricochet bullets, and most important of all, save mankind. children’s fantasies to adults with no direct experience with
Fans of ‘mass stars’ frequently refer to their cherished stars
4
Boss here stands for Bachelor of Social Science
423
computers allocating human attributes to the machines, bodies of thought in development studies. Here, I aim to very
“Computers can teach us English” in clear seriousness. It is gently open the windows into the concerns of critical theory
far fetched to ascribe such notions specifically to films, but and discourse analysis.
it’s worth looking briefly at some of the omnipotent deeds of This analysis of computers and cinema in India brings to
computers. Perhaps a compelling analogy would be that light a number of interesting factors about how filmmakers
computers can do anything, just like Rajnikanth. It is not have chosen to portray technology in Indian cinema, and it is
surprising in this light to know that movie star Mammootty particularly interesting to look at where these portrayals are
was selected as the Brand Ambassador for the rural computing comparable to other cinematic traditions, and where, like in
initiative in Kerala, Akshaya. the aspirational characterizations, it is unique and deeply
related to the prevalent discourse of technology being a means
If we look back before the computer-using days of the of social and economic leapfrog. What is also remarkably
mass stars, shades of fantasy and omnipotence among telling is the much greater prevalence of technology-related
computers in India film were akin to western cinema of the themes and characterizations in south Indian films than in
1960s, especially in early Indian science fiction [150, 151]. north Indian cinema, given the disproportionately higher
Computers and their ability for visual magic broke box office concentration of technology-related industries and
records with ‘Miss World’ Aishwarya Rai’s first screen opportunities in the major south Indian metropolises. These
appearance in Jeans [152]. Here, her brother creates a visual cities like Hyderabad, Chennai, and Bangalore saw
double of Aishwarya to convince twin brothers that they were tremendous demographic, landscape, and social
both in love with two different girls, a reasonable exercise in transformations in the late 1990s, the same time that
computers as movie characters spiked up the charts.
balancing technological savvy with the need for romance in
scripts. Jeans and the hugely popular MTV video of Michael Many of the aspirational characters may well draw their
Jackson’s single ‘Black or White’ [153] featuring faces original inspirations from real-life heroes, often engineering
morphed into one another, or simply the glut of Photoshop graduates from the hinterland who moved to the cities, and
engineers in India, also helped create one of the most enduring saw remarkable growths in their own incomes and social
computer tricks in Indian films, still a common feature of statuses, often moving abroad, enjoying opportunities that
‘mass films’ – the ability of a machine to take a scanned would have been much more challenging to aspire to a
picture of a child lost several years ago, and morph out of it a generation back. Most importantly, many of these graduates
perfectly accurate image of the adult version, auto-adjusting were women. What the cinematic analysis shows us is indeed
for sartorial grace and the few extra facial pounds on the likes a reflection of a larger perception of what counted as the
‘India Rising’ metaphor for the urban middle classes and the
of Vijaykanth. [154, 155].
rural aspiring middle classes, which incidentally, largely
comprise the regular film-going population. It is the
CONCLUSIONS
aspirations of these classes that we see portrayed in a
Is there intentionality in these portrayals? My opening the subsection the South Indian cinema of today. It is thus
article with my personal sob story of failing to sell my scripts decidedly interesting to compare the output geared at these
was meant to imply at least partly that there is at least some aspirational markets with the poorer hinterland markets of the
level of production-level engineering of characterizations. I mass films which take on a simplistic form, carrying forth in
would argue that that question impossible to answer given the selective filters the transplant of middle class imagination of
recursive loop on whether society influences media or vice- technology down to the economic and social chain.
versa.
In conclusion, we turn briefly to the iconoclastic
The idea for this research emerged as after confounding proposition of popular Indian cinema being proactively used
outcomes in other research that indicated a mismatch between to impact development. While the producers may indeed have
peoples’ stated interest in technology and their actual use of been sitting by with whips, were screenwriters of the films we
ICTD projects in rural India. People were very excited about see above thinking of issues of technology and development,
computers and the possibility of their own access to them, but and of impacting society, or is what we see here a sheer
unclear on how technology could be practically useful in their reflection of popular psyche. Irrespective of that, as
lives. Such ideas about technology were further seen to not demonstrated, research has already shown that certain types of
just influence researchers’ estimation of what the likely entertainment media experiences in India have had positive
adoption for certain ICTD projects may be, given the apparent social outcomes, especially in gender-related issues like
enthusiasm about technology, but could also influence the female child protection and domestic abuse prevention. This
populations’ own propensities to invest in ICTD projects, could well be dangerous knowledge for the ICTD community,
without a necessarily clear idea of the value of such a significant part of which is comprised of scholar-activists.
technology. For serious scholars of ICTD, closing this Indeed some of us may stand up and ask if cinema can and
phenomenon off as the ‘buzz’ of technology is not adequate, should be proactively be used to impact technology uptake. It
and an examination of the discourse encompassing this buzz is is important that we think of these questions since we are after
necessary. As ICTD matures as a strong body of independent all in development and for many of us, all research implies the
literature, it will be vital to incorporate learnings from existing responsibility of considering real world intervention
424
capability. When we do start thinking in that direction, 19. Nulens, G. and L. Van Audenhove, An Information
perhaps a social scientist in the room can stand up and point Society in Africa?: An Analysis of the Information
out the parallels of such ideas with Leninistic cinematic Society Policy of the World Bank, ITU and ECA.
propaganda, and ask if technology intake ought to be International Communication Gazette, 1999. 61(6):
promoted at all. After all, asking questions is what we do, p. 451.
decisive action is Rajnikanth’s domain. 20. Heeks, R., Information Systems and Developing
Countries: Failure, Success, and Local
Improvisations. The Information Society, 2002.
ACKNOWLEDGEMENTS
18(2): p. 101-112.
21. Keniston, K., Grassroots ICT Projects in India:
Thanks to Eric Brewer for bankrolling my whimsical work, Some Preliminary Hypotheses. ASCI Journal of
and to Divya Ramachandra, Kentaro Toyama and Jenna Management, 2002. 31(1).
Burrell for their interest in this particular research. 22. Madon, S., Evaluating the Developmental Impact of
e-Governance Initiatives: An Exploratory
REFERENCES Framework. EJISDC, 2004. 20(5): p. 1-13.
23. Donner, J., The Use of Mobile Phones by
1. Roy, B., Naukari. 1954, Bimal Roy Productions:
Microentrepreneurs in Kigali, Rwanda: Changes to
India.
Social and Business Networks. Information
2. Chandra, N., Ankush. 1986: India.
Technologies and International Development, 2007.
3. Bapaiah, K., Aaj Ka Daur. 1985: India.
3(2): p. 3-19.
4. Muthuraman, S., Aarilirunthu Arubathu Varai. 1979.
24. Castells, M., The Informational City: Information
5. Chidambaram, S., Vyapari. 2007: India.
Technology, Economic Restructuring, and the
6. Castells, M., Rise of the Network Society. 1996.
Urban-regional Process. 1991: Blackwell Publishers.
7. Negroponte, N., Being Digital. 1995, Cambridge,
25. Castells, M., Information technology, globalization
MA: MIT Press.
and social development. 1999, Geneva: UNRISD.
8. Hoffman, D.L. and T.P. Novak, Bridging the Racial
26. Roche, E.M. and M.J. Blaine, Information
Divide on the Internet. Science, 1998. 280(5362): p.
technology, development and policy. Information
390-91.
Technology, Development and Policy, EM Roche &
9. Compaine, B.M., The Digital Divide: Facing a Crisis
MJ Blaine (eds.). Avebury: Aldershot, UK, 1996.
Or Creating a Myth? 2001: MIT Press.
27. Mansell, R., Knowledge Societies: Information
10. Hendry, J.D., Social inclusion and the information
Technology for Sustainable Development. 1998:
poor. Library Review, 2000. 49(7): p. 331-336.
Oxford Univ Press.
11. Warschauer, M., Reconceptualizing the Digital
28. Webster, F., Theories of the Information Society.
Divide. First Monday, 2002. 7(7): p. 0-262. 2002: Routledge.
12. Singh, N., Information Technology as an Engine of 29. Radhakrishnan, S., Examining the “Global” Indian
Broad-Based Growth in India, in The Future of India Middle Class: Gender and Culture in the Silicon
and Indian Business, P. Banerjee and F. Richter, Valley/Bangalore Circuit. Journal of Intercultural
Editors. 2002, Macmillan: London.
Studies, 2008. 29(1): p. 7-20.
13. Heeks, R., India's software industry. 1996: Sage
30. Kuriyan, R. and K.R. Kitner. Constructing Class
Publications Thousand Oaks, Calif.
Boundaries: Gender and Shared Computing. in
14. Arora, A., et al., The Indian software services
Second International Conference on Information
industry. Research Policy, 2001. 30(8): p. 1267- Technologies and Development, Dec 2007, IEEE
1287. Conference Proceedings. 2007. Bangalore.
15. Madon, S., Information-Based Global Economy and 31. Roman, R., Diffusion of Innovations as a Theoretical
Socioeconomic Development: The Case of Framework for Telecenters. Information
Bangalore. The Information Society, 1997. 13(3): p. Technologies and International Development, 2003.
227-244. 1(2): p. 53-66.
16. Saxenian, A.L. Bangalore: The Silicon Valley of 32. Warschauer, M., Technology And Social Inclusion:
Asia? in Conference on Indian Economic Prospects: Rethinking the Digital Divide. 2004: MIT Press.
Advancing Policy Reform,. 2000. Stanford, 33. Sein, M.K. and G. Harindranath, Conceptualizing the
California: Center for Research on Economic ICT Artifact: Toward Understanding the Role of ICT
Development and Policy Reform. in National Development. The Information Society,
17. Jhunjhunwala, A., Can Telecom and IT be for the
2004. 20(1): p. 15-24.
Disadvantaged? Rural Development, 1998. 17(2): p. 34. Bailur, S., Using Stakeholder Theory to Analyze
321–37. Telecenter Projects. Information Technologies and
18. James, J., Low-cost computing and related ways of
International Development, 2006. 3(3): p. 61-80.
overcoming the global digital divide. Journal of 35. Burrell, J. and K. Anderson, "I have great desires to
Information Science, 2001. 27(6): p. 385-392. look beyond my world:" Trajectories of Information
425
and Communication Technology use among 58. Dickey, S., The Politics of Adulation: Cinema and
Ghanaians. New Media and Society, 2008. 10(2): p. the Production of Politicians in South India. The
203-224. Journal of Asian Studies, 1993. 52(2): p. 340-372.
36. Morton, H., Islanders in Space: Tongans Online. 59. Hardgrave, R.L., Politics and the Film in Tamilnadu:
Small Worlds, Global Lives: Islands and Migration, The Stars and the DMK. Asian Survey, 1973. 13(3):
1999: p. 235–254. p. 288-305.
37. Franklin, M.I., I Define My Own Identity: Pacific 60. Dwyer, R. and D. Patel, Cinema India: The Visual
Articulations of'Race'and'Culture'on the Internet. Culture of Hindi Film. 2002: Rutgers University
Ethnicities, 2003. 3(4): p. 465. Press.
38. Rai, A.S., India On-line: Electronic Bulletin Boards 61. Appadurai, A. and C. Breckenridge, Public
and the Construction of a Diasporic Hindu Identity. modernity in India. Consuming Modernity: Public
Diaspora: A Journal of Transnational Studies, 1995. Culture in a South Asian World, 1995: p. 1–20.
4(1): p. 31-58. 62. Rajadhyaksha, A., The Bollywoodization of the
39. Bastian, M.L., Nationalism in a Virtual Space: Indian cinema: cultural nationalism in a global
Immigrant Nigerians on the Internet. West Africa arena. Inter-Asia Cultural Studies, 2003. 4(1): p. 25-
Review, 1999. 1(1). 39.
40. Méliès, G., Le Voyage dans la Lune. 1902: France. 63. Mankekar, P., Screening Culture, Viewing Politics:
41. Chaplin, C., Modern Times. 1936, United Artists: An Ethnography of Television, Womanhood, and
USA. Nation in Postcolonial India. 1999: Duke University
42. Atkinson, P. Technology and the Vision of Utopia. in Press.
Imagining the Future: Utopia, Dystopia and Science 64. MacPherson, Y., Images and icons: harnessing the
Fiction. 2005. Monash University, Melbourne. power of the media to reduce sex-selective abortion
43. Bendle, M. Zarathustra’s Revenge: The Sordid in India. Gender & Development, 2007. 15(3): p.
Utopia of Contemporary Science Fiction Films. in 413-423.
Imagining the Future: Utopia, Dystopia and Science 65. Fernandes, L., Nationalizingthe global': media
Fiction. 2005. Monash University, Melbourne. images, cultural politics and the middle class in
44. Youngblood, D.J., Movies for the Masses: Popular India. Media, Culture & Society, 2000. 22(5): p. 611.
Cinema and Soviet Society in the 1920s. 1992: 66. Jensen, R.T., E. Oster, and R. National Bureau of
Cambridge University Press. Economic, The Power of TV: Cable Television and
45. Singer, B., Melodrama and Modernity: Early Women's Status in India. 2007: National Bureau of
Sensational Cinema and Its Contexts. 2001, New Economic Research Cambridge, Mass., USA.
York: Columbia University Press. 67. Holbert, R.L., D.V. Shah, and N. Kwak, Political
46. Akudinobi, J., Tradition/Modernity and the Implications of Prime-Time Drama and Sitcom Use:
Discourse of African Cinema. 1995, IRIS. Genres of Representation and Opinions Concerning
47. Lopez, A.M., " Train of Shadows": Early Cinema Women's Rights. The Journal of Communication,
and Modernity in Latin America. Multiculturalism, 2003. 53(1): p. 45-60.
Postcoloniality, and Transnational Media, 2003. 68. Dolnick, S., Keep Movie Heroes Alcohol-Free,
40(1): p. 48-78. Bollywood Urged, in Washington Post. 2008:
48. Mukherjee, R., Hum Hindustani. 1960. Washington, DC. p. C 07.
49. Maté, R., When Worlds Collide. 1951, Paramount: 69. Pandian, M.S.S., The Image Trap: MG
USA. Ramachandran in Film and Politics. 1992: Sage.
50. Zaslove, M. and C. Nichols, The Jetsons. 1962, 70. Jacob, P., From co-star to deity: Popular
Hanna-Barbera Productions: USA. representations of Jayalalitha Jayaram. Women: A
51. Newman, S., The Avengers. 1961, American Cultural Review, 1997. 8(3): p. 327-337.
Broadcasting Company (ABC): USA. 71. Sharma, A., Blood, Sweat and Tears: Amitabh
52. Roddenberry, G., Star Trek. 1966, Paramount: USA. Bachchan, Urban Demi-God. You Tarzan:
53. Kubrick, S., 2001: A Space Odyssey. 1968, MGM / Masculinity, Movies and Men, 1993: p. 167-180.
Warner: USA. 72. Lehmann, C. and L.S. Starks, Spectacular
54. Bonnière, R., Hide and Seek. 1977, Public Shakespeare: Critical Theory and Popular Cinema.
Broadcasting System (PBS): Canada. 2002: Fairleigh Dickinson Univ Press.
55. Faden, E., The Cyberfilm: Hollywood and Computer 73. Bordwell, D., Making Meaning: Inference and
Technology. Strategies, 2001. 14(1). Rhetoric in the Interpretation of Cinema. 1989:
56. Kepley Jr, V., Soviet Cinema and State Control: Harvard University Press.
Lenin's Nationalization Decree Reconsidered. 74. Gabriel, T.H., Towards a Critical Theory of Third
Journal of Film and Video, 1990. 42(2): p. 3-14. World Films. Film Theory: Critical Concepts in
57. Mishra, V., Bollywood Cinema: Temples of Desire. Media and Cultural Studies, 2003.
2002: Routledge. 75. Bose, N., Dharti Mata. 1938, New Theatres: India.
76. Menon, R., Kandukondain Kandukondain. 2000.
426
77. Khan, M., Mother India. 1957: India. 118. Kathir, Kadhalar Dhinam. 1999, A.M. Rathnam:
78. Roy, B., Madhumati. 1958, Bimal Roy Productions: India.
India. 119. Barjatya, S., Hum Aapke Hain Koun? 1994, Rajshri
79. Chopra, Y., Deewar. 1975: India. Productions: India.
80. Krishnamurthy, R., Thee. 1981: India. 120. Roshan, R., Koi...Mil Gaya. 2003, Film Kraft: India.
81. Majumdar, P., Aarti. 1962, Rajshri Productions: 121. Rathnam, M., Roja. 1992, Hansa Pictures: India.
India. 122. Kranthikumar, Kanden Seethaiyai. 2001: India.
82. Sahu, K., Dil Apna Aur Preet Parai. 1960, Mohan 123. Vasanth, Satham Podathey. 2007: India.
Pictures: India. 124. Renjith, Nandanam. 2002: India.
83. Balachander, K., Manathil Uruthi Vendum. 1987: 125. Narayana, K.L., Rakhi. 2006, Sri Durga Arts: India.
India. 126. Selva, V., Youth. 2002, Ayngaran International:
84. Chopra, Y., Daag. 1973: India. India.
85. Padmarajan, P., Koodevide. 1983: India. 127. Chopra, Y., Aadmi aur Insaan. 1969: India.
86. Kapoor, R., Shree 420. 1955, RK Films: India. 128. Chopra, R., Mazdoor. 1983: India.
87. Kapdi, M., Aaitya Bilavar Nagoba. 1979: India. 129. Chopra, Y., Kala Patthar. 1979: India.
88. Prakash Rao, K.S., Secretary. 1976: India. 130. Hingorani, A., Dil Bhi Tera Hum Bhi Tere. 1960.
89. Balachander, K., Avargal. 1977: India. 131. Anthikad, S., Nadodikattu. 1987.
90. Chopra, B.R., Pati Patni Aur Woh. 1978, B.R. Films: 132. Chatterjee, B., Chhoti Si Baat. 1975: India.
India. 133. Menon, R., Kandukondain Kandukondain. 2000:
91. Rawail, H.S., Patanga. 1949, Varma Films: India. India.
92. Neelakantan, P., Raman Thediya Seethai. 1972, 134. Menon, G., Minnale. 2001: India.
AVM Productions. 135. Bhargavan, Thiru Ranga. 2007: India.
93. Paranjpe, S., Chashme Buddoor. 1981, PLA 136. Ravishankar, Varushamellaam Vasantham. 2002:
Productions: India. India.
94. Kanwar, R., Laadla. 1994: India. 137. Rasool, Unnai Paartha Naal Muthal. 2004, Anandi
95. Mirza, A., Raju Ban Gaya Gentleman. 1992, Sippy Arts: India.
Films: India. 138. Singh, A., Raqeeb. 2007: India.
96. Rama Rao, T., Andhaa Kanoon. 1983: India. 139. Ghosh, K., Fida. 2004: India.
97. Dixit, C.P., Fakira. 1976, Eros Entertainment: India. 140. Roshan, R., Krrish. 2006, Film Kraft: India.
98. Bhogal, A., Zakhmi Aurat. 1988, Manta Movies: 141. Anand, S., Bachna Ae Haseeno. 2008, Yash Raj
India. Films: India.
99. Chopra, B.R., Insaf Ka Tarazu. 1980, B.R. Films: 142. Sivamani, Law and Order. 2002, Jyotirmayi Pictures:
India. India.
100. Sippy, R., Shakti. 1983: India. 143. Shankar, S., Sivaji. 2007, AVM Productions: India.
101. Hussain, N., Yaadon Ki Baaraat. 1973, United 144. Murugadoss, A.R., Ramana. 2002, Oscar Films:
Producers: India. India.
102. Balachander, K., Arangetram. 1973: India. 145. Vinayak, V., Tagore. 2003, Oscar Movies: India.
103. Sasi, I.V., Avalude Ravakul. 1980: India. 146. Vasu, P., Paramasivan. 2006: India.
104. Samanta, S., Amar Prem. 1971, Shakti Films: India. 147. Vishnuvardhan, Billa. 2007, Sri Keerthi Creations:
105. Ramachandran, M.G., Ulagam Sutrum Valiban. India.
1973, Emgeeeaar Pictures: India. 148. Shankar, S., Anniyan. 2005, Lakshmi Ganapathi
106. Anand, V., Jewel Thief. 1967, Navketan Films: India. Films: India.
107. Bhadran, Olympian Anthony Adam. 1999: India. 149. Jayaraj, R., 4 The People. 2004: India.
108. Madan, Pellaina Kothalo. 2004, AA Naluguru Films: 150. Parvez, S., Jumbish. 1986, Ghalib Studio: India.
India. 151. Srinivasa Rao, S., Aditya 369. 1991, Sridevi Arts:
109. Selvaraghavan, K., Aadavari Matalaku Ardhalu India.
Verule. 2007, Sri Saideva Productions: India. 152. Shankar, S., Jeans. 1998, Sri Surya Films: India.
110. Yelleti, H., Ankit, Pallavi and Friends. 2008: India. 153. Landis, J., Black or White. 1991, Michael Jackson:
111. Goud, R.K., Jodi No. 1. 2003, RK Movies: India. USA.
112. Shankar, S., Kunjikkoonan. 2002: India. 154. Perarasu, Dharmapuri. 2006, A.M. Rathnam: India.
113. Prabhakar, T., Neethone Vuntanu. 2002, Lakshmi Art p. 150 min.
Pictures: India. 155. Vidhyadharan, Vaitheeswaran. 2008, Annamalai
114. Vamsy, Avunu Valliddaru Ista Paddaru. 2002, Films: India.
Anandi Arts: India.
115. Selvaraghavan, 7G Rainbow Colony. 2005,
Ayngaran: India.
116. Kammula, S., Anand. 2004: India.
117. Dasarath, Swagatam. 2008, Aditya Ram Movies:
India.
427
Abstract— This study analyzes the nature of regulatory ministries and other government departments is a key feature
independence and its influence on wireless market development of that regulatory quality [5]. Further, empirical studies have
in Tanzania and Botswana. The study finds that the level of found that independence has significant effects on both
regulatory independence is associated with improved market market and regulatory performance [e.g. 4, 5]. However,
conditions. The research has implications for theories of many of these studies are carried out either exclusively in
regulation and market development in low income countries. In
relatively high income countries or in global studies covering
particular the Tanzania case suggests that the independence of
regulation can have secondary effects such as diversity of a broad range of institutional environments. Consequently,
technologies and faster transitions to advanced technologies, they often contain two implicit assumptions that may not
while the reversal of independence in Botswana highlights the apply in low income countries, which in turn raise interesting
need for greater insights into the under-theorized dynamic questions.
nature of regulatory independence. The first assumption is that ministries and other
government departments are able to wield power, from which
Index Terms— regulatory independence, Botswana, Tanzania, independent NRAs are insulated. However, in some low
mobile market development income countries ministries and other government
departments are not well-functioning entities and are unable
to wield power. In these environments, is regulatory
I. INTRODUCTION independence still important and if so how?
collected through in-depth interviews and document analysis. component of general regulatory quality (as will be discussed
Comparative case analyses generate insights into the nature further in the following section), research has suggested it has
of independence and stability as well as their implications for a unique contribution to market development [4, 5], although
particular characteristics of the wireless market development. findings are mixed [e.g. 11]. Studies of the relationship
The paper is structured as follows. Section 2 provides between independence and market development have
background on regulatory independence and its effects. employed both qualitative and quantitative methods, with the
Section 3 describes our research method and data collection latter becoming increasingly prevalent along with data
techniques and is followed by Section 4 in which the two availability and relying predominantly on econometric
country cases are presented. Section 5 provides a brief analyses of the formal components of independence.
cross-case analysis and in section 6 discussion and Measures of formal independence vary in their complexity,
conclusions are presented. ranging from simple dummy variables to highly complex
measures [e.g. 4, 5, 11]. The most recent and extensive of
II. REGULATORY INDEPENDENCE AND ITS EFFECTS these is the EURI-I index, which is based on eleven general
While research has found that regulatory independence has measures 1 of formal regulatory independence including: 1.
significant effects on telecommunications market multi-sector, 2. multi-member, 3. funding, 4. reporting, 5.
development, to date it is unclear the extent to which these shared roles, 6. legislative appointment, 7. fixed terms, 8.
findings apply in lower income countries. In particular, given renewable terms, 9. staff, 10. budget, and 11. experience [5].
the differences in the state of institutional development in While many of these variables are self-explanatory, several
some lower income countries and the potential challenges of deserve further discussion.
maintaining institutional gains, it is unclear how The multi-sector measure is based on the presumption that
independence will influence market development if at all. a multi-sector regulator will achieve a higher level of
These questions of influence are particularly salient for the independence from any one ministry, with similar reasoning
wireless market both because it has experienced such rapid for multi-member regulators (1) versus single member (0).
growth in low income countries and because evidence Reporting indicates whether regulators are required to report
suggests the influence of regulatory quality is lower in this only to the Minister (0), to both the Minister and the
sector. These issues are discussed in turn below. legislature (.5) or only the legislature (1). The shared roles
measure indicates whether or not a regulator shares its duties
A. Regulatory Quality with the government (0) or is granted exclusive powers (1).
For years scholars have argued for the necessity of Finally, experience is measured as whether or not the NRA
regulatory independence both for proper functioning of the has been in existence for at least two years.
regulator and to enhance both regulatory and market While the EURI-I index provides a complex measure of
performance [6, 10]. Regulatory performance is the extent to independence it is developed only for the European context.
which regulations fulfill their immediate objective, such as Hence, it is unclear the extent to which their results that
lowering interconnection prices, collecting universal service regulatory independence reduces interconnection rates only
revenues or lowering entry barrier for new entrants. These when the public telecommunications operator is partially
outcomes are expected to create conditions which in turn owned by the state, are generalizable to lower income
improve overall sector performance by, for example, contexts. In addition, a question remains to what extent this
lowering consumer prices, extending networks, and operationalization of regulatory independence relates to
improving service quality. regulatory issues other than interconnection.
Regulatory independence has been characterized by That effects of regulatory quality might be different in
institutional arrangements that foster clarity of roles of the lower income countries are suggested by the research of
regulator, and accountability and transparency in the process Gutierrez [4] and Wallsten [12]. Gutierrez [4] found that
of regulatory decisions [6, 10]. The delegation of authority when separating low versus high income Latin American
and need for independence is driven in part by the desire to countries the influence of regulatory quality on market
establish credibility but also in some cases to insulate the development is indeed different. In particular, regulatory
regulatory policies from future politically-driven changes quality has less of an impact on market development in the
(Gilardi 2007). Further, independence is frequently low income countries. The author proposes this may be result
established at the time the autonomous regulator is of the greater difficulty in implementing a positive regulatory
developed, a frequent driver of which is privatization and environment in low income countries or that their
liberalization of monopoly public telecommunications institutional changes were more recent and therefore have yet
operators (PTOs).To this extent, in a study of PTO ownership to produce changes in the market. Conversely, the variables
in Europe, Bauer [11] found that levels of independence were of competition and privatization have a greater effect in lower
highest for those nations that were slow to privatize their income countries, which may be attributed to the greater
PTO. Thus, he concludes the level of independence may be room for improvement. Similarly, in a study of the impact of
driven in part by the level of state ownership in the PTO.
Independence is a multidimensional construct, with both 1
Here we exclude the study-specific measure of interconnect powers,
formal and informal components, each with several which measures the extent to which the regulator has powers over
interconnection. This measure is useful to studies of interconnection but may
dimensions. While it is often conceptualized as one not be appropriate for studies of other regulatory issues.
429
an independent regulator on market performance in Africa regulator, (2) the degree of freedom from political and
and Latin America, Wallsten [12] found the regulatory industry interference, clarity of regulatory functions,
variable on its own is insignificant in explaining teledensity. accountability and transparency, and (3) the legal basis of the
Hence, while research concerned with the implications of creation of the regulatory body.
regulatory quality is becoming more nuanced, with These studies suggest that independence will be influenced
independence, competition, and privatization having distinct by a variety of contextual factors. In particular, factors that
contributions to market development, it is unclear the extent influence the ability of governmental departments to fulfill
to which the direction of these developments are relevant for their role in terms of providing checks and balances will
studying independence is low income countries. The impact independence, positively or negatively. For example,
following section discusses the broader context of as described by Jain [18] and Samarajiva [19] in India and Sri
independence and its implications for regulatory Lanka respectively, the judiciary played an important role in
independence in low income contexts. mediating the relationship between the PTO, ministry and
regulator. Whereas in the Indian case the judiciary challenged
B. The Context of Independence
the position of the regulator, in Sri Lanka it was supported.
While at some level challenges to independence are Further, in both cases the judiciary fulfilled its role in
universal, in different contexts they are likely to vary in their providing checks and balances. However, does this imply that
degree. For instance, while nearly all government in other environments where ministries, judiciaries and other
departments face resource constraints, in low income administrative bodies are unable to wield power, that
countries the lack of resources of the regulator and society in independence is still important?
general raise special issues for independence. Examples
include the inability to pay competitive wages to regulatory C. Changes in Regulatory Independence
staff, resulting in high levels of employee turnover as well as A second assumption implicit in much of the research on
the general scarcity of qualified personnel resulting in the the effect of regulatory quality and independence on market
necessity of hiring staff with potential conflicts of interest development is that gains in quality and independence are
(e.g. are former Ministry or PTO employees). This occurred, maintained. As noted by Stern [6, p. 69], “Even when
for example, in Sri Lanka where the former Managing formally independent regulatory agencies have been set up,
Director of Sri Lanka Telecom (SLT) was named Director as in Argentina or Hungary, the question remains as to how
General of the regulatory authority [13, 14]. far their independence is (a) genuine and (b) sustainable.”
This lack of resources may not only affect the regulator Despite the possibility of reversal little attention has been
itself, but also government departments that serve as partners paid to its implications.
and possibly opponents of the regulator as well. For example, This is not to say, however, that the dynamic nature of
also in Sri Lanka, Balasooriya et al. [13] found that the Fair regulation is not considered. Indeed, several econometric
Trading Commission, which by law was to serve as the studies use time series data that capture changes in regulatory
competition authority for the country, was “almost quality and independence over time. For example, in the
inoperative despite its investigative and quasi-judicial power EURI-I index, which measures independence in the 15
assigned by its Act” (p. 388). original EU member states between 1998 and 2003,
Such circumstances suggest that measures of a regulator’s regulatory institutions declined in two countries, remained
independence that focus solely on the regulator may not constant in four countries, and improved in nine countries [5].
suffice. Earlier studies on the effects of governance or However, because declines are far fewer than improvements
regulatory quality on telecommunications market their implications are not obvious from the broader analysis.
development focused on the broader governmental It is also important to note that even where the changes in
environment. For example, studies examined the effect of the regulatory quality and independence are positive, the effects
credibility of the policy regime [15, 16] through use of the may not be. Gutierrez [4], in a study of 22 Latin American
POLCON index [15], a measure of veto points within a countries during the 1980-1997 period, examines changes in
government that in turn constrain any one political actor from regulatory quality and suggests that effects on market
changing government policy. The studies found that over a development may be nonlinear. In particular, he proposes that
wide range of high and low income countries those with at low levels of regulation, further enhancement of the
lower likelihood of arbitrary policy changes had higher regulatory framework at first increases telecom market
growth in fixed teledensity. performance but then over time its impact is slowly
Examining the relationship between general regulatory diminished.
quality exclusively in the telecommunications realm is the The possibility of nonlinear effects of institutional
work of Gutierrez and Berg [17] and Gutierrez [4]. In these improvements begs the question as to how declines in
two studies, the measure of regulatory quality is developed institutional quality affect market performance. However,
from a dichotomous measure of of the the presence (1) or prior to understanding the effects of declines, more
absence (0) of a regulatory framework, to one in which information is required about their fundamental nature (i.e. in
regulatory development is measured as an index. The index what areas of regulatory quality are declines most common)
includes (1) the separation of operations and regulatory and in what conditions are they most likely to occur.
activities, although not necessarily the existence of a separate
430
D. Regulatory Quality and Wireless Market Development regulatory governance, albeit a general one. Further, both
The dynamic nature of regulatory independence raises a countries have in recent years experienced relatively strong
second issue which is its differential impact on wireless telecommunications regulatory governance, with Botswana
market development. In particular, Andonova [8] finds that being recognized by the ITU [22] and Tanzania by the 2006
whereas an increase in the institutional quality positively ICT Investment Summit [9]. Furthermore, both countries
affects the number of Internet hosts, there is no effect for have recently implemented a converged licensing framework
cellular penetration. Further, the static effects of regulatory as two of the first countries in the continent. The
quality were less for cellular penetration than for fixed line implementation of such licensing frameworks has the
services. potential to significantly affect wireless market development,
Andonova [8] proposes these findings are due to wireless as indeed has been the case in Tanzania [23].
infrastructure being less of a sunk investment, one that can be In addition to these similarities there are differences as
redeployed and thus less prone to expropriation and hostage well. First, the two countries come to the strong regulatory
taking. While this may be true in theory, the rapid pace of positions via different routes, with Botswana having a past
technological change may hinder the redeployment of characterized by political stability and strong governance
wireless infrastructure. Other reasons for the lack of influence throughout the government, as compared to Tanzania that has
of regulatory quality on wireless services are possibly the had a history of political and economic turmoil. Second, most
lower level of regulation, greater prevalence of private recently while Tanzania’s regulator has maintained its
ownership and the high level of demand. independence, Botswana has decreased [9]. This difference
These factors suggest that a closer examination of the enables a comparative analysis of regulatory quality as well
relationship between wireless market development and as its link with wireless market development. These dynamics
regulatory quality is required. Earlier studies of regulatory in regulatory governance in both countries, historically and
quality [e.g. 12, 17, 20] actually used wireless market most recently, facilitate comparisons that can identify their
development as a predictor of fixed line market growth. underlying factors.
Clearly greater understanding of the influence of regulatory Data were collected through 46 face to face interviews
quality on wireless market growth (both fixed and conducted in Tanzania and Botswana during September
mobile/cellular) is required. Also, whereas studies have –November 2006, with managers at incumbent and mobile
focused primarily on market growth in terms of access and/or operators and Internet Service Providers (ISPs), as well as
users as well as efficiency of the sector, the relationship with policy makers from ministries responsible for
between regulation and the wireless market given the less telecommunications and regulators from the national
central role of regulation may require a more nuanced regulatory authorities. Additional data were collected through
understanding. Further, given the rapid growth of wireless policy and document analysis from 2006 until present.
markets worldwide, it may be time to look at higher level Data were analyzed through a combined deductive and
service criteria than merely access. inductive analysis. The goal of the deductive analysis was to
Thus, this research seeks to provide further nuance to characterize the degree of independence according to the
theories of the role of independence in telecommunications EURI-I index as developed by Edwards and Waverman [5]
market development. Research on regulatory independence and changes in policies and market development, while the
in the 80’s and 90’s was primarily normative, given the inductive analysis was used to identify possible underlying
limited experience of regulators. Next, qualitative case study factors influencing these phenomena, with a particular focus
research began to emerge and provide insights into on the dynamics of regulatory independence (i.e.
independence as a contribution to regulatory quality. As sustainability of gains in independence), and the role of, and
experience with independence grew and data became relation between, Minister, judiciary, and other
available econometric studies began to provide systematic administrative bodies with the regulatory authorities.
evidence of its effects. Here, we seek to contribute to research
based on in-depth case studies that can subsequently inform IV. REGULATORY INDEPENDENCE AND WIRELESS MARKET
more systematic analyses. In particular, we seek to provide DEVELOPMENT IN TANZANIA AND BOTSWANA
insights into the nature of independence, how and why it In the following sections the development of regulatory
changes and how it affects wireless market development in bodies and markets for wireless services are discussed. Both
low income countries. These insights may subsequently cases start with an overview of regulatory and wireless
generate greater nuance in both qualitative and econometric market developments as related to the implementation of
studies. converged licensing frameworks, followed by a more detailed
analysis of these countries’ evolving regulatory
III. METHODS independence.
The study takes a comparative, explanatory case study
approach [21]. Tanzania and Botswana have been selected as A. Tanzania Case
cases for their similarities as well as differences. They are
both members of the Southern African Development
Community, which generates a common approach to
431
by industry, Parliament and Minister play a role in approval July 2007 Sasktel International of Canada started a three year
of TCRA’s budget. contract to take over management control8.
Nevertheless, as per the EURI-I index, there are a number Further, Tanzania’s broader institutional endowment is
of indicators that would actually suggest ‘dependence’. For known for a highly dominant Executive (including President
example, Tanzania’s staff number of 97 is far below that in and Ministers) in national policy making processes. While
Europe, however for African standards is relatively high [9]. Parliament has powers of scrutiny, due to the large majority
Additionally, TCRA primarily reports to the Minister while of one party in Parliament, at the end the Legislature thus
the legislature is hardly involved. Parliament does receive cannot easily hold the Executive accountable [27]. This
TCRA’s budget and annual report from the Minister, but no might explain the limited role of Parliament in
specific approval powers etc. are stated in the TCRA Act of telecommunications matters; besides enacting the TCRA Act
2003. The board members of TCRA are also not appointed by of 2003, Parliament’s single formal role in relation to TCRA
Parliament, but through a committee established by the regards to how it may provide funds to TCRA, and that it
Minister. Further, as per the TCRA Act of 2003, the Minister shall receive TCRA’s annual report from the Minister.
has the power to engage in many regulatory activities. For However, this does not explain the limited role the ministry
example, TCRA is not allowed to award or cancel licenses has played.
with an exclusivity period, universal service obligations or The limited attempts at protecting the incumbent may
any license for a term of more than five years without partially be due to the ministry’s lack of resources. Currently,
consulting the Minister. the Ministry of Infrastructure Development employs roughly
These factors implying certain degrees of dependence vs. 100 people responsible for departments spanning from postal
independence have also changed throughout the years. In to transportation, but as of 2006 only 6 staff are assigned to
favor of independence is first TCC’s evolution from a single telecommunications and ICT matters [23]. Consequently, the
sector regulator into a converged regulator responsible for ministry realistically does not have a strong influence on ICT
postal, telecom and broadcasting. Second, in some areas the and telecommunications policy or regulation. As one
Minister’s powers and influence in the regulatory process manager from the private sector put it, “the regulator is trying
have decreased. Third, during the time that the regulator was to compensate for lack of policy from the Ministry’s side”.
purely in control of telecom, under the name TCC, board Furthermore, the manager mentions that the Minister has an
members could renew their terms an unspecified number of overseeing role, but that in practice TCRA is “independent”.
times, whereas currently more stringent rules apply where Another manager says “frankly speaking, the ministry
board members can only renew their appointment once. doesn’t have enough staff. They should actually give more
However, regardless of developments in the direction of input”. This relative lack of involvement of the ministry in the
more independence for the regulator, the Minister is still telecommunications sector might be further reflected in the
awarded an extensive formal role. Hence, the question lack of engagement of operators with the ministry (as
remains: why is TCRA referred to as an “independent” opposed to the regulator), as is explained by regulatory affairs
regulator? Particularly interesting is the fact that the managers at two mobile operators.
incumbent TTCL remains partially state-owned, which in
many countries has led to government protection and delayed
B. Botswana Case
implementation of liberalization due to incumbent-initiated
court proceedings.
First, in Tanzania, little evidence exists of Regulation and Wireless Market Development in Botswana
incumbent-initiated court proceedings. In interviews with The case of Botswana shows a very different history of
managers at TTCL’s competitors, including mobile operators regulatory quality and wireless market development than
and ISPs, attempts at delaying the implementation of Tanzania. Botswana, due to its population of 1.8 million, has
liberalization mechanisms through court proceedings were a very small market with, besides an incumbent fixed line
not mentioned, and stands in stark contrast with a country like operator, until very recently only two mobile operators.
South Africa [see e.g. 9]. There exist two potential underlying While seemingly a small number of mobile operators, at the
reasons for the lack of judiciary involvement. First, Tanzania time of licensing in 1996 when market liberalization started
is known to have a fairly weak judicial system, which has and the Botswana Telecommunications Authority (BTA) was
restrained market development in many sectors due to the established, many thought that only one mobile operator
lack of (threat of) enforcement, leaving government would be viable due to the small population size. By 2007
unchecked by law [25, 26]. This may have refrained TTCL already 1.2 million people used mobile phones [28]. But
from taking action. nevertheless, it was felt that competition in the Botswana
A second reason for limited action from the side of TTCL telecommunications sector developed “unevenly across
might lie in the problems with management control, which different regions of the country and at different levels” [29, p.
first changed from government to Celtel during TTCL’s 3], as outside of cities and major districts, provision of
partial privatization in 2001. Even though in the years after telecommunications services often remained limited or even
Celtel extended its shareholdership to 65% shares while non-existent. A new service neutral licensing framework was
government remained 35%, problems in management control 8
led to the government taking back management control. In See e.g. http://wirelessfederation.com/news/sasktel-takes-over-at-ttcl/
Last accessed February 18, 2009.
433
developed in hopes of “this imbalance [to] be corrected” [29, generated a significant new impulse to market entry as it has
p. 4]. done in Tanzania. Moreover, even though the PTO market
On June 20, 2006 the Minister of Communications, Science segment shall only in 2009 be considered for further
and Technology announced the intent to further liberalize the liberalization and market entry [31], as explained at the
telecommunications sector through five separate changes to Ministry of Communications, Science & Technology, “The
license conditions, as well as having the incumbent rebalance current regime means there is no intent to get more than three
tariffs and allowing new entrants to tender for service neutral main voice operators”. The future licensing of new national
national licenses [30]. The first five measures were network providers will depend on the Minister, who is in
accomplished by March, 2007 through the introduction of the charge of awarding licenses.
“Service Neutral Licensing Framework in the Era of Since April 2008 BTC has started its mobile operations
Convergence” [31]. Prior to the change in the licensing under the brand name Be Mobile. Managers from industry
framework, the market was categorized into “Fixed, Cellular, explain that indeed more competition is felt due to the
Internet Service Providers (ISP), Satellite and Data etc.” regulatory changes. Perhaps as a result of this, more advanced
Market segments determined as “non-competitive” were services are introduced: Mascom Wireless has launched a 3G
restricted in terms of the number of players within the and 3.5G HSDPA network10, while Orange has introduced
segment. In Botswana these were Fixed and Cellular, whereas the Blackberry in February 2008 on its GSM/GPRS/EDGE
ISPs, data service providers and paging services providers network11. And, even though the new regulations had ISPs
have been determined to be competitive9. Hence, only one remain dependent on infrastructure provision by third parties
fixed and two cellular providers were licensed, while a [9], as of July 2008 value added network services (VANS)
significantly larger number of ISPs and data providers providers (e.g. ISPs) may tender for Fixed Wireless Access
obtained licenses. Spectrum in a number of frequency bands [32].
The new service neutral licensing framework required the
existing fixed and mobile operators (BTC, Mascom Wireless Regulatory Independence in Botswana
and Orange Botswana) to obtain so-called “public Similar to the Tanzanian case, Botswana’s regulator BTA
telecommunications operators” (PTO) licenses. Under this scores well on a number of variables of Edwards and
license, any one of them is eligible to provide both cellular Waverman’s [5] EURI-I index. BTA scores well on (1)
and fixed services. In addition, all PTO licensees have multi-member board control, with 5 non-executive members;
become eligible to operate the international gateway and to (2) fixed term appointments of board members (4 years); and
“self-provide” – a relief to mobile operators that previously (3) experience – with 12 years of experience, Botswana has
were required to use incumbent BTC’s backbone. The three significant regulatory experience. In addition, BTA has both
PTOs however are still not eligible to provide value-added telecommunications and broadcasting departments, and as
Internet services. Hence, BTC continues to use its subsidiary such may be perceived as a multi-sector regulator. However,
Botsnet for Internet services provision [9]. there are separate boards for both sectors.
The introduction of a service neutral license nevertheless There are a number of factors related to independence as
was a surprise to some market players. As explained by identified by Edwards and Waverman (2006) that BTA does
regulatory managers at incumbent BTC, it was expected that not score very well on, however. First, similar to Tanzania,
a third mobile operator was to be licensed instead of a fully the staff number of BTA is low compared to for example
revamped licensing framework to be implemented. BTC European regulatory authorities, at 70 staff in 2006.
applauded this decision, as BTC was “not sure” if it could Nevertheless, for African standards it is very high;
have applied for a mobile license otherwise – as two particularly when taking into account the low population size
managers explain. of Botswana (1.8 million) [9]. A second factor that BTA does
Interestingly however, three regulatory staff at regulatory not score very well on is the lack of involvement of the
authority BTA explain that BTA actually recommended the legislature. Botswana’s Parliament (National Assembly) does
Minister to license a 2nd fixed and 3rd mobile operator after not have a dedicated communications portfolio committee,
stakeholder consultation processes. Nevertheless, the and although the Parliament has passed the
Minister had the authority to either accept or refuse BTA’s Telecommunications Act of 1996 as well as the Amendment
recommendations. Further, as a manager at regulator BTA Act of 2004, Parliament does not have significant formal
explains, the incumbent BTC lobbied the government – powers in relation to BTA. It is only eligible to (1)
concerned that under BTA’s recommended plan BTC could appropriate money for the Authority’s fund; and (2) receive
not have a mobile license. Hence, according to a BTA an annual report and auditing account within 30 days after the
employee, the Minister decided to convert the three major Minister’s reception of both [33]. This also implies that there
operators’ licenses into one service neutral license. are two other factors not in favor of independence: there is no
Thus, while throughout the years the growth in ICT
connectivity in Botswana has been greater than expected, the 10
See http://www.cellular-news.com/story/33201.php. Last accessed
introduction of the new licensing framework has not September 20, 2008.
11
See http://www.orange.co.bw/press_room/news_page.php?newsID=11
“Orange and RIM Introduce BlackBerry for the first time in
9
See http://www.bta.org.bw/licensing.html Last accessed March 30, Botswana”, February 14, 2008. Last accessed September 20,
2008. 2008.
434
legislative appointment of board members, and BTA reports budget has been observed immediately: in 2006, BTA
only to the Minister. Third, BTA’s board members may provided to the treasury Pula 10 million, and in 2007 Pula 2.5
renew their appointments an unspecified number of times. million (respectively about USD $1.5 million and USD
Finally, the shared roles of BTA and Minister, and BTA’s $375.000) [28]. Additionally, the government can now take
budget, seemingly are indicative of a limited degree of part of the profits of BTA. A manager at BTA suggests that
independence. The Minister has the power to interfere in a the Minister taking back power is a trend observed in more
number of regulatory areas. First, as per the southern African countries, including Lesotho, South Africa,
Telecommunications Amendment Act of 2004, the Minister and Namibia [9].
has to approve all licensing of fixed line and cellular Indeed, that these developments represent a transfer of
telephone service and may set licensing fees. Additionally, power from the regulator back to the Ministry is confirmed by
the Minister has the “power to make regulations” [33, B101], people in the industry and at the regulator as well. However,
and thus may interfere in many areas of regulation. while theoretically the Minister does have more power,
Thus, as much as six indicators have been identified that reality might be slightly different due to underlying resource
could imply a low level of independence. Further, some issues. As one of the Directors at BTA states, “even though
issues are indicative of the dynamics, including reversal, of the Minister legally has more power […] at present that is not
independence. Especially in its early days, BTA was an issue”. Further, another Director at BTA indicates, “there
perceived as a model regulator, and moreover, governance in is a lot of consultation between the ministry and BTA – BTA
Botswana in general has been of a high standard. BTA’s has a lot to say. Liberalization was initiated by BTA. The
exemplary regulatory governance and independence is for ministry relies a lot on BTA because it is better resourced”.
example reported in a 2001 report by the ITU which states The Director furthermore continues: “The Ministry is really
that “the Botswana experience also offers a number of world under-resourced. Most work is carried out by BTA. A policy
models. Among these are that BTA has achieved a high level direction should come out, which BTA would then have to
of independence as measured by the lack of influence from implement.” This however does not always happen. As the
the government in implementing its mandate. Its virtually Director continues, “a problem […] is that BTA basically
unfettered authority to license operators and self-financing made the national plan. This is not desirable for checks and
operation may also develop as a world model. BTA further balances.”
provides good models of strong legal processes in carrying Nevertheless, regardless of these issues, generally
out its regulatory mandate.” [22]. Additionally, with regard to speaking Botswana is still perceived by many people,
licensing, the ITU (2001) report even states that “BTA is one including those directly involved in Botswana’s private
of the rare regulatory bodies that has been given almost sector, as having a very good regulator.
complete freedom to decide which services are to be licensed,
how many licenses should be granted for each service and V. CROSS CASE ANALYSIS
which operators are to be awarded a license” [22, p. 27]. Not
only the ITU has noticed Botswana’s good governance; The cases of Tanzania and Botswana bring to bear two key
throughout the region regulators refer to Botswana’s strong aspects related to independence: the relation between
governance. For example, a former regulator from South Minister and regulatory authority as well as sustainability of
Africa perceives BTA to be a “model regulator”, which by regulatory independence. First, according to the
industry managers and regulators across the Southern African independence indicators by Edwards & Waverman’s EURI-I
continent is seconded [9]. Botswana’s impressive regulatory index [5], Tanzania currently scores better. The difference
governance is furthermore acknowledged in previous however primarily lies in terms of board members of the
research. McCormick [34], suggests in her article with the regulatory authority of Botswana being renewable an
revealing title “Telecommunications reform in Botswana: a unlimited number of times. Further, in Botswana, the
policy model for African states” that Botswana has been able Minister has significant control over the budget of the
to develop a model of policy and regulatory governance regulatory authority.
known by significant transparency in decision making. However, in both countries there are shared roles between
However, the amendments to the Telecommunications Act the Minister and NRA, and in both countries the Minister
of 2004 have led to BTA’s degree of dependence becoming theoretically has significant room to implement regulations.
point of debate. The Telecommunications (Amendment) Act The case of Tanzania however showed that even though this
of 2004 substitutes a number of sections of the 1996 formally might be the case, it does not necessarily stand in the
Telecommunications Act that give more ‘power’ to the way of independence in day to day operations. Due to limited
Minister. In particular, these changes enable the Minister to 1. resources at the ministry, Tanzania’s regulatory authority has
determine the use of surplus funds that accrue to the enjoyed full freedom to design regulations, and even has
Authority, 2. make regulations, on the recommendation of the played a major role in drafting policies; which typically is a
Board, 3. set licensing fees, and 4. approve all decisions on Ministerial responsibility.
the licensing of fixed line and cellular telephone service [33]. While over the years in Tanzania formal independence has
Thus, a number of decision making powers formerly under increased, in Botswana it has actually decreased. The
the authority of the BTA Board have been transferred back to Minister’s decision to ignore BTA’s recommendation to
the Minister. Further, the Minister’s control over BTA’s tender for a third mobile operator but instead to allow the
435
incumbent fixed line operator to start offering mobile services Indeed, Bauer [11] suggests that complaint records and
illustrates this point. Thus, the cases provide evidence of court decisions may be an adequate measure of one
independence being a dynamic phenomenon. Under what component of independence, as long as they are free of a
conditions are reversals most likely? Comparing Botswana systematic bias. While specific case records may be difficult
and Tanzania one can conclude, although tentatively given to obtain and bias even more difficult to assess, as the above
the limited number of cases, that reversals are more likely in quote suggests, a measure that reflects simply the
countries where the government owns a larger stake in the involvement of the courts may adequately differentiate
incumbent, and thus has incentives to protect the incumbent. different levels of independence.
Finally, the cases provided evidence that suggests that Also, the case data, together with other instances of
regulation has an important role to play in determining the reversals, suggest there are a variety of reasons for these
number of carriers and hence competition, which has been reversals, which may be more or less strategic. The case of
shown to have positive market effects in many countries. Botswana reflects a strategic reversal in that it enabled the
Tanzania’s implementation of its converged licensing ministry to create market conditions favorable to the
framework with an open approach to market entry led to incumbent. However, for example the regulator in Sweden,
market entrance of 4 new operators and additionally an which was recently identified by the EU as needing greater
impetus to innovation (i.e. the offering of more advanced independence due to a court ruling that resulted in reduced
technologies). Botswana’s approach led to significantly less powers for the regulator 12 , reversals can occur for other
impetus, even though competition of course was stimulated reasons as well. This together with the previously mentioned
with there being a third mobile operator now. declines in independence reported in the EURI-I index (see
section 2.3), indicate that independence reversal is an issue
VI. DISCUSSION AND CONCLUSIONS common to both high and low income regions. Hence, a more
nuanced understanding of the dynamic aspects of
This study analyzed the nature of regulatory independence independence, as well as the day to day practices in case of
and its influence on wireless market development in shared roles between Minister and regulatory authority, as
Botswana and Tanzania, and as such, the study has well as the causes and market effects of reversals is required.
implications for theories of regulatory independence in low Further, insight into reversals in regulatory independence
income countries. may also contribute to understanding reversals in the more
First, we find that traditional notions and measures of general realm of regulatory quality [see e.g. 35].
independence may under- or overestimate a regulatory Finally, the research findings reported here and elsewhere
authority’s status, particularly when the broader institutional suggest that regulatory independence has different effects on
context is not taken into account. When other administrative fixed versus wireless market development. Indeed, the lower
bodies, such as the ministry, legislature or judiciary, are levels of regulation in these markets and their rapid growth
weak, independence of an NRA may generate greater across all nations, independent of institutional arrangements,
freedoms than suggested by their legal status. As reflected in has led some researchers to question the role of regulation
the case of Tanzania, to date these freedoms have generated altogether. First, these questions may be more salient in high
positive outcomes for market development. However, this income countries where the diversity of checks and balances
freedom exists with minimal checks and balances, which are is in general greater. Second, if indeed the effects of
typically a requirement for stable political systems. This regulation on traditional measures of market development
finding suggests that independence of the regulator should such as teledensity and efficiency are universal, it may be that
not be assessed solely by characteristics of the regulator itself. new measures of market development are needed. The case of
Measures must further consider: what is the regulator Tanzania suggests that measures such as the diversity of
independent from? technologies and speed of transition to advanced
This may be interpreted as a call to return to studies that technologies, both likely important to consumer satisfaction
examined the broader construct of regulatory quality and not and to enhancing the true benefits of wireless technologies,
just independence. Indeed, some measures of regulatory may be fruitful areas for future research.
quality such as the POLCON index, which measures veto
points in a government, reflect the existence of checks and
balances. However, the checks and balances related to the REFERENCES
independence of the regulator are more subtle. As suggested [1] Saunders, R.J., J.J. Warford, and B. Wellenius, Telecommunications
and economic development. 1983, Baltimore and London: Johns
by Samarajiva [19, p. 81] who described the increasing
Hopkins University Press for the World Bank. 395.
credibility and independence of the regulator in “The [2] Bonciu, F. and H. Williams, The contribution of the mobile industry to
incumbent operator appealed against one of the economic development in Romania. Info, 2006. 8(5): p. 37-55.
interconnection decisions, and for the first time the appeal
went to the courts, not to the political and administrative 12
See “Commission opens three new cases on independence and
authorities as in the past.” In this case the incumbent may effectiveness of telecoms regulators in Latvia, Lithuania and Sweden”,
have had the right to appeal to the courts all along, but never IP/08/1343, September 18, 2008. Retrieved from
http://europa.eu/rapid/pressReleasesAction.do?reference=IP/08/1343&type
needed to. =HTML&aged=0&language=EN&guiLanguage=fr. Last accessed
September 22, 2008.
436
[3] Williamson, O.E., Credible commitments: Using hostages to support Botswana, at the BTA Offices. 13 june 2007. 2007: Gaborone,
exchange. American Economic Review, 1983. 73(4): p. 519-540. Botswana.
[4] Gutierrez, L.H., The Effect of Endogenous Regulation on [30] Venson-Moitoi, P., Press Statement June 20, 2006: Further
Telecommunications Expansion and Efficiency in Latin America. Liberalization of the Telecommunications Sector. 2006: Gaborone,
Journal of Regulatory Economics, 2003. 23(3): p. 257-286. Botswana. Retrieved September 20, 2007 from
[5] Edwards, G. and L. Waverman, The Effects of Public Ownership and http://www.bta.org.bw/Liberalisation/Press%20Statement%20-%20L
Regulatory Independence on Regulatory Outcomes. Journal of iberalisation%203.pdf.
Regulatory Economics, 2006. 29(1): p. 23-67. [31] BTA, Service-Neutral Licensing Framework in the Era of
[6] Stern, J., What makes an independent regulator independent? Convergence. 2007, BTA: Gabarone, Botswana. p. 7.
Business Strategy Review, 1997. 8(2): p. 67-74. [32] BTA, Public Notice Radio Spectrum Licensing Tender No. BTA
[7] Varoudakis, A. and C.M. Rossotto, Regulatory reform and BTA/PT/02/2008.2009: Public Invitation to Tender for Fixed Wireless
performance in telecommunications: Unrealized potential in the Access Spectrum in the Frequency Bands 1785 – 1805 MHz, 2500 –
MENA countries. Telecommunications Policy, 2004. 28: p. 59-78. 2690 MHz and 3400 – 3600 MHz. 2008: Gaborone, Botswana:
[8] Andanova, V., Mobile phones, the Internet and the institutional http://www.bta.org.bw/tenders/2008/Spectrum%20public%20announ
environment Telecommunications Policy, 2006. 30(1): p. 29-45. cement%20(3).pdf.
[9] Van Gorp, A.F., Increasing Regulatory Capacity: The Role of the [33] BotswanaGovernmentGazette, Telecommunications (Amendment)
Region in Shaping National ICT Policy in Southern Africa. PhD Bill, 2004. Bill No. 15 of 2004. 2004: Gaborone, Botswana. p. 3.
Dissertation, the College of Information Sciences & Technology, the [34] McCormick, P.K., Telecommunications Reform in Botswana: A Policy
Pennsylvania State University, University Park, PA, USA. 2008, p. Model for African States. Telecommunications Policy, 2001. 25(6): p.
207. 409-420.
[10] Melody, W.H., On the meaning and importance of 'independence' in [35] Kaufmann, D., A. Kraay, and M. Mastruzzi, Governance Matters VII:
telecom reform. Telecommunications Policy, 1997. 21(3): p. 195-199. Aggregate and Individual Governance Indicators 1996-2007, in
[11] Bauer, J.M., The coexistence of regulation, state ownership and Policy Research Working Paper 4654. 2008, The World Bank
competition in infrastructure industries: Quagmire or Institutional Development Research Group Macroeconomics and Growth Team &
Innovation?, in Quello Center of Telecommunication Management World Bank Institute Global Governance Program: Washington, D.C.
and Law Working Paper Series. 2003, Michigan State University: East
Lansing, Michigan.
[12] Wallsten, S.J., An Econometric Analysis of Telecom Competition,
Privatization, and Regulation in Africa and Latin America. The
Journal of Industrial Economics, 2001. 49(1): p. 1-19.
[13] Balasooriya, A., Q. Alam, and K. Coghill, The effectiveness of the
telecommunications regulatory regime: The case of Sri Lanka
Telecom. Public Administration and Development, 2006. 26: p.
383-393.
[14] Samarajiva, R., The Role of Competition in Institutional Reform of
Telecommunications: Lessons from Sri Lanka. Telecommunications
Policy, 2000. 24(8-9): p. 699-717.
[15] Henisz, W.J. and B.A. Zelner, The institutional environment for
telecommunications investment. Journal of Economics and
Management Strategy, 2001. 10(1): p. 123-147.
[16] Henisz, W.J., The institutional environment for infrastructure
investment. Industrial and Corporate Change, 2002. 11(2): p. 355-389.
[17] Gutierrez, L.H. and S. Berg, Telecommunications Liberalization and
Regulatory Governance: Lessons from Latin America.
Telecommunications Policy, 2000. 24(10-11): p. 865-884.
[18] Jain, R., Interconnection regulation in India: Lessons for developing
countries. Telecommunications Policy, 2006. 30: p. 183-200.
[19] Samarajiva, R., Regulating in an Imperfect World: Building
Independence through Legitimacy. Info, 2001. 3(5): p. 363-368.
[20] Gutierrez, L.H., Regulatory Governance in the Latin American
Telecommunications Sector. Utilities Policy, 2003. 11: p. 225-240.
[21] Yin, R.K., Case Study Research: Design and Methods. Third ed. 2003,
Thousand Oaks, CA: Sage Publications.
[22] ITU, Effective Regulation Case Study: Botswana 2001. 2001: Geneva,
Switzerland. p. 48.
[23] Van Gorp, A.F. and C.F. Maitland, Regulatory Innovations in
Tanzania: The Role of Administrative Capabilities and Regulatory
Governance. Info: The Journal of Policy, Regulation and Strategy for
Telecommunications, Information and Media, 2009. 11(1): p. 64-77.
[24] Moshiro, S., Licensing in the Era of Liberalization and Convergence.
2005, ITU.
[25] Mwaikusa, J.T., The Limits of Judicial Enterprise: Judicial Powers in
the Process of Political Change in Tanzania. Journal of African Law,
1996. 40(2): p. 243-255.
[26] Gloppen, S., The Accountability Function of the Courts in Tanzania
and Zambia. Democratization, 2003. 10(4): p. 112-136.
[27] Lawson, A. and L. Rakner, Understanding Patterns of Accountability
in Tanzania. 2005, Oxford Policy Management, Chr. Michelsen
Institute and Research on Poverty Alleviation (REPOA): Oxford, UK.
[28] BTA, Annual Report 2007. 2007, BTA: Gaborone, Botswana.
[29] Pheko, T.G., Speech by the Chief Executive, Mr T.G. Pheko, at the
Award Ceremony of the Service Neutral License to Mascom Wireless
437
Abstract – It is estimated that there are 100 million “street square kilometer [4]. About 4 million live in one of about
children” in the world [1]. Street children typically live 5000 slum communities, and 42% earn between US$132 and
independent of families in makeshift living arrangements, and
$372 dollars per year. These slum communities have an
survive on little but their wits and the camaraderie of small
overall literacy rate of 28.5%, and only 3.9% of males and
gangs.
1.3% of females have graduated from secondary school [5].
To better understand the lives of street children, we conducted Few have access to opportunities, social networks, or
150 days of ethnographic investigations in and around marketable skills that would provide them with a steady
Calcutta, with an emphasis on street children who live in train stream of reasonable income; the majority find irregular
stations. Our hope was to understand their typical challenges, income streams from the informal service industry or small-
and to see whether they could be addressed with information scale industries such as tailoring, carpentry, blacksmithing, or
and communication technology. Among other things, this manufacturing [6]. Indian street children are routinely
study led to the identification of an unreported knowledge gap detained illegally, beaten, tortured and sometimes even killed
among the children. Namely because their world is limited to by police, based on the general population’s mistrust of street
geographical “islands” within a couple-block radius of train children, and the corruption of the police force [7] [8] [9].
stations, they were entirely unaware of the of micro-
In this paper, we describe the three phases of research we
entrepreneurial possibilities beyond their islands.
conducted to provide a technology-based system that provides
In the second half of this paper, we describe a preliminary trial older street children with information about available income-
with an NGO to expose street children to entrepreneurial generating opportunities: First, we conducted an ethnographic
possibilities using mediated video instruction. Such a system exploration of street children in and around Calcutta. Second,
shows promise, if combined with ongoing social support: we tried a couple of short-term interventions to understand the
some children were both quick to understand and try their impact they would have on youths who live in train stations.
hand at new professions, and were to continue their These early investigations led to the novel understanding that
professions for at least two months with ongoing NGO street children have an extremely limited view of the job
support. opportunities available to them, due to a kind of tunnel vision
about the world around them. Third, we then implemented and
observed a preliminary trial of a new system called Social
I. INTRODUCTION
Enterprises, which informs street children of a diverse set of
[5] M. Huenerfauth. 2002. Developing Design Recommendations for [24] A. Chavan. 2007. Around the World with 14 Methods.
Computer Interfaces Accessible to Illiterate Users. Thesis. Master of http:// humanfactors.com/downloads/whitepapers.asp#CIwhitepaper.
Accessed on August 22, 2008.
Science (MSc). Department of Computer Science. National University
of Ireland: University College Dublin. [25] A. I. Rudnicky. Mode Preference in a Simple Data-Retrieval Task.
[6] S. Deo, D. Nichols, S. Cunningham, I. Witten, 2004. Digital Library Proceedings of the ARPA Workshop on Human Language Technology.
Access For Illiterate Users. Proc. 2004 International Research San Mateo: Morgan Kaufmann, 1993, 364-369.
Conference on Innovations in Information Technology
[26] W. Ong. Orality and Literacy: The Technologizing of the Word. London:
[7] I. Medhi, A. Sagar, K. Toyama. Text-Free User Interfaces for Illiterate
Routledge, 2002.
and Semi-Literate Users. Proc. International Conference on Information
and Communications Technologies and Development, 2006. [27] J. Sherwani, N. Ali, R. Rosenfeld. Orality-grounded HCID:
[8] S. Grisedale, M. Graves, A. Grunsteidl, 1997. Designing a Graphical Understanding the Oral User. Submitted to the Information Technology
User Interface for Healthcare Workers in Rural India, ACM CHI 1997 and International Development journal.
[9] V. Anantaraman, et al. Handheld computers for rural healthcare, [28] D. Jurafsky, J. H. Martin. Speech and Language Processing. Prentice
experiences in a large scale implementation. In Proceedings of Hall, 2008.
Development By Design, 2002.
[29] J. Sherwani, S. Tomko, R. Rosenfeld. Sublime: A Speech- and
[10] T. Parikh, G. Kaushik, and A. Chavan, Design studies for a financial Language-based Information Management Environment. In Proc. IEEE
management system for micro-credit groups in rural India. Proc. of the Int.l Conference on Acoustics, Speech and Signal Processing, Toulouse,
ACM Conference on Universal Usability, ACM Press (2003). France, May 2006.
458
Abstract—Nearly 40 million people in Africa suffer from Despite the subsidies offered by pharmaceutical companies
HIV/AIDS. African governments and international aid agencies for expensive AIDS drugs, and the enormous effort of African
have been working to combat this epidemic by vigorously promot- governments to promote HAART programs, these initiatives
ing Highly Active Anti-Retroviral Therapy (HAART) programs.
Despite the enormous subsidies offered by governments along have not been adopted on a wide-scale. The fundamental
with free Anti-RetroViral (ARV) drugs supplied by agencies, barriers to large-scale adoption are two-fold. The first problem
the introduction and implementation of HAART programs on is one of lack of accountability in the system, due to the
a large scale has been limited by two fundamental problems: theft of expensive ARV drugs, counterfeiting of drugs, and
(a) lack of adherence to the ARV therapy regimen; (b) lack of corruption. The Global Fund crisis that gripped much of
accountability in drug distribution due to theft, corruption and
counterfeit medication. Africa, with losses amounting to hundreds of millions of
In this paper, we motivate the case for SmartTrack, a telehealth dollars was directly related to the lack of accountability in
project which aims to address these two problems facing HAART the system. In fact, many pharmaceutical companies that
programs. The goal of SmartTrack is to create a highly reliable, provide subsidized AIDS medications are demanding better
secure and ultra low-cost cellphone-based distributed drug in- accountability practices as a prerequisite to their continued
formation system that can be used for tracking the flow and
consumption of ARV drugs in HAART programs. In this paper,
participation.
we assess the potential benefit of SmartTrack using a detailed The second serious problem relates to the lack of patient
needs-assessment study performed in Ghana, using interviews adherence to the medication regimen. A typical HAART pro-
with 516 HIV-positive rural patients in a number of locations gram requires every patient to consume two to five medications
across the country. We find that a system like SmartTrack would per day; in addition, required medication and dosages may
immensely benefit both patients and healthcare providers, and
can ultimately lead to improved patient outcomes and better
change rapidly based on the side effects observed by the
accountability. patient. Lack of medical oversight, or inappropriate use of
Index Terms—SmartTrack, telehealth, drug tracking, drug ARVs not only harms the health of patients but may also
monitoring, patient adherence encourage drug-resistant strains of HIV, posing a substantial
public health risk. Hence, to improve patient outcomes, it is
I. I NTRODUCTION essential for doctors to continuously observe the health status
and the medication consumption regimen of patients. In fact,
According to the World Health Organization, effective it is relatively common for doctors and health-workers to
HIV/AIDS care requires antiretroviral therapy (ART) [1] as physically track patients who have not reported for regular
a treatment method. Without access to antiretroviral therapy, medical visits.
people living with HIV/AIDS cannot attain the fullest possible In this paper, we motivate the case for SmartTrack, a project
physical and mental health and cannot play their fullest role as which aims to address these two fundamental problems with
actors in the fight against the epidemic, because their life ex- HAART programs. The vision of the SmartTrack project is to
pectancy will be too short. While ART is commonplace in de- create a highly reliable, widely available and ultra low-cost
veloped countries, these life saving medications have reached cellphone-based distributed drug information system that can
only a small percentage of the more than 40 million Africans be used for tracking the flow and consumption of ARV drugs
infected with HIV or suffering from AIDS [2]. In order to be in HAART programs, ultimately leading to improved patient
successful in combating this deadly disease, strict adherence outcomes.
to a highly active antiretroviral therapy (HAART) [3] program The motivation to use cellphones in SmartTrack to track the
must be followed. flow of AIDS drugs is in large part driven by the explosive
459
growth in the adoption of low-cost cellphones in the rural II. M OTIVATION AND R ELATED EFFORTS
developing world. For example, around 80% of Rwanda has The SmartTrack project vision is centered on three primary
cellphone coverage [4], [5] and the cost of cellphones is so premises: (a) the need for better accountability in ARV drug
low (less than $50) that the devices have become affordable distribution; (b) the need for improved patient adherence to
by the poor [6]. Cellphones have enormous potential for medication regimen in HAART programs; (c) the utility of
enhancing rural healthcare, given their ability to act as a cellphones as a healthcare platform for monitoring the flow
low-cost computing platform for distributed applications. The and consumption of drugs. In this section, we discuss these
open-source movement in cellphone software [7], [8], [9] three aspects in more detail, along with the corresponding
has also opened up the potential for a wide range of new related efforts.
applications targeting these low-cost devices.
The vision of SmartTrack is to enable patients, health- A. Maintaining Accountability in the Supply Chain
workers and doctors to use cellphones to record, track and The first premise of SmartTrack builds on the position of
transmit information about pharmacotherapy utilization. The the World Health Organization (WHO), which has argued for
idea is to tag every ARV drug bottle with a “smart” tag: an the need for a drug tracking system that can monitor the
RFID or a barcode, that will enable patients to identify drugs flow of medications from the supplier to the patient. The
easily and also enable them to report their drug consumption high cost of ARV drugs makes them an attractive target for
remotely using a cellphone. The drug information will be theft. This is a particular threat in developing countries where
compiled on the device, sent to a central server, and stored corruption can be widespread and persistent. International
at different levels (regional level, district level) to be used HAART programs are not immune to theft, as was illustrated
by the healthcare teams to tailor patient-specific therapies. In by the Global Fund scandal in Uganda in 2006; in that
addition, the device will also be able to remind patient end- incident, an investigation revealed that ”tens of millions” of
users about when to take their specific medications. Thus, dollars in grants intended for AIDS treatment were misspent
SmartTrack will provide the patient with information about by officials, often to finance lavish lifestyles [11]. Similar
his or her specific regimen, and provide the care provider with problems have also been reported in other recipient nations
information about patient compliance to the regimen. [11], and the widespread theft of medicine for resale has been
SmartTrack is based upon eMedonline1 , a technology pre- exposed in a number of instances [12], [13].
viously developed and patented by our partner, Leap of Faith Perhaps more disturbing than outright theft is the growing
Technologies, Inc. [10]. This technology leverages cellphones practice of replacing legitimate medication with counterfeit
and RFID technology to optimize medication compliance, drugs. While it is difficult, if not impossible, to determine the
track medication usage and extend patient care to remote actual scope of this problem, WHO has estimated that in parts
settings. The previous project has been successfully used in the of Africa, Asia and Latin America, over 30% of all drugs sold
US for tracking adverse events and improving compliance with are counterfeit [14]. Consuming counterfeit medication can
oral antineoplastic (anti-cancer) therapy; efficacy is currently lead to drug resistance and death, and many counterfeits have
being demonstrated among cardiovascular disease populations. been found to contain highly toxic substances.
A key requirement for the success of the SmartTrack is In addition to the serious health consequences that can
to first assess needs and user acceptability. In this paper, we result from fake or stolen HAART drugs, the phenomenon
describe our experiences in realizing this first important step could also lead to a loss of confidence in the health system
in the SmartTrack project vision. We recently performed a as a whole. This threatens to undermine support for aid
detailed needs-assessment study, interacting with 516 AIDS programs, as well as discourage patients from seeking out
patients in rural parts of Ghana to determine the potential and adhering to treatment. A supply-chain management system
role and effectiveness of cellphones in improving AIDS care that could track medicines from their acquisition point to the
delivery. To this end, we developed simple cellphone-based end user would be able to detect drug theft quickly, and
healthcare applications which we used as part of our study to would make it easier to discover sources of counterfeit drug
understand user acceptability issues. Our study was conducted substitution within the supply-chain. This added accountability
by medical students who interacted with each patient for over could greatly improve the safety and effectiveness of HAART
30 minutes and used a translator (when language was an issue) programs in developing countries.
to obtain answers to a detailed survey in combination with
B. Tracking Patient Adherence and Symptoms
user interface testing. Based on our detailed patient study
evaluations, we determined that a system like SmartTrack Proper adherence to a HAART regimen is critical to prevent
would immensely benefit both patients, healthworkers and drug resistance and ensure survival [15]. However, HAART
doctors and can ultimately lead to improved patient outcomes treatment requires multiple drugs, taken multiple times per
and better accountability. day, and often results in unpleasant side effects. These dif-
ficulties, combined with the high cost of the drugs and the
1 developed by Leap of Faith Inc., with support from the National Cancer
potential stigma of being identified as HIV+ can act as a
Institute (Contract No.HHSN261200644005C) and the National Institute on deterrent to proper adherence. A number of studies have
Aging (Grant No.2 R44 AG022271-02A1) attempted to measure adherence rates in Sub-Saharan Africa,
460
and they have produced promising numbers, with an average in healthcare environments in Ghana, Uganda, and Kenya.
of 77% of participants reporting proper and consistent use of The project put PDAs into the hands of physicians, medical
the medication [16]. However, these studies rely on unverified students and community volunteers in different settings in
self-reporting, which has been demonstrated to exaggerate order to demonstrate their viability and usefulness, especially
the level of compliance in studies conducted in developed for the collection of health data and dissemination of medical
countries. information.
Tracking adherence is especially difficult in developing Bridges.org, a technology NGO, performed an independent
regions, where lack of medical, transportation, and commu- evaluation of the Satellife PDA trial [28], in which it validated
nication infrastructure can limit contact between patients and the use of handheld computers in healthcare environments in
health-care workers [17]. For example, our study of HAART Africa. Specifically, it found handheld computers to be an
patients in Ghana found that more than 37% of participants appropriate technology for use in the African context, and
went at least three months between clinic visits, that over 94% concluded that they provide an inexpensive alternative to PCs
had never placed a phone call to their clinic, and that the same in terms of computer power per dollar. The technology of
number had never received a phone call from their clinic. This Satellife was found to be simple to use and easily integrated
lack of contact can make it very difficult for health workers into the daily routines of the healthcare professionals.
to keep tabs on patient adherence and progress. Likewise, it Cell-Life [29], [30], a research team at the University
can prevent patients from notifying health workers of critical of Cape Town, has demonstrated the effectiveness of an
symptom information that might require a trip to the clinic or information gathering system using cellphones for improving
a change in treatment. the health outcomes in ART. Their system leverages health
A low-cost method of tracking medication, adherence, and workers to collect real-time information on HIV/AIDS along
symptoms in developing countries could provide great benefit with the spatial infrastructural requirements (person, their
to its users. By preventing fraud, it could save precious finan- environment, and access to basic amenities). Cell-Life has
cial resources for providers, and ensure that medicine gets to been deployed in South Africa and Zambia.
those who most desperately need it. By tracking adherence and
symptoms, it could improve the responsiveness of healthcare III. S MART T RACK
workers to the needs of patients, and ultimately improve and
The SmartTrack project was in part motivated by prior suc-
prolong the lives of those receiving HAART treatment.
cessful research and development by our partner, Leap of Faith
C. Cellphones as a Healthcare Platform Inc., of a cellphone-based telemonitoring system for improving
medication adherence. This system, called eMedonline2 uses
In the past few years, several research and developmental
smartphones equipped with RFID scanners to read smart-tags
efforts around the world have explored [18], [19], [20], [21] the
attached medicine bottles. The phone provides reminders to the
use of cellphones as a potential tool for improving healthcare
patient, and the patient scans each bottle with the phone when
in both the developed and the developing world. For a detailed
taking that particular medication, updating a remote database
overview of globalization and health related issues, please
and providing doctors with detailed information on regimen
refer to [22]. In developed nations, the problem of tracking
adherence. The system validated the integration of cellphones
medication compliance has been addressed successfully with
and radiofrequency identification (RFID) as a therapeutic solu-
smartphone-based telehealth tools [23]. Recent growth in cell-
tion to medication compliance and supply chain management.
phone penetration in developing regions has made mobile tele-
Feasibility and functional tests of the system in a sample
health solutions a real possibility for users in less-developed
of oncology patients demonstrated that drug and compliance-
nations. In Sub-Saharan Africa in 2006, the overall cellphone
related data can be reliably collected, analyzed, and exported
penetration level was estimated at 15%, and it has more than
for use in other clinical monitoring systems. Patient acceptance
doubled in the last two years [4], [5]. Furthermore, certain
and value of the system was very high.
countries, such as South Africa, have attained adoption rates
of over 70% [24]. In our own study, conducted among HAART Building on this prior success, the SmartTrack project vision
recipients in Ghana, we found that cellphone usage was quite aims to develop and deploy a cellphone-based telemedicine
high in the cities and relatively large even in rural areas: and supply-chain sytem that exploits this trend, with the
Nationwide, 54% of participants reported using cellphones, twin goals of tracking the flow of drugs to the field, and of
with more urbanized areas such as the Greater Accra (77%), monitoring the status and adherence of HAART patients. In
Northern (67%) and Volta (57%) regions ranking higher in addition to tracking adherence, we also envision a system that
usage than rural areas, such as the Central region (29%), and acts as a reminder system, that allows users to record symptom
Upper West (26%). information which can be tracked and analyzed by their doctor,
Several groups have examined the use of PDAs and cell and that allows health workers to contact patients. There are,
phone-based tools to augment patient monitoring relative however a number of limitations on the ground that prevent
to ARV therapy, and to extend medical follow-up capabili- 2 developed by Leap of Faith Inc., with support from the National Cancer
ties [17], [25], [26], [27], [18], [19], [20], [21]. One important Institute (Contract No.HHSN261200644005C), and with support from the
project was led by Satellife [25], which tested the use of PDAs National Institute on Aging (Grant No.2 R44 AG022271-02A1)
461
SmartTrack architecture, we need to address several technical AIDS patients. This system used a voice and text interface
research challenges to make SmartTrack into a low-cost, to prompt users to respond to various questions. In addition,
secure, highly reliable and widely available system. We briefly the system used RFID tag identification which we did not
outline the basic technical ideas that we intend to use in use in our study. We used a HTC P6300 phone for this
SmartTrack to achieve these properties. implementation.
1) Information aggregation: Using an inbuilt network cost Cellphone based system: As a comparison point, we
model, SmartTrack will intelligently aggregate multiple developed an alternative implementation using JavaME [31]
updates from a cell-phone and semantically compress on Nokia 3110 phones that used a text-free pictogram based
(not standard compression) to the smallest number of user interface of specific healthcare symbols to primarily un-
messages possible to reduce transmission costs. derstand whether the end-user population would be conversant
2) Strong identity and secure updates: To provide strong with a pictogram-based interface.
security properties, the system will leverage public-key A. Methodology
and symmetric-key cryptographic operations to develop
an unforgable identity for every cellphone and also The user study for SmartTrack was performed at several
provide the ability to verify the validity of individual sites across Ghana. To perform the study, we developed a
updates using aggregate signatures. survey tool that was administered to cohorts of HIV+ patients
3) Privacy: To provide privacy, the system will develop an in each location. The cohorts were patients affiliated with
anonymized indexing system that cannot be traced back either the West Africa AIDS Foundation (WAAF) [32] or
to individual patients without appropriate information. Korle Bu Teaching Hospital. Participants were referred to us
4) Reliability and Availability: the system increases avail- by these organizations, and we did not attempt to implement a
ability without incurring additional operational cost by randomized study. The ages reported by patients ranged from
maintaining a detailed local data-store within every 17 to 101, and 73.3% of the patients were female, 26.7% male.
cellphone that contains most of the appropriate infor- 77.3% of patients were currently receiving ARV therapy.
mation required by the end-user (health-worker, doctor We interviewed 516 HIV+ patients with a survey consisting
or patient). This local-store is constantly synchronized of 72 questions divided into eight categories:
with the main information store either using aggregate 1) Laboratory values: CD4 count, viral load and WHO HIV
updates or physical synchronization. To improve reliabil- stage (taken from blood tests/patient records).
ity, SmartTrack will intelligently replicate and partition 2) Demographics: Age, housing, occupation, education,
critical information in the data store between the local language skills, access to water and electricity (11
server and the smart-phones. questions).
While these technical issues are certainly important, a 3) Health Care Access: Health practices, frequency of
detailed discussion of them is outside the scope of this paper. medical care, frequency of contact with health worker,
The primary focus of this paper is to make the case for why distance to the hospital (9 questions).
rural users need a system like SmartTrack, and, if provided, 4) Symptoms: HIV-related symptoms, typical response to
will they actually use such a system. The rest of the paper certain symptoms, and histories of other conditions like
will focus on answering these user-centric questions. tuberculosis or malaria (11 questions).
5) Medications: Medications being taken, adherence to
IV. N EEDS -A SSESSMENT U SER S TUDY regimen, acquisition of drugs (10 questions).
When working on any “technology for development” 6) Narrative Text: Difficulties of the disease, difficulties
project, two fundamental questions always arise: the need for of treatment, perceived effectiveness of treatment (9
the technology and user acceptability of the technology. Many questions).
projects in the development space have been successful in the 7) Mobile Phone Usage: Access to mobile phone, usage
beginning phase but have often not been adopted on a wide habits, amounts paid for service and hardware (20 ques-
scale. It is this barrier that we wish to cross in SmartTrack. tions).
Hence, we began with the endeavor of performing a detailed 8) Patient Education: What is the patient’s primary source
needs-assessment study to answer the two questions about of health information (1 question)?
need and user acceptance: (a) Do rural users need a system like The patients gender and use of a translator were not asked
SmartTrack? (b) If provided with a system like SmartTrack, directly, but were recorded. When necessary the aid of a
will they use it? Rather, how would one build SmartTrack to translator was enlisted.
maximize user acceptance. 278 patients were interviewed via healthcare associations
To address these questions, we began with the modest for patients living with HIV, coordinated through WAAF.
goal of designing a detailed survey covering a wide-range of Researchers traveled to five cities in four different regions of
questions. We also designed two simple user-interfaces as part Ghana to interview a total of six cohorts of patients: Volta
of this study: Region (Ho), Central Region (Cape Coast), Northern Region
Smartphone based system: We tailored our smartphone (Tamale), Upper West Region (Wa and Lawra). Blood draws
based telehealth system for cancer patients into a system for to be used for CD4 counts were provided free-of-charge by
463
percentage
were required to give written consent. The written consent
40
forms were collected by Korle Bu Hospital staff to be included
in the patients chart, and will not be published or disclosed in
this or any other report to maintain patient confidentiality. The 20
R
not yet started ARV therapy. Only patients who were currently
rb
ur
an
al
Translator required
taking ARV therapy were interviewed as their experiences are
more relevant to the study.
Adjustments were made to the survey and methods during
Fig. 2. Percentage of patients requiring translators, with urban compared to
the course of the study. The survey itself was edited after the rural
first two administration sessions in order to more effectively
elicit the target information. Several questions were altered,
discarded, or added. The changes have been recorded and will
be taken into account. The overwhelming majority of patients 100
Education percent of patients
were interviewed individually. In certain situations however,
there were time constraints and patients were interviewed in 80
A-
JS
SS
U
o
ni
-le
h
le
ve
sc
Fo
ve
ve
ho
rs
rm
ity
o
developing world, and is particularly severe in rural areas. We l
Education level
gave participants a number of questions designed to gauge
the level of literacy and education within our population.
Questions we asked included: i) what languages does the
participant speak? ii) in which languages is the participant able Fig. 3. Maximum education level
to read or write? iii) what is the participant’s highest level of
education?
We found that 55.6% of interviewees in urban areas, and
60% in rural areas needed a translator for our interviews. solution might solve the literacy problem, though the voice-
Interviews were conducted in English, which is also the official based UI would not solve the multiple language problem, and
language of Ghana. Additionally, our survey showed that urban would have the additional drawbacks of lack-of-privacy, in
areas tend to have higher literacy rates than rural regions. addition to feasibility and cost problems when using low-cost
Figure 3 shows the distribution of educational attainment phones with expensive network access.
among patients, and Figure 4 shows the percentages of patients To this end, we are exploring pictogram interfaces for
able to read or write in each of the locally-used languages. this application. We developed a demonstration interface for
Notably, at least 35% of patients had never attended school of our Ghanaian study in order to test our ability to convey
any kind, and only 46% reported being able to read English. healthcare-related concepts through graphical symbols. Al-
This high rate of illiteracy, particularly when combined with though extremely preliminary, this demonstration was well-
the lack of a common language among patients, presents a received by users and showed promise for expressing health
fundamental challenge to the deployment and acceptance of concepts. Using simple pictograms, combined with shape or
any telemedicine application. In order for SmartTrack to be color symbols to represent different medications, it should be
effective, it must be operable by a patient with no reading possible to make an effective interface that is usable despite
skills whatsoever. Either a voice-based or a pictogram-based barriers of literacy and language.
464
100 100
Languages known (read or write) percent of patients percentage of ages of patients
80
80
60
percentage
60
percentage
40
40
20
20
0
Ar
En
En
En
En
Ew
Fa
Fr
Tw
on
a
en
ab
nt
gl
gl
gl
gl
i
e
e
is
is
is
is
ch
ic
h,
h,
h,
Ar
Fa
Tw
ab
nt
i
i
ic
20
30
41
51
61
-2
-4
-5
-6
-7
9
3
Languages Age
Fig. 4. Percentage of patients able to read or write, by language Fig. 5. Age distribution of participants
B. Age and Sex To better understand the economic situation of our popula-
A patient’s age is an important factor in determining proper tion, we asked a number of questions pertaining to their basic
medical care, and age statistics can offer some insight into the employment, income, and living conditions. The following are
impact of AIDS on the society at large. Figure 5 shows the age some selected findings:
distribution of the patients, and illustrates the disproportionate • Type of house: 58.3% lived in a compound house, 24.4%
number of people seeking treatment who are in the 30 to 50 in a self-contained house and 16.3% in a rented house.
year-old range, with over 69% of patients falling within that The rest were either homeless or lived in mud huts.
group. For so many people in the prime of their work and • Water source: 75.6% used piped water and 15% used
family lives to be stricken can only have devastating effects on well-water.
a society. The lost productivity of working-aged people slows • Electricity: 26% did not have reliable access to electricity.
the economic development of an already-depressed region, and • Refrigeration: 56% lacked any sort of refrigeration.
the death or debilitation of so many parents will have social • Occupation: The most common job categories reported by
ramifications we have barely begun to understand [33]. patients were petty trading(34.7%), service work(23.1%)
Improving adherence can prolong lives and improve the and agriculture(12.6%).
quality of those lives. In addition to the direct and obvious To live in such a setting can be challenging enough for
benefit to the patient, this can also have the wider benefit of a healthy person, but for an HIV+ person, the difficulty is
allowing the patient to be more productive economically, and multiplied. People with AIDS can experience weakness and
to play a greater role in the lives of family and the community. any number of other symptoms that interfere with their ability
Interestingly, 73% of patients were female. While it is to work and earn money. Of our study participants, 64%
estimated that almost 61% of HIV+ people in Sub-Saharan responded that AIDS has a considerable effect on their ability
Africa are female[34], it is not clear if our higher number to work, and 60% reported missing work as a result of HIV
reflects the actual proportion of women with HIV in Ghana, symptoms.
or if it is a result of our non-random selection process. It is Social stigma for those with HIV has been reported to be a
even conceivable to us that it is a result of the self-selection serious issue in many cultures, particularly in parts of Africa,
process of people seeking HIV treatment or getting diagnosis. but interestingly, 61.5% of our participants said that they did
[34] not have any social problems due to AIDS. If true, this is
wonderful news, but we suspect the response does not tell the
C. Economic and social issues whole story. One possibility is that patients were hesitant to
It is impossible to divorce the problem of medical treatment admit that they had experienced such discrimination. Another
in the developing world from the economic environment in is that they had not experienced discrimination because they
which the patients live. The cost of care is only the most ob- had successfully hidden the fact that they had HIV.
vious element. The patient also must contend with the expense All of these details bring home a key point: Any designer of
of travel to distant clinics, and the cost of missed work during a telehealth system for the developing world must be keenly
time spend traveling. The frequent lack of electricity, clean aware of the tenuous existence its users may lead. Costs must
water, and safe, sanitary conditions makes staying healthy even be kept to a bare minimum, as the pennies charged even
more difficult. to send simple SMS messages may accumulate to become a
465
diseases which may be common to people suffering from it was much more expensive than SMS. Given that the
AIDS, and the percentages of patients who reported having information per patient is extremely limited, we felt that
these conditions at any point. The second column contains the SMS is a much cheaper alternative than voice or data
proportion of patients who traveled to see a doctor as a result services.
of the corresponding condition. These trips can be quite diffi- 3) The system can potentially be extended to other chronic
cult for patients, and might be avoided in certain cases if the illnesses like malaria, TB and Cholera.
doctor can consult with the patient remotely and recommend 4) It is critical to perform a needs-assessment study before
a course of action. Again, a cellphone-based health system performing the actual design. While this may seem
could enable doctors to track patient symptoms remotely and obvious in retrospect, doing a detailed needs assessment
act appropriately, contacting the patient if necessary. study is a tedious and time consuming process that is
often ignored.
VI. U SER R EACTIONS AND L ESSONS L EARNED
When we experimented with two different cellphone plat- VII. C ONCLUSIONS
forms with each user, the reactions were fairly varied across As Bill Easterly wrote in The White Man’s Burden, “2.3
users. Many users were completely new to the concept of trillion dollars of aid over 50 years and we have nothing much
smartphones and had difficulties navigating the screen. This to show for results.” [35] This so poignantly summarizes the
is not to rule out smartphones completely from such an current state of affairs of many philanthropic organizations
environment, but it is well known that rural populations are funding AIDS relief efforts in Africa. While it is indeed true
not accustomed to sudden changes. The cellphone has been that some organizations have made significant progress in
considered a user friendly device due to its similarity with specific areas, this is more of an exception than the norm.
the telephone. While many users found the voice option fairly Our effort is largely centered around technological issues
attractive, our voice interface was based on English which facing the adoption of AIDS relief schemes. We believe that,
was not understandable to over 50% of the users. The users a combination of technological and sociological approaches
we interviewed spoke several languages/dialects that made it are needed to tackle the healthcare(HIV/AIDS) problem in
hard for translators we used in our study to converse with developing regions.
local Ghanaians in rural areas. Another concern that patients From a technical perspective, to make AIDS relief efforts
expressed with voice was that a “talking” phone could disclose more effective, it is essential to improve accountability in the
the HIV status of a patient. health system, and to enhance patient adherence to medication
In contrast, users found pictograms very easy to understand. regimen through constant monitoring. The vision of Smart-
While we used only a limited set of universal healthcare Track is to address these two problems, using a distributed
symbols as pictograms in our study, it remains to be seen as to cellphone-based platform to achieve these goals. Though this
how many different types of questions can be phrased purely project is still in a relative state of infancy, it has undertaken
using pictograms. For example, to represent different drugs, a detailed needs-assessment study using interviews with over
we found that simple color and symbol coding would work 500 AIDS patients in Ghana. Our next step in future work
sufficiently well for several illiterate users. Overall, we felt is the actual deployment of SmartTrack with the aid of the
that a pictogram based user interface holds much promise and West Africa AIDS Foundation and Korle Bu Hospital. We
can also deal with the stigma issues that voice based interfaces plan to use the lessons learned from the needs-assessment
sometimes might present. study, with particular attention to the user interface issues,
The overall response to a system like SmartTrack was to deploy a widely acceptable version within Ghana to non-
very positive. The vast majority of patients (91%) seemed English-speaking and non-literate populations. If successful,
extremely positive and interested in a reminder system that this effort can be replicated in other rural developing parts of
could be placed on their cellphones. Typically, patients have the world.
to travel great distances to reach their treatment facilities;
ACKNOWLEDGMENT
hence, being able to remotely communicate with the doctor
is a huge positive factor for most patients. Social workers, We would like to thank Korlebu Hospital and the West
nurses, community health workers, and doctors seemed very Africa AIDS Foundation for enabling us to do the detailed user
positive and amenable to the system. Many health-workers study and Microsoft Corporation for supporting this project.
also wanted a similar system for other common diseases like R EFERENCES
malaria, TB and Cholera.
[1] “World Health Organization,” http://www.who.int/3by5/en/.
Other important lessons we learned include: [2] “HIV and AIDS in Africa,” http://www.avert.org/aafrica.htm.
1) From a cost perspective, smartphones are much more [3] “Treatment of HIV infection,” http://www.niaid.nih.gov/
expensive than cellphones. In Ghana, users can purchase factsheets/treat-hiv.htm.
unlocked cellphones at extremely low cost, thus we felt [4] “Upwardly mobile in Africa-Business Week,” http://www.
businessweek.com/globalbiz/content/sep2007/gb20070913
the system should be built around people’s cellphones. 705733.htm.
2) From a communication perspective, many rural areas we [5] “Africa Telecom News,” http://whiteafrican.com/2008/08/01/
surveyed did not have GPRS coverage. Even if available, 2007-african-mobile-phone-statistics.
467
[6] “Low cost cellphones,” http://www.cameraphonesplaza.com/ [23] B. Rapchak, “Cell phone technology to enhance medication
new-low-cost-nokia-cell-phones/. adherence and therapeutic outcomes,” in TEPR 2009, Medical
[7] “Open Handset Alliance,” http://www.openhandsetalliance.com. Records Annual Meeting.
[8] “Symbian OS: the open mobile operating system,” http://www. [24] K. Kalba, “The adoption of mobile phones in emerging markets:
symbian.com/. Global diffusion and the rural challenge,” International Journal
[9] “Android - An Open Handset Alliance Project,” http://code. of Communication, vol. 2, Jun. 23 2008. [Online]. Available:
google.com/android/. http://ijoc.org/ojs/index.php/ijoc/article/view/216
[10] “Leap of faith technologies, inc.” [Online]. Available: http: [25] “Satellife: The Global Health Information Network,” http://
//www.leapoffaith.com/ www.healthnet.org.
[11] R. Scheier, “African graft stings donors,” The Christian Science [26] B. DeRenzi, N. Lesh, T. Parikh, C. Sims, W. Maokla,
Monitor, Jun. 1 2006. M. Chemba, Y. Hamisi, D. S. hellenberg, M. Mitchell, and
[12] C. Mangwiro, “Mozambique links health officials to drug G. Borriello, “E-imci: improving pediatric health care in low-
thefts,” Reuters, Aug. 28 2007. income countries,” in CHI 2008. ACM, pp. 753–762.
[13] F. Phiri, “Drug theft poses threat to battle against TB,” Inter [27] I. Medhi, A. Sagar, and K. Toyama, “Text-free user interfaces
Press Service News Agency, Mar. 24 2004. for illiterate and semiliterate users,” in ITID ’07: Information
[14] “Counterfeit medicines fact sheet,” World Health Organization, Technologies and International Development, 2007, pp. 37–50.
Nov. 14 2006. [Online]. Available: http://www.who.int/ [28] “Evaluation of the Satellife PDA Project, 2002: Testing the use
[15] P. Sax, “Patient adherence and response to therapy,” in AIDS of handheld computers for healthcare in Ghana, Uganda, and
Clinical Care August 1, 1999. Kenya,” http://bridges.org.
[16] E. J. Mills, J. B. Nachega, I. Buchan, J. Orbinski, A. Attaran, [29] S. Anand and U. Rivett, “Ict in the management of hiv treat-
S. Singh, B. Rachlis, P. Wu, C. Cooper, L. Thabane, ment: Cell-life: a south african solution,” vol. 6, pp. 56–59, 2004.
K. Wilson, G. H. Guyatt, and D. R. Bangsberg, “Adherence [30] “Cell-Life,” http://www.cell-life.org/.
to Antiretroviral Therapy in Sub-Saharan Africa and North [31] “Java Micro Edition,” http://java.sun.com/javame/index.jsp.
America: A Meta-analysis,” JAMA, vol. 296, no. 6, pp. [32] “West Africa Aids Foundation,” http://www.waafweb.org/.
679–690, 2006. [Online]. Available: http://jama.ama-assn.org/ [33] C. Bell, S. Devarajan, and H. Gersbach, “The long-run
cgi/content/abstract/296/6/679 economic costs of AIDS : Theory and an application to
[17] D. Werner, C. Thurman, and J. Maxwel, Where there is no south africa,” The World Bank, Policy Research Working
doctor? Paper Series 3152, Oct. 2003. [Online]. Available: http:
[18] “OpenMRS,” http://www.openmrs.org. //ideas.repec.org/p/wbk/wbrwps/3152.html
[19] “Partner in Health,” http://www.pih.org. [34] “2007 AIDS epidemic update,” UNAIDS, Tech. Rep., Dec.
[20] “Dimagi,” http://www.dimagi.com. 2007. [Online]. Available: http://data.unaids.org/pub/EPISlides/
[21] “D-Tree,” http://www.d-tree.org. 2007/2007 epiupdate en.pdf
[22] W. Kaplan, “Can the ubiquitous power of mobile phones be [35] W. R. Easterly, The white mans burden : why the Wests efforts
used to improve health outcomes in developing countries?” in to aid the rest have done so much ill and so little good. New
Globalization and Health 2006. York: Penguin Press, 2006.
468
Fixed
Lines
per
100
telephony. This study investigates the usage of mobile phones 0.4
Inhabitants
in this immediate post-conflict setting. In particular, we adopt 0.3
the uses and gratifications approach to media research, giving 0.2
focus to both instrumental and intrinsic motivations for use. 0.1
Mobile phone users in both the capital city of Monrovia and 0
in various rural areas were surveyed using the Q methodology,
1989
1995
1998
2001
2007
1980
1983
1986
1992
2004
which identified distinct perspectives within these urban and
rural groups. Participants were then sorted into groups where
Year
each group contained users with similar perspectives on their
mobile phones. These identified groups included sets of users who
saw their phones as productivity enhancers, means of connectivity (a)
to family and friends, essential business tools, technological
curiosities, and sources of personal security. The idea of a phone Mobile Teledensity
as a stylish object was markedly rejected, especially in rural
areas. We contrast these Q-sort results from Liberia with previous 16
Subscribers
per
100
related to security. 8
I. I NTRODUCTION 4
Considerable attention has been given to the role of in- 0
1989
1995
1998
2001
2007
1980
1983
1986
1992
2004
formation and communication technologies as tools for de-
velopment within Africa, and increasing levels of excitement
Year
have concentrated on the use of mobile phones. With some
fanfare the Economist [1] announced that the “real digital
(b)
divide” was in terms of the differential access to mobile
telephones while computers and the Internet were of less use. Fig. 1. Fixed-line and mobile teledensity in Liberia from 1980–2007. Steep
Many writers have disagreed with their pessimistic assessment drops in fixed lines are evident in 1991 and 2003. Introduction of competing
mobile carriers in the mid 2000’s produced a soaring number of mobile
of computers and the Internet (e.g. [2]). Nonetheless, it is subscribers. Note the difference in vertical scale between the two charts.
clear that mobile telephones are playing a substantial and
important role in development within the global south. Indeed,
compelling evidence of the macro and microeconomic effect of
reports that average year-on-year growth rate for mobile phone
mobile phones in low-income countries has been mounting [3].
subscribers in Sub Saharan Africa from 1999-2004 was double
For instance Waverman et al. [4] find that mobile phones offer
what it was in Europe. Indeed Sub Saharan Africa is a
a significant macroeconomic growth dividend and one that is
continent driven by mobile telephony and in 2001 the total
“twice as large in developing countries compared to developed
number of mobile subscribers exceeded the number of fixed
countries”. Microeconomic benefit is also evident. For instance
line subscribers [6]. In 2004, the mobile teledensity across
Jensen [5] shows that mobile phone use by farmers in Southern
all of Africa was 9.1, with the vast majority, 87%, making
India increases productivity, enhances revenues, reduces waste,
use of prepaid cards. Considering only Sub Saharan Africa
and lowers consumer prices.
the mobile teledensity is best approximated at 6.2% [7]. And
A. Mobile phones in Africa while this number describes subscriber penetration it does not
Mobile phone penetration growth rates are today highest give an adequate sense of overall access and usage due to
in Africa compared to all other continents [6]. The ITU widespread sharing of phone subscriptions. Clearly, mobile
469
telephony is the central communication technology for much Furthermore, retrospective empirical scholarship has
of Sub Saharan Africa. demonstrated the critical nature of communication amongst
the people of a nation if there is to be a lasting peace instead
B. Mobile phones in Liberia of an all-too-frequent return to civil conflict [10]. Modern
Liberia, established as a state in 1847 by freed African information and communication technologies can therefore,
slaves from the U.S.A., is situated on the Atlantic coast of on their face, serve as tools in this process of national
West Africa with Sierra Leone, Guinea, and Côte d’Ivoire as reconciliation if they are ably applied to these communication
bordering countries. A relatively small country with approxi- activities. However, to understand what it would mean to
mately 3.3 million inhabitants, it is attempting to right itself “ably” apply modern ICT’s, including mobile telephony,
after decades of civil conflict. to the process of post-conflict development requires at a
Unrest has been a staple within Liberia for more than minimum an adequate understanding of the current uses and
15 years with two major civil wars (1989-1996 and 1999- meanings of mobile telephony in that environment. Such was
2003) in this time period. These years of conflict have seen the motivation for this study.
nearly one-third of the population displaced and taken the
lives of approximately 250,000 people. A peace was brokered II. U SES AND G RATIFICATIONS
and transitional government was established in 2003. A UN In seeking to uncover the everyday, micro-level motivations
peacekeeping mission was positioned to keep this peace, and for mobile phone use among Liberians, this study draws
democratic elections were held in the fall of 2005. This inspiration from the uses and gratifications (U&G) research
resulted in the selection of Africa’s first elected female head tradition. U&G as an approach originated in communications
of state, President Ellen Johnson-Sirleaf. research in the mid-1970’s [11], advancing the view that
An outcome of these years of civil conflict was the complete consumers of mass media make active choices and selectively
destruction of the fixed-line telephone infrastructure. The consume media in order to satisfy specific needs. More tersely,
copper network was wholly destroyed or looted and all but U&G can be said to focus on what people do with media, as
one switch was destroyed [8]. It is clear from Figure 1(a) the opposed to what media does to people [12]. More recently,
steady decline in mainline penetration, starting from a very low the U&G approach has been applied to study adoption and
level to begin with. The precipitous drops evident in 1991 and use of new media technologies, including telephones [13], the
2003 are the outcome of the two major civil wars such that Internet [14], and mobile phones [15]. Also notable about the
by the time of the establishment of peace all mainlines where U&G tradition is attention to a broad range of motivations,
gone. including those which go beyond the purely instrumental or
On the other hand, mobile telephone adoption in Liberia has utilitarian (such as increased productivity or personal safety)
recently been growing at a staggering rate, as shown in Figure to the intrinsic, social, or to quote McClatchey [16], ‘hedonic’
1(b). It has been shown that teledensity phone penetration rates motivations for use.
are likely to over count the number of actual subscribers (due A typical U&G based study proceeds in one of two ways:
to purchased but inactive accounts) and, as already mentioned, either by starting with a hypothetical set of possible uses
significantly undercount the number of actual users (due to and seeking to confirm or deny each one, or in a more
sharing) [7]. In Liberia we estimate the subscribers to users exploratory fashion, starting off with no such initial set. As will
ratio to be as high as one to five. Competition within the be seen, our study walks a line between these two alternatives.
Liberian mobile phone sector is also robust with four active However, it must be noted that U&G in itself is not a method.
operators. Indeed, usage costs are reportedly the lowest in West Indeed, previous studies have employed a variety of different
Africa [34]. All county capitals and most other population methods to investigate uses, including surveys [17], semi-
centers currently receive signal from at least one of the mobile structured interviews [13], and focus groups [18].
providers’ services, and two providers currently offer GPRS A recent investigation of mobile phone usage among mi-
mobile internet services. Operators are actively extending both croentrepreneurs in Kigali, Rwanda [15] also drew upon the
their networks and services. U&G approach. In using the Q-sort methodology (also used in
The striking success of Liberia’s mobile sector, which this study and described in the next section), Donner identified
continues to develop at a feverish pace despite the resource- four archetypal ‘factors’ which speak to predominant uses of
strapped country, is a cause for optimism. mobile phones in Kigali. They were: convenient, intrinsic,
indispensable, and productive. Donner remarked on the diverse
C. Mobile phone usage in post-conflict settings nature of those factors, saying that they ‘’‘suggest numerous
Regrettably, civil conflicts such as those experienced in paths for future research.”
Liberia are not unusual in contemporary times. Indeed, while Our research is intended as an extension of this body of
inter-state wars are increasingly less common, the incidence research on uses of mobile phones to an immediate post-
of civil conflict is on the rise [10]. Thus the study of ICT’s conflict context. To our knowledge, ours is the first study
within countries emerging from civil conflict is an area of of mobile phone uses and gratifications in such an environ-
considerable importance though we note a paucity of work in ment. We believe that this context may give rise to unique
this area [9]. motivations for use, especially given the vibrancy of Liberia’s
470
Concept Statement F1 F2 F3 F4
Connectivity I use my mobile phone to stay in touch with my cus- Urban # of Defining Participants 9 13 12 4
tomers. % of Variance Explained 10% 13% 11% 7%
I use my mobile phone to stay in touch with my suppliers. Rural # of Defining Participants 13 8 5 6
My phone gives me access to new customers. % of Variance Explained 20% 12% 11% 12%
I use my mobile phone to stay in touch with my family.
My mobile phone helps me come and go without worry-
ing about missing calls.
components analysis to identify initial factors within the data,
I use my mobile phone to stay in touch with my friends.
followed by a varimax rotation to arrive at the final set of
Information My mobile phone helps me find work.
factors.
My mobile phone helps me keep informed about prices
in my business. Each such factor can be thought of as an archetypal per-
Intrinsic Having a mobile phone makes me feel more important. spective; a sorting of the statements that defines one group
Having a mobile phone makes me feel more connected of subjects against the others. Once these factors have been
to the world. identified, a loading score is computed for each partici-
I like customizing my mobile phone with accessories like pant/factor combination, which measures the similarity of that
special sounds and carrying cases
participant’s perspective to the archetypal perspective of that
I enjoy talking to my friends and family on my mobile.*
factor. A participant is said to load on or define a factor if
Having a mobile phone makes me happy.
their loading score for that factor crosses a certain threshold.
My mobile phone is stylish.
As a result of this process, each factor is associated with a set
Productivity My business is easier now that I have a mobile phone.
of participants defining it. In a sense the set of participants
My family is better off because I have a mobile phone.
have now been clustered into a small number of factors (we
My mobile phone saves me time.
find four factors in our study) with each participant assigned
My mobile phone lets me get more done during the day.
to that factor that best represents them.
My mobile phone helps my business save money.
In the final step, the statement rankings for each participant
My mobile phone helps me make more money in a day.
are weighted according to that participant’s loading score for
Security I use my mobile phone for emergency calls.
the factor they are assigned to (therefore if they define the fac-
My mobile phone makes me feel more secure.
tor more closely their ranking will have more weight). Then all
Other Getting a mobile phone changed the way I do business.
I am interested in learning about new features or mobile
of these weighted ranks are combined among the participants
phone models. assigned to each factor such that each factor is then described
I can’t do business without my mobile phone. by a list of single Z-scores, one per statement, along with a
I was among the first of my friends and business asso- p value assessing the value’s statistical significance. Finally,
ciates. these Z-scores are re-projected back into the original space of
I give my mobile phone number to many people. values from -4 to +4 (from “describes me least” to “describes
I share my mobile phone with my family or friends. me best”) as shown in Figure 2. And as an aid to comparison
I keep my mobile phone with me at all times. between the groups each set of Z-scores are projected into the
My mobile phone gives me more control over who I talk space described by the other factors as well.
to, and how/when I talk to them.
PQMETHOD also determines for each a factor a set of
I use my phone more for business more than for social
calls. “distinguishing statements” which differentiate the factor from
* Due to a miscommunication, in the urban study, this statement was the others. These statements are of special importance as they
replaced with “I bought my mobile phone for business.” This change are most representative of the differences between the factors.
was incorporated into the analysis that follows. It is in examining these representative statements that insight
into the meaning of each factor can finally be gained.
Tables III and IV, which we will go on to study below, show
D. Analysis Methods these sets of factors and their most distinguishing statements,
Q-analysis is usually performed using PQMETHOD, a along with those statements’ Z-scores and the -4 to +4 values
software package developed specifically for the task. A typical associated with them.
Q-analysis involves several steps. Initially, a large correlation
matrix is created, describing the similarities between the Q- IV. R ESULTS
sorts of all pairs of participants. We then look for ways to The Q-sort data we obtained from urban and rural par-
reduce the information in this matrix into an interpretable ticipants in Liberia have been analyzed separately. This has
form, a process which is both iterative and partially subjective. allowed us to examine differences in mobile phone use and
There are several routes to this goal, a review of which goes perception between these populations.
beyond the scope of this work. In our analysis we chose a Following the procedures described above, PQMETHOD
procedure similar to Donner [15]; we performed a principal was used to perform our analysis. After the principal com-
472
TABLE III
U RBAN FACTORS
473
TABLE IV
RURAL FACTORS
ponent analysis, several factor rotations were computed and with ranks for the present factor of interest of +3 or +4, or
examined. For both data sets, a set of four factors was found -3 or -4, are placed under the headings ‘Describes Me Best’
to provide the best balance between explanatory power and and ‘Describes Me Least’, respectively. Other distinguishing
succinctness. statements with a high (or low) rank for the present factor
Table II displays the number of participants found to load relative to the other factors are placed under the heading
on each factor, as well as the percentage of initial variance ‘Relatively High’ (or ‘Relatively Low’).
that each explains. Each factor can be taken to represent an
archetypal perspective regarding phone use among Liberians. Below, we review the distinguishing statements for each
The explained variance proportions we obtained are similar to factor in an effort to interpret the nature of the archetypal
those obtained in previous studies. perspectives they represent. We also review commonalities
Tables III and IV show the commonalities across factors for across factors for the urban and rural groups. These common-
both datasets, and the statements distinguishing each factor, alities are statements which are consistently ranked positively
ordered by Z-score. The computed ranks for each factor are or negatively for each of the four factors, and thus indicate
also shown for each statement (F1, F2, F3, and F4). Statements agreement across most participants.
474
use my phone for emergency calls” was ranked +4, +3, +2, and I use my mobile phone for emergency calls
+1, while “I keep my phone with me at all times” was ranked F1 F2 F3 F4
+2, +2, +2, and +4. It was clear that many participants felt Urban Liberia 2 1 2 4
strongly about this function of the phone. Several participants Rural Liberia 1 2 3 4
offered compelling stories of using their phone to call for help Urban Rwanda 2 1 -4 -3
during a robbery, to call for medical care for a loved one, or
as a deterrent against sexual violence. My mobile phone is stylish
F1 F2 F3 F4
Factor 1: Business: Thirteen participants defined this factor.
Of the statements distinguishing it from the others, few were Urban Liberia -4 -2 -1 1
positive. Most prevalent among them was the assertion that Rural Liberia -4 -4 -4 -3
business is easier to conduct thanks to the phone, which was Urban Rwanda -3 1 -3 -3
rated +4. Participants in this group also claimed that their
phones helped their business save money. On the other hand, Having a mobile phone makes me feel
members of the group do not share their phone with friends more connected to the world
or family (-4), or especially rely on it for communication with F1 F2 F3 F4
family (0) or friends (-2). Overall, this group views the phone Urban Liberia 0 2 2 3
as a serious tool that has improved their ability to conduct Rural Liberia 3 3 2 3
business, much as in the urban factor of the same name. Urban Rwanda 2 1 3 3
Factor 2: Mixed: This factor describes a variety of personal TABLE V
P OINTS OF COMPARISON BETWEEN RURAL AND URBAN L IBERIAN DATA ,
uses of the phone. Above all, the eight participants defining AND URBAN RWANDAN DATA FROM D ONNER [15].
this factor enjoyed talking on the phone with their family and
friends (+4). However, they rated the statement “My phone
makes me feel more secure” quite high as well (+3), and they
view the phone generally as a time saver (+3). No business-
We discuss them below. Table V presents comparisons of
related statements were rated highly—the phone is not seen
several noteworthy statements.
particularly as making business easier (0), and the phone is not
used to stay in touch with customers (-2). We find this factor Perhaps the most striking and unique result of this study
to be a mix of personal, productivity and security uses with a is the prevalence of security and emergency use reported by
clear bias against business. Unlike the previous factors in this Liberian mobile phone users. Data from Monrovia suggested a
study, this group does not admit to a clear, single description. distinct factor emphasizing security, while rural data revealed
Factor 3: Intrinsic: The five participants defining this factor security as an item of consensus. In both cases, the emphasis
were clearly enthusiastic about intrinsic uses of their phones as on security was considerably stronger than that reported by
technological artifacts. Their most highly rated distinguishing Donner [15] for Rwandan users, as shown in Table V. This
statement was “Having a mobile phone makes me happy” (+3), greater emphasis could be due to Liberia’s much more recent
and they were also uniquely keen to learn about new features civil conflict. From informal discussions with participants, it
or models (+2), and, less strongly, to accessorize their phone was clear that the safety and security of self, of loved ones,
(0). On the other hand, participants in this group were not and of personal property is still a major concern in Liberia.
very interested in business functions of the phone (-2), or in This is by no means surprising. Despite the 15,000-strong
using it to find work (-2). They also reported not giving their UN peacekeeping mission, the country’s police force is still
number out to many people (-4), the reason for which is not under development, and many ex-combatants have turned to
clear. crime as a source of financial support. In such a situation, it
is understandable that a mobile phone is seen as providing
Factor 4: Mixed: Our analysis identified a large number of
security, as it allows the user to call a family member or an
statements with broad meaning as distinguishing this factor.
authority in the event of a crime or transgression. In several
The six participants defining this factor keep their phone with
cases, even police officers themselves spoke of their mobile
them at all times (+4), use it to find work (+3), and stay in
phones as a source of security.
touch with friends (+3). They also find that it makes their
business easier (+2), and they use it to stay in touch with Of all the findings of this study, this emphasis on security
their customers (+1). Thus whereas Factor 2 reveals a variety carries the most implications for possible future technology
of personal uses with a bias against business uses, this factor designs. Current phones, while providing access to centralized
suggests an even more general blend of valued uses. network security services such as 9-1-1, were not designed
for an environment with weak state institutions and a lack of
V. D ISCUSSION centralized security apparatuses. Instead one could imagine a
phone design incorporating a “panic” button feature, which
The results of our Q-sort analysis have highlighted a set of emits a loud noise and flashing light, and automatically
uses and gratifications for mobile phones in Liberia, some of contacts other phones, either in the immediate geographical
which are particularly interesting in light of previous research. area, or on a predefined emergency contact list. This feature
476
could be activated in an attempt to stop a crime in progress, internet cafs, leaving the mobile phone as the only link to the
or to act as a deterrent against potential offenders. outside world.
Another result of interest highlights a difference between Also common to the two studies was the finding of strong
urban and rural users within Liberia. While analysis of urban business-related factors. Our study identified clear business
data revealed four fairly well defined factors, two of the factors for both the urban and rural populations. In addi-
four rural factors admitted to a more blurry description. We tion, several other factors rated productivity-related statements
feel that this may be due to the multifaceted livelihoods highly, such as ‘My mobile phone helps me find work.’ It is
characteristic of rural citizens and/or the fact that the mobile clear that the mobile phone plays a crucial role in the largely
phone is often the only available ICT service in rural areas. We informal Liberian economy.
encountered many rural participants who reported a number of
different occupations. One claimed to work for the Ministry VI. C ONCLUSION
of Immigration in addition to being a farmer. Another taught This study has employed the uses and gratifications ap-
secondary school in addition to selling rubber. Several students proach and the Q-sort methodology to investigate mobile
reported also working in various family businesses. Such users phone usage among urban and rural Liberians. As in previous
are not likely to fit neatly into any one category, such as work, several distinct user groups were identified. Urban users
‘business’ or ‘connectivity’—their use of the phone is more saw their phones variously as productivity enhancers, means
varied. of connectivity to family and friends, essential business tools,
On the other hand, rural users displayed an overwhelming and security providers. A group of business users was also
rejection of the phone as an object of style. As reported, identified among rural users, as was a group of techno-
participants often openly scoffed at the “My phone is stylish” enthusiasts, and two groups which eluded definite description.
statement upon reading it from the card. It seemed that the The multifaceted nature of these groups replicates Donner’s
idea of a phone being stylish was absurd, and to consider principal finding [15] in highlighting the diverse uses and
it so would be shameful. This is in contrast to both the gratifications characteristic of mobile phone users in a low-
urban Liberian and Rwandan data, both of which revealed a income region.
factor in which the same statement received a positive rating, However, the chief difference between these two studies—
as shown in Table V. Rural Liberians universally seem to our identification of the prevalence of security use among
view their phones as serious tools, not fashionable accessories. Liberians—is potentially important in its own right. As stated,
While in central Monrovia there exists a group of relatively we believe that this finding may be related to the nature of
successful business people that exhibit consumerist behaviors, Liberia’s immediate post-conflict environment. Not only does
consumerist populations are mostly absent throughout rural such a finding have implications for future technology designs
Liberia. This is due to poverty and the pervasive lack of a (which we are interested to explore), it also suggests further
formal economy in many of these areas. We suspect that in research into the role of ICTs in the process of stabilizing and
such a context the idea of flaunting or fetishising a phone rebuilding a nation following a civil conflict. The unfortunate
as stylish seems disassociated with local realities. Also due to fact of widespread civil conflict in today’s world makes
limited resources, most participants owned the least expensive, understanding such phenomena even more important.
most basic phone models, and seemed to be aware of the A more general theme that we encountered is the sheer
humbleness of their devices in comparison to the phones indispensability of the phone for most users. In many cases,
available in the city. This awareness is sure to preempt any the mobile phone is their only option for communications other
pretensions of fashionability. than physical travel, which is costly and time consuming. In
With this in mind, the emergence of an intrinsic factor other places where methods of communication are various and
focusing on technological enthusiasms in the rural data be- many, the idea of a single modality being so essential is harder
comes noteworthy. Members of that group expressed interest to fathom. But many participants we spoke to related stories
in learning about newer more advanced phones, despite the of the phone saving them many miles of travel. Businesspeo-
fact that they were likely to be unaffordable. Participants often ple celebrated the time saved in ordering goods from their
spoke about this interest in technology as if it were a civic suppliers over the phone, instead of traveling by costly public
duty - that any good citizen should be up to speed with the transportation, sometimes only to find the supplier out of stock.
latest technology. We suspect that this group is ultimately One participant described a hypothetical situation in which
aspirational ascribing to a vision where technology serves as her daughter had fallen ill and she was without a phone. How
an engine of their personal, and the nation’s development. should she know where to take her, when the only doctor in
One finding from Donner’s study which was mostly repli- the area could be in any of several different towns, each a
cated in our data was the feeling that the phone supports considerable distance away? Add to this the security role they
connectedness to the world, as also shown in Table V. This apparently perform, and it is clear that the phone is a truly
finding was especially prevalent among rural users, which is indispensable item.
not surprising given the fewer options available to rural users A weakness of this work arises from the requirement that
for communicating internationally. Some parts of Liberia, such participants be able to read. This was the unfortunate reality,
as Sinoe County, are without radio stations, newspapers, or since performing a Q-sort requires rapid and repeated visual
477
scanning of the statement cards. Unlike in a traditional survey, [14] J. Eighmey and L. McCord, “Adding value in the information age: Uses
we felt that simply reading the statement to the participant and gratifications of sites on the World Wide Web,” Journal of Business
Research, vol. 41, no. 3, pp. 187–194, 1998.
once would not be sufficient to support the sorting process. Es- [15] J. Donner, “Microentrepreneurs and mobiles: An exploration of the uses
pecially in rural areas, this unfortunately excluded a significant of mobile phones by small business owners in Rwanda,” Information
number of potential participants. In future, we are interested Technology and International Development, vol. 2, no. 1, pp. 1–22, 2004.
[16] S. McClatchey, “The consumption of mobile services by Australian
in experimenting with study designs using iconography and/or university students,” International Journal of Mobile Marketing, vol. 1,
sounds in order to allow non-literate participants to share their no. 1, pp. 1–9, 2006.
view. [17] K. Aoki and E. J. Downes, “An analysis of young people’s use of and
attitudes toward cell phones,” Telematics and Informatics, vol. 20, no. 4,
ACKNOWLEDGEMENTS pp. 349–364, 2003.
[18] F. Williams, H. Dordick, and H. Jesuale, Focus group and questionnaire
We thank Jonathan Donner for kindly providing his data for development for exploring attitudes towards telephone service. Herbert
the purpose of comparison. We also thank the various Liberian Dordick and Associates, 1985.
[19] W. Stephenson, The study of behaviour: Q technique and its methodol-
property owners who provided facilities for this study to take ogy. Chicago: University of Chicago Press, 1953.
place, and the guards at the Carter Center for their help in [20] E. B. Brownlie, “Young adults’ constructions of gender conformity and
pilot testing our study apparatus. Finally we thank the three nonconformity: A Q methodological study,” Feminism & Psychology,
vol. 16, no. 3, pp. 289–306, 2006.
anonymous reviewers for their constructive comments on this [21] D. Shemmings, “‘Quantifying qualitative data: An illustrative example
work. of the use of Q methodology in psychosocial research,” Qualitative
Research in Psychology, vol. 3, no. 2, p. 147, 2006.
R EFERENCES [22] M. Lister and D. Gardner, “Engaging hard to engage clients: A Q
methodological study involving clinical psychologists,” Psychology and
[1] “The real digital divide,” The Economist, 2005, March 12. Psychotherapy-Theory Research and Practice, no. 79, pp. 419–443,
[2] E. J. Wilson, M. L. Best, and D. Kleine, “Moving beyond ‘The real dig- 2006.
ital divide’,” Information Technologies and International Development, [23] P. H. D. Stenner, G. Bianchi, M. Popper, M. Supekova, I. Luksik, and
vol. 2, no. 3, pp. iii–v, 2005. J. Pujol, “Constructions of sexual relationships—a study of the views
[3] M. L. Best and C. Kenny, “ICT’s, Enterprise and Development”. In of young people in Catalunia, England and Slovakia and their health
ICT4D by T. Unwin. Cambridge, UK: Cambridge University Press, implications,” Journal of Health Psychology, vol. 11, no. 5, pp. 669–
2009. 684, 2006.
[4] L. Waverman, M. Meschi, and M. Fuss, The Impact of Telecoms on [24] S. J. Snelling, “Q methodology and the political opportunity of feminist
Economic Growth in Developing Countries. Vodafone, 2005. psychology,” Feminism & Psychology, vol. 14, no. 4, p. 519, 2004.
[5] R. Jensen, “The digital provide: IT, market performance and welfare [25] A. Risdon, C. Eccleston, G. Crombez, and L. McCracken, “How
in the South Indian fisheries sector,” Quarterly Journal of Economics, can we learn to live with pain? A Q methodological analysis of the
2001. diverse understandings of acceptance of chronic pain,” Social Science
[6] V. Gray, The un-wired continent: Africa’s mobile success story. ITU, & Medicine, vol. 56, no. 2, p. 375, 2003.
2006. [26] L. D. Bryant, J. M. Green, and J. Hewison, “Understandings of
[7] J. James and M. Versteeg, “Mobile phones in Africa: How much do we Down’s syndrome: A Q methodological investigation,” Social Science
really know?” Social Indicators Research, no. 84, pp. 117–126, 2007. & Medicine, vol. 63, no. 5, p. 1188, 2006.
[8] M. L. Best, K. Jones, I. Kondo, D. Thakur, E. Wornyo, and C. Yu, [27] R. M. Baker, “Economic rationality and health and lifestyle choices for
“Post-conflict communications: The case of Liberia,” Communications people with diabetes,” Social Science & Medicine, vol. 63, no. 9, p.
of the ACM, vol. 50, no. 10, pp. 33–39, 2007. 2341, 2006.
[9] D. Thakur and M. L. Best, “The telecommunications policy process [28] I. Goldman, “Q methodology as process and context in interpretivism,
in post-conflict developing countries: The case of Liberia,” in The communication, and psychoanalytic psychotherapy research,” The Psy-
36th Research Conference on Communication, Information, and Internet chological Record, vol. 49, no. 4, p. 589, 1999.
Policy (TPRC), 2008. [29] M. W. Kramer and P. M. Pier, “Students’ perceptions of effective and
[10] W. Long and P. Brecke, War and reconciliation: Reason and emotion ineffective communication by college teachers,” The Southern Commu-
in conflict resolution. Cambridge, Mass.: MIT Press, 2003. nication Journal, vol. 65, no. 1, p. 16, 1999.
[11] J. G. Blumler and E. Katz, The uses of mass communications: Current [30] N. J. A. van Exel, G. de Graaf, and W. B. F. Brouwer, “‘Everyone dies,
perspectives on gratifications research. Beverly Hills: Sage Publications, so you might as well have fun!’ Attitudes of Dutch youths about their
1974. health lifestyle,” Social Science & Medicine, vol. 63, no. 10, p. 2628,
[12] W. Chigona, G. Kankwenda, and S. Manjoo, “The uses and gratifications 2006.
of mobile internet among the South African students,” in PICMET 2008: [31] A. C. Borthwick, T. Stirling, A. D. Nauman, and D. L. Cook, “Achieving
Portland International Conference on Management of Engineering & successful school-university collaboration,” Urban Education, vol. 38,
Technology, 2008, pp. 2197–2207. no. 3, p. 330, 2003.
[13] J. W. Dimmick, J. Sikand, and S. J. Patterson, “The gratifications of [32] J. Szente, J. Hoot, and J. Ernest, “Parent/teacher views of develop-
the household telephone: Sociability, instrumentality, and reassurance,” mentally appropriate practices: A hungarian perspective,” International
Communication Research, vol. 21, no. 5, pp. 643–663, 1994. Journal of Early Childhood, vol. 34, no. 1, p. 24, 2002.
[33] A. C. McKennel, “Surveying attitude structures—discussion of princi-
ples and procedures,” Quality & Quantity, vol. 7, no. 2, pp. 203–294,
1974.
[34] R. Southwood. “Liberia: Four mobile companies bring lowest
prices in West Africa” Balancing Act. http://www.balancingact-
africa.com/news/back/balancing-act 352.html#head (accessed February
20, 2009).
478
Abstract—This work advances the state-of-the-art in assistive which monitors the student’s writing and transmits data in real time
technology for the visually impaired by enhancing an automated to a computer to provide immediate audio feedback to the user. The
tutor designed to teach beginner-level braille writing skills. The latest version of the BT consists of two rows of 16 braille cells and
hardware component of the tutor is a low-cost device that six buttons placed over the top of the two rows to work as an input
provides an intelligent tactile interface that connects to a laptop area. The stylus is a standard braille stylus that connects to the BT by
or desktop computer. We demonstrate the many uses of this its metal tip. Moreover, students can press on one of two control
tutor including its multilingual capabilities, interactive exercises, buttons placed on the sides of the BT to perform a variety of mode
and educational games. Preliminary field tests of the tutor in changes. Further details of the BT are discussed in prior publications
India, Zambia, and Qatar have demonstrated the usefulness of by Kalra et al. ([1] and [2]). A new design for the packaging the BT
this automated tutor, and its potential to enhance braille literacy was also explored (Fig. 1.).
in developing regions.
I. INTRODUCTION
Fig. 1. Prototypes and preliminary field testing of the Braille Writing Tutor
Abstract—FRUILEMA, a Malian fruit and vegetable economic further inform partners about the supply process and improve
interest group (EIG), is acquiring the tools and resources needed north-south linkages that promote not only trade, but also
to meet the requirements (GLOBALGAP certification) for mutual understanding and collaboration. This platform allows
exporting Malian farm products to international markets. for the better identification of producers, to know their
Starting with the highly marketable mango fruit, an ICT solution collective production forecasts, and to trace the entire chain of
has been developed to enable the identification of producers, to
activities, from the purchasing of agricultural products back to
trace farm plots and produce, and to obtain quantitative and
qualitative information on produce to support analysis and the producer. During the first year of the system being in
decision-making. The information is accessible via a web-enabled operation, FRUILEMA was able to use the platform to
platform and can be updated using Internet and PDAs. The convince European fruit importers to work with them,
information is disseminated to producers, trackers, collectors, resulting in two contracts for the 2008 mango harvesting
importers, and other actors in the mango value chain. season.
The platform was developed during 2007-2008, and
Index Terms—Agriculture, Geographic information systems, involved the identification and translation of all processes in
Information systems, Software the production and "conditioning" of mango products,
software development as well as capacity building of the five
exporting organizations involved. The platform was
I. CASE (CONTEXT, NOVELTY AND SIGNIFICANCE)
developed in collaboration with the five exporting
F
relevant. Global Voices' Ethan Zuckerman said recently "I'm
reedom Fone is a free open source software tool that can
surprised there hasn’t been more work done making
be used to build and update a dial-up information service
interactive voice response systems usable for development
in any language.
purposes [1]." Freedom Fone provides exactly that.
Its easy to use interface lowers the barriers to using Interactive
Other projects using Interactive Voice Response to make
Voice Response (IVR) for outreach to communities small and
critical development information more available local
large, near and far. Freedom Fone empowers non-technical
communities are VoiKiosks, which “present an alternative
organizations to build automated information services that are
platform to create and host such information kiosks in the
available to the public 24 hours a day, 7 days a week – power
telephony network,”[2] and MobilED, whose “Audio Wiki”
and connectivity allowing.
programme uses a range of software products including
Based on the concept of Interactive Voice Response (IVR),
Asterisk, Kannel and Flite to make audio information
pre-recorded audio files are stored by Freedom Fone in a
available to learners [3].
Content Management System (CMS). This is updated through
a simple to use browser interface. Content is made available to
REFERENCES
telephone callers when they dial phone numbers associated
with the service. Users are presented with an audio menu of [1] Ethan Zucherman. “Delivering Ethiopian teff via Czech taxi?” My
options and using their telephone keypads select channels Heart’s in Accra Blog. October 13, 2007.
carrying information of interest to them. They are able to http://www.ethanzuckerman.com/blog/2007/10/13/delivering-ethiopian-
teff-via-czech-taxi/
listen to a selection multiple times if they so choose. [2] Sheetal Agarwal, Arun Kumar, Amit Anil Nanavati, Nitendra Rajput.
Deployment in any language is possible as key global files for “VoiKiosk: Increasing Reachability of Kiosks in Developing Regions”,
menu prompts can be uploaded through the browser interface WWW 2008 / Poster paper.
to the CMS. [3] Server – MobilED website http://mobiled.uiah.fi/?page_id=98
Individuals can contribute questions, content and feedback
by leaving voice messages via the IVR interface. Freedom
Fone can be operated as a collective, with different groups
managing different channels (IVR menu options) of
information from the same installation.
Currently available as a prototype, Freedom Fone can be
connected to telephony services via the internet using VoIP –
this is the most scalable option; or via telephony cards to
either analog or digital phone lines. Services can be offered at
Abstract—FrontlineSMS and Ushahidi are both free and open developed in 2005 and updated in 2007 – is being used around
source tools which allow for the collection, dissemination and the world for a wide range of non-profit activities including
visualization of data collected to and from the field. This paper the sending of market prices and other agricultural data to
outlines details of live demonstrations of the tools at the ICTD smallholder rural farmers in Aceh, Cambodia and El Salvador,
2009 conference in Qatar.
the dissemination of news in Iraq, the sending of security
Index Terms—FrontlineSMS, Ushahidi, data, messaging, open alerts to fieldworkers in Afghanistan, for human rights work
source tools, mobile, SMS in places such as Zimbabwe, Pakistan and the Philippines,
and the running of a rural healthcare network for 250,000
people in Malawi. Because the software can be used on a
I. INTRODUCTION single laptop computer without the need for the internet, it has
been widely adopted among the grassroots non-profit
FrontlineSMS and Ushahidi are two fully free and open
community and nominated for several awards.
source tools which leverage information and communication
technologies – of which mobile is a key component – to allow
individuals or groups to collect and/or disseminate FrontlineSMS is currently the focus of a number of academic
information to and from the field. The tools work studies, and the recent UN Foundation/Vodafone report on
independently, but can also work seamlessly together, as they mHealth described it as “one of the largest and most
are currently doing in a deployment in DRC. ambitious mHealth programs in the world”.
Abstract—ICT enables peaceful change in communities war, Ziad from Gaza City commented: "The situation is really
affected by conflicts, poverty and injustice. In this demo, we bad in Gaza (…) I am happy about the posts on the blog. It is really
illustrate how young people use multimedia, cell phones and the nice to know that there are people who care about us and support
internet to become global citizens responsible for building secure, us…this is encouraging.” Rebuilding hope is sometimes harder than
productive and just communities at home and around the world. infrastructure but a necessary condition for peaceful recovery.
Global Youth Connectivity allows disaffected youth to express According to a UNDP report on Arab youth, global communication
thoughts and feelings, while also scaling global citizenship systems also provide a gateway for youth to delve into creativity,
education and civic engagement across borders. Thousands of innovation, learning, professionalism, and democratic governance:
Middle-Eastern and American youth have exchanged content
“Their self-confidence is boosted when they expand their practical
and discussed online since 2005. Innovations and partnerships
should enable Mercy Corps to empower millions more living in knowledge and communicate with their peers around the world” 4.
extreme vulnerability.
III. INFORMATION + COMMUNICATION = PROGRESS
Index Terms— Civic engagement, Global Citizenship, Global Many young people in the Middle-East suffer from
Youth Connectivity
repression, isolation and limited opportunities for social or
civic engagement. Mercy Corps and IJMA35 are working on a
grassroots news network with information organized by youth
I. INTRODUCTION
for youth. After a newspaper-based pilot in Iraqi villages, the
Abstract—Organizations in developing regions want to effi- clinic operations were limited to using paper forms to ensure
ciently collect digital data, but standard data gathering practices timely information access. Only after a labor-intensive delay
from the developed world are often inappropriate. Traditional did the medical researchers enjoy the benefits of digital data
techniques for form design and data quality are expensive and
labour-intensive. We propose a new data-driven approach to form for research and analysis.
design, execution (filling) and quality assurance. We demonstrate We have built a system called U SHER that maximizes data
U SHER, an end-to-end system that automatically generates data quality at entry-time using statistical data modeling, dynamic
entry forms that enforce and maintain data quality constraints interfaces, and collaborative insight. Guided by prior data,
during execution. The system features a probabilistic engine that U SHER learns probabilistic relationships in the data to train a
drives form-user interactions to encourage correct answers.
model, which is then applied to automatically generate forms
I. I NTRODUCTION with the appropriate constraints. U SHER then provides real-
Governments, companies, and individuals routinely make time feedback during the data entry process to dynamically
important decisions based on inaccurate data stored in suppos- guide (or usher) the user toward better data quality.
edly authoritative databases. In healthcare, a simple error may Based on a list of form questions and a sufficient set
have fatal consequences. While data quality can be addressed of answers, U SHER optimizes the form’s question-ordering
at every stage of the data life-cycle, from creation to archival, and layout, mimicking survey design principles. During form
we believe that entry-time is the first and best opportunity to entry, U SHER provides dynamic data-quality feedback to the
improve the quality of manually-entered data. There is much user. When the user enters a value, U SHER automatically
prior work on improving the quality of data that already resides decorates the interface with hints and warnings if the answer
in a database [1]. However, relatively little attention has been is deemed “risky.” Decoration choices are probabilistically
paid to improved techniques for data entry. guided, and include auto-complete, correctness-thermometers,
Survey design [2] has long informed the design of data entry warning/error flags, and other scented widgets [5]. U SHER
forms, applying principles for data encodings, constraints, also invites the user to write and view comments about
and validation rules. For electronic forms, quality assurance form questions or data instances for and by other users.
during entry has centered on the ubiquitous and costly practice Finally, U SHER mimics double-entry by choosing to re-ask
of double-entry [3]. Current standards have failed to take questions with responses likely to be erroneous, based on the
advantage of new technology: pervasive cellular networking probabilistic model.
and low-cost mobile devices allows even remote users to II. D EMONSTRATION
interact with data entry systems that could potentially provide
rich feedback. Our demonstration will show U SHER’s ability to approxi-
For organizations with limited resources, existing standards mate expert form design and double-entry based only on prior
are neither practical nor attainable. In such settings, designing data, both on a PC and a mobile device. Using a real dataset
data collection instruments is too often an ad hoc practice, from a rural health organization, users will be able to 1)
consisting of mapping desired information elements to a set automatically extract training data from a Microsoft Access
of entry widgets (text fields, combo boxes, etc.), guided only database; 2) refine the automatically designed form; and 3)
by the designer’s intuition. According to recent work on data execute the forms with and without smart decorations and
collection in resource-poor settings, lack of expertise and quality assurance.
difficulty of remote data collection are the chief obstacles R EFERENCES
to high data quality [4]. In our previous fieldwork with a [1] J. M. Hellerstein, “Quantitative data cleaning for large databases,” United
well-funded HIV/AIDS treatment program in East Africa, we Nations Economic Commission for Europe, 2008.
found that little thought was given to form design, and a [2] R. M. Graves, F. J. Fowler, M. P. Couper, J. M. Lepkowski, E. Singer,
and R. Tourangeau, Survey Methodology. Wiley-Interscience, 2004.
haphazard double-entry program bottlenecked the data entry [3] S. Day, P. Fayers, and D. Harvey, “Double data entry: what value, what
process to a substantial degree; in fact, the program’s health price?” Controlled Clinical Trials, 1998.
[4] J. V. D. Broeck, M. Mackay, N. Mpontshane, A. K. K. Luabeya,
K. Chen, N. Conway and J. M. Hellerstein are with the Computer M. Chhagan, and M. L. Bennish, “Maintaining data integrity in a rural
Science Division, University of California at Berkeley (email: {kuangc, nrc, clinical trial,” Controlled Clinical Trials, 2007.
hellerstein}@cs.berkeley.edu). [5] W. Willett, “Scented widgets: Improving navigation cues with embedded
H. Chen is with the Computer Science and Artificial Intelligence Labora- visualizations,” IEEE TVCG, 2007.
tory, Massachusetts Institute of Technology (email: harr@csail.mit.edu)
H. Dolan and T. S. Parikh are with the School of Information, University
of California at Berkeley (email: {dolan, parikh}@ischool.berkeley.edu)
Manuscript accepted Feb 10, 2009.
488
REFERENCES
I. WHITEBOARDS [1] BECTA (2003). What the research says about interactive whiteboards.
Abstract - It has been estimated that 2/3rds of the individuals focus on mastitis was twofold. First, mastitis, historically, is
living on less than $2/day own or keep livestock. However, difficult for farmers both in the North and in the South to
research has illustrated the difficulties that this population face in diagnose. Second, delayed diagnosis has a detrimental impact
obtaining accurate and relevant information regarding animal
health. Therefore, the demonstration offers two tools to support
on milk production and consequently income. As such, the
the animal health needs of the poor in India. First, the creation of disease has a large impact on the livelihoods of poor dairy
a mobile phone based knowledge transfer tool is described and producers, particularly in India [1].
second, a mobile phone-based diagnostic tool to enhance the The reason for choosing a mobile phone application is
capacity of the poor to diagnose diseases is also detailed. The twofold. First, mobile phones are better from both a cost and
demonstration of the tools will support the ‘hands-on’ exposure of battery life perspective [3]. Second, by connecting via
conference goers with problems faced by the poor and the role of
Bluetooth to the Livestock Guru kiosk, the MDS can link to
ICTs in their solution.
learning information provided by the Guru in relation to the
Index Terms - Information Delivery System, Livestock Diseases, livestock disease in question.
Mobile Phone, Poverty Nevertheless, a version of the MDS entirely contained within
a mobile phone was created and will be demonstrated as many
poor livestock keepers may be too remote from a server (The
I. INTRODUCTION Livestock Guru kiosk), in this manner, a farmer may take a
ICTs are critical to addressing the animal healthcare needs of picture of his or her animal and the mobile MDS will provide a
diagnosis. The software was created based upon a combination
poor livestock keepers in the South [4]. First, such tools may
be applied as a means of knowledge transfer and secondly, of prototype matching and statistical pattern recognition.
ICTs can help redress historical gaps in animal health service
II. DEMONSTRATION
delivery to the poor. Indeed, research has identified that women
and children, while primarily livestock care-takers are often the In the demonstration, a mobile phone based on Symbian
most excluded from knowledge transfer [5]. Equally, disease S40/60 OS will be utilized to display the Mobile Livestock
diagnosis is particularly problematic for the poor [5]. Guru. Its application among poor farmers in India (including
Therefore, in recent years there has been an increasing interest women and children) will be detailed as part of the display.
in tele-veterinary medicine to address animal healthcare gaps in Additionally, use of the MDS will also be shown. As such, a
southern countries. Mobile diagnostic technologies are a display of symptom pictures will be offered and conference
rapidly emerging field in tele-medicine (both human and goers may utilize a mobile phone to take and send symptom
animal) and are defined by two key characteristics: mobility pictures to a kiosk computer to make a diagnosis of a sick
and remote diagnosis [2]. animal. Alternatively, a mobile phone stand-alone version
In the following demonstration, the authors display three MDS will be also available for conference goers to test.
linked and inter-related tools to address these issues: first, a Further, a video of the tool in use in India will also be provided.
mobile phone version of the Livestock Guru, a multi-media
learning program for poor farmers in the South; second, REFERENCES
multi-media learning games for children and third, a Mobile [1] C. Heffernan, D. Pilling, J. Yu and Y. Lin, “Listening to the Voices of the
Diagnostic System (MDS), which supports the diagnosis of Poor: Enhancing evidence-based decision-making in livestock
development”, Nottingham University Press, Nottingham, United
Mastitis among poor dairy producers in India. The MDS Kingdom, 2008.
enables farmers to take a picture of their sick animal by a [2] M. Papageorges, P. Herbert, “Telemedicine using standard internet
mobile phone and then send this picture to a kiosk, which technologies (telemedicine for less than $1000)”, Clinical Techniques in
Small Animal Practice, vol. 16, no. 2, pp. 102-107, May 2001.
subsequently makes the diagnosis. The reason for the initial [3] L. Naismith, “Literature Review in Mobile Technologies and Learning: A
Report for NESTA Futurelab”. Bristol: NESTA Futurelab, 2004.
[4] S. Warren, “A distributed infrastructure for veterinary telemedicine, in
Manuscript received February 2, 2009. proceeding of the 25th annual International Conference of the IEEE
Jun Yu is a PhD student at the Livestock Development Group, School of EMBS, September 17th-21st, Cancun, Mexico, 2003.
Agriculture, Policy and Development, University of Reading, UK (e-mail: [5] Yu, J. “Improving Development Information flows: The creation of ICTs
aar04jy@ reading.ac.uk). for poverty alleviation in the livestock sector”, Thesis (submitted). The
Claire Heffernan is the director of the Livestock Development Group, University of Reading, Reading, UK.
School of Agriculture, Policy and Development, University of Reading, UK
(Tel.: +44 1189 318213, e-mail: c.l.heffernan@reading.ac.uk).
490
Abstract—Traditional Single Display Groupware (SDG) solu- work. To do this mapping effectively, we have to map mouse
tions have been used to create software for developing regions. movements as well as mouse clicks.
SDG allows for greater utilization of the limited infrastructure
available in such areas. This technology is particularly appropri- A. Movement
ate in classrooms where groups of children often use educational
software on a single computer. However, SDG has faced several To map the many cursors down to one, we use a well known
challenges working with legacy applications. Our technology, SDG technique. We average the mouse locations, and display a
called metamouse, takes a step toward an integrated multi-user metacursor at that position. With this, each user’s actions have
application by allowing users to collaborate within unmodified a small, but noticeable, impact on the metacursor’s location.
legacy educational software. Likewise, if all of the users agree on the position of the cursor,
I. P ROBLEM S TATEMENT the metacursor will be at that same location. This encourages
collaboration, as the users must discuss and agree to place the
Several million children, especially in the developing world,
metacursor where they want it.
never use a computer without a partner sharing the mouse
and keyboard [2]. Single Display Groupware techniques [4] B. Clicking
are particularly relevant to this scenario, allowing for users to We can’t allow the users to click at any time, as the
more effectively share computing resources by using multiple metacursor may be at an inappropriate location. To solve this,
mice. Existing research has shown that these techniques are we only allow clicking when the mice are in close proximity
intuitive, usable, and beneficial in the developing world [3]. to each other. When distant, all of the clicks are ignored. The
Despite these encouraging early findings, two important intuition is that if the mice are close to each other, then the
barriers remain in deploying shared input technologies in real users have agreed on the correct location to make progress,
world settings. First, most legacy software would have to and clicking should proceed. This encourages collaboration;
be significantly modified to effectively use multiple mice, all users must agree on a location before making progress.
but among other issues, many business challenges exist in
doing this. The second barrier, of interest from the HCI III. D EMONSTRATION
perspective, is that of efficiently encouraging coordinated on- We are going to demonstrate the point-and-click metamouse
screen decision-making using multiple mice. Most trials of system running on Microsoft Windows XP. We will run a
multimouse, while emphasizing the importance of collabora- set of educational games from the following vendors using
tion, have been impeded by working within a “fastest-finger- metamouse and four individual mice.
first” race-clicking, or repetitive click modes which require all • The Azim Premji Foundation [1]
users to click on the same link for the function to continue. Our • Disney online
challenge was to work with legacy software, and yet provide • Leapfrog
an alternate way for allowing coordinated clicking to capture
the proven learning gains of collaboration. R EFERENCES
II. S OLUTION [1] Azim Premji Foundation. http://www.azimpremjifoundation.org.
[2] J. Pal, U. S. Pawar, E. A. Brewer, and K. Toyama. The case for multi-user
To solve these problems, we developed the metamouse. design for computer aided learning in developing regions. In WWW ’06:
The metamouse does two things. First, it maps multiple user Proceedings of the 15th international conference on World Wide Web,
pages 781–789, New York, NY, USA, 2006. ACM.
mice and cursors down to one metacursor that interacts with [3] U. S. Pawar, J. Pal, and K. Toyama. Multiple mice for computers
the application. This allows for the use of unmodified legacy in education in developing countries. In IEEE/ACM Intl Conf. on
applications. Secondly, we have to make sure that this mapping Information and Communication Technologies for Development, 2006.
[4] J. Stewart, B. B. Bederson, and A. Druin. Single display groupware:
is intuitive and encourages collaboration among users. a model for co-present collaboration. In CHI ’99: Proceedings of the
We began by trying to create an effective mapping for point- SIGCHI conference on Human factors in computing systems, pages 286–
and-click applications. Tackling other tasks is left as future 293, New York, NY, USA, 1999. ACM.
491
A New Generation of
Open Source Data Collection Tools
Yaw Anokwa, Carl Hartung, Adam Lerer, Brian DeRenzi, Gaetano Borriello
Abstract—Organizations in low income regions need tools JavaRosa and AndroidRosa. We will also demonstrate Server-
for collecting and reporting on data. Existing paper-based ap- Rosa, software which processes and hosts data submitted by
proaches are often slow and incomplete when compared to data mobile phones.
collection tools on mobile devices. In this demonstration, we
present the authors’ Android-based system in addition to broader To demonstrate JavaRosa, we will use a few examples of
work by the OpenRosa Consortium – a community dedicated phones which support Java 2 Platform Micro Edition (J2ME).
to building open source, standards-based tools for mobile data The Nokia 3110c and 6085 both sell for around $100 and are
collection, aggregation, analysis, and reporting. widely available in low income regions. The Nokia N95 is
Index Terms—data collection, mobile phones, ICT available for around $400 and is one of the most advanced
smartphones available. We will demonstrate form filling and
submission functionality designed for community health care
I. I NTRODUCTION workers on these phones.
variety of tools are urgently needed to address the lack AndroidRosa will be demonstrated on the T-Mobile G1
A of data in low-income countries. This data is required for
decision making and research in many sectors. Due to recent
phone running Android. Android is an open source operat-
ing system developed by Google that is more feature-rich
technological advancements, there is great excitement for and easier to develop for than J2ME. AndroidRosa uses the
using mobile devices to address current gaps in information. JavaRosa core code and leverages Android’s unique abilities.
Phones and PDAs have proven to be dramatically faster and We will demonstrate form filling which builds on the G1’s
more complete than traditional methods of pen and paper, and input devices (touchscreen and keyboard) and sensors (SMS,
can be more accurate and less expensive as well[5], [6]. location and images).
The OpenRosa Consortium[1] is a group working together To collect the completed form data sent from the phones,
to foster open source, standards-based tools for mobile data we will demonstrate ServerRosa. This server component is
collection, aggregation, analysis, and reporting. Participants in- written in Java and can run on local servers as well as in
clude AED-Satellife, Cell-Life, CIDRZ, D-Tree International, the cloud. Phones can submit completed form data to the
DataDyne, Dimagi, Google, MRC-SA, University of Bergen, server, which can export submitted data and generate reports
Makerere University, and the University of Washington. The if necessary. The server can also summarize and aggregate
group has active developers in Bangladesh, Kenya, India, submission data and send it to a separate Android application,
Norway, Pakistan, South Africa, Tanzania, Uganda, and the which supervisors can use to monitor data collection.
United States primarily working on JavaRosa – an open source
framework for data collection on Java-enabled phones. III. C ONCLUSION
By developing open source solutions and conforming to The OpenRosa Consortium is close to consolidating the
standards based on the W3C XForms specification, OpenRosa necessary functionality for the first public release of JavaRosa,
projects can interoperate their components. So while many AndroidRosa and ServerRosa. Together, we are working to-
of the member organizations are currently using JavaRosa wards solutions to allow organizations with minimal technical
for form collection and management systems, others are capacity to collect data using a variety of mobile phones.
building complex solutions like clinical trial software[4], data
reporting on servers[3], and even community health worker ACKNOWLEDGMENT
management[2]. The authors thank all members of OpenRosa, especially
Neal Lesh, Drew Roos and Clayton Sims.
II. D EMONSTRATION
R EFERENCES
In this demonstration, we will present some of the data
collection and management tools under development by the [1] OpenRosa, http://openrosa.org
[2] CommCare, http://commcare.cs.washington.edu
authors for the OpenRosa Consortium. We will show form [3] Gather, http://blog.gatherdata.org
filling, submission and monitoring on mobile phones using [4] OMEVAC, http://epihandy.org/index.php/OMEVAC
[5] K. Shimira et al., The use of personal digital assistants for data entry at
Yaw, Carl, Brian, and Gaetano are affiliated with the University of Wash- the point of collection in a large household survey in southern Tanzania,
ington and Adam is affiliated with the Massachusetts Institute of Technology. Emerging Themes in Epidemiology, 2007.
This work was done while Yaw, Carl, Adam and Gaetano were interns at [6] T. Parikh, et al., Mobile Phones and Paper Documents: Evaluating a New
Google. All authors can be reached at yanokwa@cs.washington.edu. Approach for Capturing Microfinance Data in Rural India, CHI, 2006.
494
1
Sai Gopal Thota(thota_gopal@daiict.ac.in), Rabin Patra(rkpatra@berkeley.edu) , Murali Medisetty(murali_krishna_2006@daiict.ac.in), Sivananda
Reddy(sivananda_reddy@daiict.ac.in), Vivek Munagala(pavan_v_munagala@daiict.ac.in), Joyojeet Pal(joyojeet@washington.edu)
495
Author Index