Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Information Overload:

Addressing the Productivity Paradox in


Face-to-Face Electronic Meetings
MARY-LIZ GRISE AND R. BRENT GALLUPE

MARY-LIZ GRIS^ is an Assistant Professor of Information Systems at the School of


Business Administration at Dalhousie University in Halifax, Nova Scotia. She re-
ceived her Ph.D. in management (management information systems/organizational
behavior) from Queen's University in 1997. Her current research interests are com-
puter support for groups and teams, human information processing, and mental
workload. Her work has been published in the Journal of Applied Psychology.

R. BRENTGALLUPEisaProfessorof Information Systems and Director of the Queen's


Executive Decision Center at the School of Business, Queen's University, Kingston,
Ontario. Dr. Gallupe also holds an ongoing Visiting Professor appointment at the Uni-
versity of Auckland, New 2^aland. He received his Ph.D. in MIS from the University of
Minnesota. His current research interests are computer support for groups and teams,
evaluation of information systems, and the history of information systems. His work has
been published in journals such as Academy of Management Journal, Sloan Manage-
ment Review, Journal of Applied Psychology, and Management Science.

ABSTRACT: The Electronic Brainstorming System (EBS) is a group support system


(GSS) tool considered particularly productive in supporting idea generation. Unfor-
tunately, as computer-supported groups are confronted with larger numbers of ideas
and supporting comments to organize and evaluate, they may experience informa-
tion overload. This study explores the problem of information overload within the
context of an idea-organization task in a face-to-face electronic meeting. Integrative
Complexity Theory provides the primary theoretical foundation, and an Information
Overload Model for GSS is introduced. Result.s from a laboratory experiment provide
support for the idea that effective GSS tools can be designed based on a theoretical
understanding of information processing, in particular, how information is processed
under conditions of high information load. Use of a GSS toot designed to regulate the
flow of information, called an Idea Regulator, led subjects to organize ideas with
higher levels of complexity, but they also reported higher levels of mental workload.
This research suggests that particular attention to the problem of information over-
load, and research focused on finding theory-based solutions, can lead to more effec-
tive meetings.

KEY WORDS AND PHRASES: electronic brainstorming, group support systems, information
overload, Integrative Complexity Theory.

Journal of Managemtnt Informution Systems / Wiaer 1999-2000. Vol. 16. No. 3. pp. 157-185
® 2000 M.E. Shan«, Inc.
0742-1222 / 2000 S9.50 + 0.00.
158 MARY-LIZ GRISt AND R. BRENT GALLUPE

INCREASING NUMBERS OF ORGANIZATIONS ARE USING COMPUTERS TO SUPPORT face-


to-face meetings. However, the rate of adoption of tools such as group support sys-
tems (GSS), group decision support systems (GDSS), or electronic meeting systems
(EMS)to facilitate these meetings has decreased.' One of the principal reasons for this
reduced adoption rate is that we still do not know how to use these tools properly for
most group meeting tasks. It seems that we have ieamed how to use some GSS tools
but not others. This is a significant problem for organizations that need to improve
the productivity of their organizational meetings. If these tools cannot effectively
support group activities in face-to-face meetings, they will fall into disuse.
The Electronic Brainstorming System (EBS) is one class of GSS tool that has been
particularly productive in supporting the idea generation stage of meetings [22].
Fueled by Osbom's [44] notion that groups of people generating ideas benefit from
synergistic effects, brainstorming is a popular meeting technique. When an EBS
supports brainstorming, synergy seems to occtir. A substantial amount of researcb in
recent years has demonstrated that groups supported by EBS tools generate more
ideas than traditional unsupported groups in the same amount of time [12]. In an EBS
session, GSS groups can generate 50 percent more ideas in 60 percent less time [ 11.
The problem arising from the use of such a productive tool occurs after the idea-
generation phase, when these ideas and concepts must be organized into meaningful
categories. Redundant ideas must be identified and removed, and the complexity of
the data needs to be simplified by categorizing and prioritizing. The very success of
GSS in supporting idea generation makes this a critical issue. As computer-supported
groups are confronted with large numbers of ideas and supporting comments that
require some form of meaningful organization, they may experience information
overload. Groups that are extremely productive in the idea-generation phase of an
electronic meeting may find themselves bogged down by an overwhelming volume
of ideas and comments to organize [22, 31]. These ideas may vary in quality and
relevance, and may reflect many different points of view. The task of organizing them
to the mutual satisfaction of all group members becomes very difficult. Groups in this
overload condition may end up taking unnecessary risks [50] by accepting impracti-
cal ideas, making errors in interpretation, or ignoring important ideas. Progress may
even grind to a halt from frustration and confusion on the part of group members [36].
Group support systems can be used to support idea organization. Researchers are
now recognizing that the task of improving GSS support for this purpose is an impor-
tant item on the GSS research agenda [4,6,7]. It is our belief, however, that real gains
in GSS effectiveness will be made only if the underlying problem of information
overload in group meetings is studied and managed. The main objective of GSS is to
provide groups with communication and in formation-processing support to offset
the limitations inherent in group communication and human information processing
[13]. Information overload fueled by the increased communication capabilities of
computers and accelerated by people's limited information-processing capabilities
is a problem that warrants theoretical and empirical consideration within GSS re-
search. This problem provides the focus for this research.
INFORMATION OVERLOAD 159

We begin by reviewing recent literature on the topic of information overload in


electronic meetings. We then develop a theoretical model that uses Integrative Com-
plexity Theory to outline the key information overload factors. Finally, we describe
an experiment that uses a GSS tool based on the theoretical mode! (the Idea Regula-
tor) to assess its impact on the information overload problem.

Background
WITHIN THE GSS LtTERATURE, TWO STREAMS—COGNITIVE RESEARCH and facilita-
tion—are key to understanding information overload. Given that information over-
load is a cognitive concept, the former approach is of particular relevance in this field
of study. The facilitation school of thought holds that the techniques adopted and the
tools used by a facilitator affect the extent to which information overload is experi-
enced within a face-to-face meeting.

Cognitive Research in GSS


From the perspective of general systems theory, information overload results from the
inability of living systems to process excessive amounts of information [36]. Within
Simon's framework [49], information overload may result from the interaction of
high information loads, high task complexity [45], and the limitations of the human
information processor. As the complexity of the task or information load increases,
the human information processor tries to reduce cognitive effort by changing to a
more effective information-processing strategy [39]. People try to minimize the ef-
fects of information overload hy employing conscious or even unconscious strate-
gies to reduce information load 19].
Specifically, within the field of GSS, there is some recent research on decision-
making tasks. In a decision task requiring processes of information exchange, infor-
mation use, and information recall. GSS groups that had more information to consider
had less desirable decision outcomes [11]. Participants may have suffered from infor-
mation overload and an inability to focus their analysis. These researchers call for
further studies in which information load is varied to determine if GSS use helps or
hinders in situations of high and low information load. There is a need to identify and
better understand those group processes that can minimize overload and enable
better integration of information.
The majority of cognitive research has focused on the task of idea generation, with
particular emphasis on two outcomes: idea quantity and idea quality. With respect to
idea quantity, the question posed in this area of research is "How can electronic
brainstorming systems (EBS) enhance productivity?" In terms of quality, the ques-
tion is. "How can EBS enhance creativity?"
An EBS can support the generation of a large quantity of ideas. The reason appears
to be that an EBS allows simultaneous, immediate entry and storage of ideas, and
enables people to enter ideas as they think of them, without having to wait their tum
[21,24,25]. People are then less likely to forget or suppress ideas that do not fit into
160 MARY-LIZ GRISE AND R. BRENT GALLUPE

the general trend of the discussion. Through simultaneous entry and storage, the EBS
acts as an aid to the human information processor. The ability of an EBS to store all
the ideas entered alleviates the need for the human information processor to maintain
the ideas in memory, and allows additional ideas to be processed [381. Thus, EBS may
be used to overcome or complement the structural limitations of the human informa-
tion-processing system.
An EBS may also affect quality by enhancing creativity. The ability of the EBS to
store ideas indefinitely may promote the generation of more paradigm-modifying
ideas because It provides the ability to access a large pool of previously generated
ideas 137]. On the other hand, simultaneity or parallel entry couid encourage clusters
of paradigm-preserving ideas to be generated, since the EBS makes it possible for a
person to generate a stream of ideas within one paradigm with only intermittent
exposure to the ideas of others [37]. Massey and Clapper [34] call the initial process of
developing a set of ideas "sensemaking"[161. When comparing the sets of ideas gener-
ated during a GSS session to the set of ideas generated individually prior to the session,
they found that when GSS groups were compared to non-GSS groups, GSS groups
generated more ideas or elements that were not in the pregroup collective pool of ideas.
The primary focus of cognitive research in GSS has been on idea generation. How-
ever, the concept of the idea-organization task in this context remains largely unex-
plored. Eor example, there is a paucity of research on the related cognitive processes
and strategies adopted by the human information processor to find and remove re-
dundant ideas, look for similarities and differences between ideas, and distribute the
ideas into categories.

Facilitation
Meeting facilitation appears to affect how people perform their tasks in groups [81. A
meeting facilitator is defined as "someone from outside the group who is trained in
skills used to assist the group interaction while remaining neutral as to the content of
discussions" [1]. An active facilitator is characterized as someone who provides a
flexible response to evolving group needs [2]. Distinctions have also been made
among three types of facilitation in GSS: (1) in chauffeured facilitation, a person
simply provides systems support; (2) in content facilitation, the individual is knowl-
edgeable about the content of the meeting [30]; and (3) in process facilitation, a
person manages the meeting process. Clawson et al. [8| describe the three key dimen-
sions of the facilitator's role: (1) to plan and design the meeting process, (2) to listen
to, clarify, and integrate information, and (3) to demonstrate flexibility. Elexibility
seems to be a key factor [8, 15, 40], not so much in a high-restrictiveness task 148]
such as idea generation, but certainly in a task such as idea organization which has
lower restrictiveness. It has been found that when facilitation is not provided, a GSS
is less likely to be appropriated [ 14] faithfully. Anson, Bostrom, and Wynne [ 1 ] found
that in an idea-organization task, all twelve facilitated groups used the organization
tool appropriately, but in the nonfacilitated groups, only seven out of twelve did,
with five groups discarding the tool altogether. In addition, only three of the
INFORMATION OVERLOAD 161
I

nonfacilitated groups fully organized their lists, compared with eleven of the twelve
facilitated groups. However, even when a facilitator is present, an inappropriate match
between process and GSS tool can lead to information overload. In one case, a facili-
tator described a group that generated 200 comments in 20 minutes. Although sum-
marizing the key points took only another 20 minutes, the process of analyzing and
rephrasing each comment required 21 hours, and led to information overload [32].

Information Overload in GSS Literature


Information overload has been identified as a general problem in the IMIS field. One
potential solution for overcoming the problem is to provide features that filter and
structure communication [27], a solution that can also be applied to group support
systems. As the problem of information overload in face-to-face computer-supported
meetings becomes more apparent, researchers are becoming more interested in under-
standing the underlying causes of overload, and in finding GSS-based solutions.
While some seek to identify the source of high productivity in electronic idea gen-
eration [ 12] others are concerned with the dysfunctional aspects of that high produc-
tivity in the subsequent idea-organization task. The GSS research community has
recognized that information overload can be a problem when many ideas need to be
organized [2]. McGrath and Hollingshead [35] define information overload as hav-
ing too many things to do at once, and stress that GSS research should pay more
attention to such temporal features. In the research stream that focuses on electronic
brainstorming, it is generally agreed that the high productivity that results from the
use of electronic brainstorming tools is due in large part to the ability of GSS to
support parallel and immediate entry of ideas [24. 38, 41 ]. Gallupe and Cooper [22]
acknow[edge that overload can be a significant problem in meetings when a large
number of ideas need to be organized and evaluated. In a field study involving
twenty-six facilitators, Kelly and Bostrom [31] note that the occurrence of informa-
tion overload due to simultaneous input was frequently identified.
Other solutions to the information overload problem are summarized in Table 1.
Some of these reduce overload by restricting idea generation, others by changing the
type of support provided in the organization stage. Restricting idea generation can
prevent information overload by reducing the demand on the human information
processor and allowing the participant to focus less on "holding on" to all of the
Ideas, and to focus more on understanding and organizing them. Other solutions are
proposed for those situations where ideas are not restricted in the generation stage.
There are three such approaches: (I) making the participants' taskof organizing the
ideas easier, for example with a "drag and drop" interface [7], (2) automatically
organizing the ideas, for example, using the concept classification tool with auto-
matic indexing and cluster analysis [6]; and (3) providing greater facilitator support,
such as with Eden and Ackermann's [ 18] "ideas management" tool.
The three areas of literature—cognitive research, facilitation, and information over-
load—have common features. When dealing one is specifically with the task of
organizing ideas, the goal is to make the job easier for group members. The next
162 MARY-LIZ GRIS^ AND R. BRENT GALLUPE

Table 1. Proposed Solutions to the Information Overload Problem in Idea


Organization

GSS feature Characteristics that Reference


help information overload

Group memory Queuing and filtering information [41]


Round robin idea Reduces total number of ideas [21 & personal
generation generated communication]
Drag and drop More intuitive user interface [7]
Chunking Eiectronic sorting of ideas into batches [10,38]
Concept classification Automatically indexes and clusters
ideas into common categories [6]
Cognitive mapping; Facilitator creates a team map [lai
ideas management with drill down features
tool (COPE)

section presents an Information Overload Model for GSS based on Integrative Com-
plexity Theory, which pulls together the concepts of cognitive processing, informa-
tion overload, facilitation techniques, and GSS tool development, all within the
context of the idea-organization task.

An Information Overload Model for GSS


THE INFORMATION OVERLOAD MODEL FOR GSS {FIGURE 1) is an inpnt-process-
output model, which can be used to illustrate some of the problems that occur in GSS
groups. Some concepts, such as environmental variables, information load, and ini-
tial integrative complexity, are derived from Integrative Complexity Theory (ICT),
which is described first. Some of the outcome variables such as task difficulty and
satisfaction are familiar in GSS research. New concepts—process enablers, adjust-
ment Strategies, and output complexity—are introduced in this section.

Integrative Complexity Theory


Integrative Complexity Theory (ICT) [51] considers how people process informa-
tion, the mental structures that aid processing, and the situational characteristics that
influence processing [53]. Integrative complexity is defined as the ability of indi-
viduals to process information using the cognitive processes of differentiation and
integration. Using differentiation, a person generates ideas from several different
dimensions, or points of view. Integration refers to the process of developing com-
plex connections between those dimensions. When a topic is viewed from several
alternative (and perhaps inconsistent) perspectives, the ability to integrate involves
reconceptualizing the problem so that two or more dimensions are combined in a way
that reconciles the inconsistencies between the different perspectives.
Information load is a situational or environmental component that is often used to
INFORMATION OVERLOAD 163

INPUT VARIABLES PROCESS VARIABLES OUTCOME VARIABLES

Gnvironrrwntal VartablBS Individual Cognitive Processes Process


Mealing Tatk Mental Workload
Group ConiBNt Initial IntesiaKvc Complexity
Load Adjustm«nt Detign Task Difficulty
Rat« of Information Flow • »
¥
Understnndlng of Tsak
Stnlcgles
Complexity Acguitmant Disign lr)ttructioni
Information t o a d StrMagiM Mottvation
Topic Domiln
Nurnb«r of ld«n
lil«a DIvaralty PorfOrmanco
Tim* Outptrt Complaxlty
Time to Complete
Process Enablars
Facilitator Tachnlquta Perceptual
OSS Tool* Satlifactlon
Comiort
Advocacy
Ptreaptlon of Work Compfattd

Figure I. An Information Overload Model for GSS

represent the effects of the other environmental variables [33, 36, 43, 46]. In its
broadest sense, information load is the amount of input an information processor
receives within a given period of time. It has been defined in terms of (1) external
stimuli [51], (2) dimensions of information, (3) diversity of information and the
number of alternatives, and (4) the "number of alternatives each unit of information
adds" [47, p. 55].
Integrative complexity is a function of information load. A person who is integratively
complex is not only multidimensional—in being able to switch between alternative
viewpoints—but is also able to reconcile or integrate these different viewpoints.
Differentiation requires the ability to accept the validity of alternate ways of think-
ing. Integration, however, requires the ability to allow these other perspectives to
change one's own understanding of an issue. The degree to which a person can apply
differentiation and/or integration depends on the level of information load. This rela-
tionship is usually characterized graphically by an inverted U-shape (see figure 2).

Input Variables
Environmental Variables

A meeting task refers to the type of activity being undertaken in a face-to-face meet-
ing. Two common GSS tasks are idea generation and idea organization. Information
overload problems often result when moving from the former task to the latter. High
productivity in the idea-generation stage can sometimes lead to information over-
load in the idea-organization stage of a meeting. These two meeting tasks differ in terms
of the cognitive processes people use to complete them. Both require differentiative and
integrative cognitive processes, but to different degrees (see figure 3).
Idea generation involves breaking up a problem into several components and de-
veloping many ideas within these components. This is largely a process of differen-
tiation, although some integration may occur since it is often through comparing two
or more ideas to each other that additional ideas emerge. Idea organization involves
arranging the problem's components and ideas in some meaningful way, and remov-
ing redundant and irrelevant ideas. This procedure uses integration, although if
164 MARY-LIZ GRIS^ AND R. BRENT GALLUPE

High Output more complex

Integrative Output less Output less complex


Complexity complex Reduce information
Add information

Low

Low High
Information Load

Figure 2. The Information Load-Integral ive Complexity Relationship (adapted from |53])

encouraged, some differentiation will also occur, leading to additional ideas and new
dimensions. For both tasks, the integrative complexity of group members affects how
well the tasks are carried out.
Group context factors used in GSS research include composition variables such as
heterogeneity (member status, subject type, history), member characteristics (atti-
tudes, values, power, personal beliefs, age, sex, preferences, self confidence, experi-
ence); initial levels of cohesiveness, task understanding, and consensus; and leadership
characteristics (such as style, attitude, power, and organizational position) |20]. Vari-
ables such as group size can affect the productivity of electronic brainstorming [25],
and group heterogeneity may increase information overload because a diverse group
may generate more ideas.
While group context variables are always important to GSS research, in idea orga-
nization, the initial stage of putting ideas into categories is often an individual
exercise. The group has generally come from an idea-generation session, which is an
individual pooled exercise [7]. It then moves into the organization stage, which is
generally a combination of individual and group processes. Finally, the group par-
ticipates in the choice stage, where ideas are ranked and/or voted on; this is also a
stage where further group discussion and negotiation take place. In the individual
activity of organizing ideas, the effect of group-level variables is less infiuential, but
as part of the larger meeting process, group context must be acknowledged. Experi-
mentally, maintaining a consistent group size can control the effect of group context.
In idea generation, ideas do not flow into group memory at regulated rates. People
enter ideas as they think of them and look at the ideas other people generate accord-
ing to their needs and at their own pace, thus achieving a self-regulated rate of
information flow. In contrast, in idea organization it is often the case that the group is
presented with a list of previously generated ideas. The presentation of many ideas at
once introduces stress to the system. Streufert and Swezey [50] hypothesize, in the
context of their research, that the optimal fiow of information is one load item of
INFORMATION OVERLOAD 165

High

Idea • Idea
Levels of organization A generation
Integrated
Information
Processing

Low

Low High
Information Load
Figure 3. The Optimal Load for an Idea-Generation Task Becomes Overload for an
Idea-Generation Task

information every three minutes, where a load item is defmed as subject-predicate-


object statement of single meaning on any one content dimension (p. 83). This would
suggest that an individual could effectively integrate only ten ideas every half-hour.
Given the high productivity of electronic brainstorming, it may be more than a
problem of too many ideas being generated; additional problems may occur because
the ideas come too quickly in the idea-generation phase, and too many are presented
at once in the idea-organization phase. A GSS Regulator tool could potentially regu-
late the rate of information flow, thus reducing the possibility of overload.

information Load

In an idea-organization task, information load has four components: task domain, the
number of ideas, idea diversity, and time. Task domain refers to the general problem
defmition or question being addressed. Some domains may trigger higher informa-
tion loads than others. Controversial topics, such as abortion, for example, may
induce higher levels of stress than more neutral topics, such as tourism promotion.
The second component is the number of ideas. In idea organization, this refers to the
number of ideas requiring organization that is presented to the group or individual
group members. Higher numbers of ideas may lead to higher information loads, The
third component is idea diversity. Some sets of ideas may represent more dimensions
than others [28]. Multiple dimensions or higher idea diversity may be associated
with higher information ]oads [55]. Finally, time has a significant impact on informa-
tion load. While time is considered an environmental constraint in many models, it is
an inherent part of information load and thus is included here, rather than in the
environmental variable section. Given a constant number of ideas to organize, reduc-
ing the amount of time available may increase information load.
166 MARY-LIZ GRISfi AND R. BRENT GALLUPE

Process Enablers

Process enablers are processors that act as "engines" to provide cognitive process
support, or to reduce information load. The two process enablers in this model are
facilitator techniques and GSS tools.
Facilitator techniques consist of the suggestions, instructions, and actions used by
a human facilitator to support a group in an electronic meeting. Facilitator tech-
niques can be applied to reduce information load or to support cognitive processes.
A facilitator can reduce information load in several ways. Prior to presenting the
ideas to the group, the facilitator can decrease the number of ideas to be organized by
removing redundant concepts, or by reorganizing ideas into similar groupings. An-
other means of reducing load is to increase the amount of time available for the task,
if the group is not finished and it appears that the group is not yet experiencing
information overload. A third load reduction technique h filtering: The facilitator
can choose with or without group participation, a subset of information on which to
concentrate (for example, the first twenty ideas or all ideas with the word "policy").
The concept of multiple channels [36] can also be applied by distributing the set of
ideas among group members, thus giving each person a lower information load.
Facilitators can also enable cognitive processes. Differentiation and integration
can be supported by suggesting categories, or by providing overall problem frame-
works. Predetermining the different dimensions ofa problem and showing how they
link together can provide individuals with the support they need to consider their
own integrative frameworks. For example, the facilitator could trigger differentiation
by putting two ideas on an overhead screen and asking "How are these ideas differ-
ent?" and conversely, couJd trigger integration by asking, "How are these ideas
similar?" Chunking is supported when the facilitator labels a group of ideas with a
single category name—and then refers to those ideas by that single name, rather than
by the individual ideas within the chunk. Errors may be circumvented when a facili-
tator encourages discussion of ideas and clarifies concepts. Finally, approximation
may be facilitated in several ways—by encouraging group members to complete one
group of ideas and move on to the next, by considering similar ideas (as well as
identical ideas) to be redundant, or by reducing the stringency of criteria for what
ideas would belong in which categories, in other words, "making do."
Integrative Complexity Theory (ICT) gives the GSS researcher an opportunity to
develop ICT-based GSS tools. An ICT-based GSS tool based on the Information Over-
load Model for GSS supports idea organization by either reducing information load or
enabling cognitive processes. For example, one problem encountered in an idea-orga-
nizing task is that the ideas generated in the previous generation task are frequently
presented to the group all at once. Beginning with a high load may lead a person to use
a load adjustment strategy, such as ignoring or filtering the ideas. As a result, some ideas
may be ignored from the start. Not only is the level of complexity affected, but also the
group pr(xluct may suffer if potentially important ideas are immediately excluded.
An Idea Regulator tool is an ICT-based GSS too! that evens out the presentation of
ideas to group members. Presenting ideas in groups of five at a time facilitates
INFORMATION OVERLOAD 167

Human Information Processor


Process Enablcr
High

Integrative
Complexity

Human Information
Processor
Low
Low High
Information Load

Figure 4. Effect of Process Enablers on Integrative Complexity Curve

chunking and integration strategies by giving the impression that a relationship may
exist among those five ideas. Chunking and integrating reduces information load,
which then makes it easier for group members to process the next group of ideas. The
Idea Regulator is facilitator-paced. The facilitator controls the interval of time al-
lowed between presentation of groups of ideas. Groups of ideas are presented at
preset intervals, ensuring that all ideas are reviewed before the end of the session.
Other GSS tools can act d& process enablers. One of these, a Dynamic Dimensiona-
lizer, can enable cognitive processes by acting as a clustering tool to sort ideas on key
dimensions [5]. This tool supports a differentiation strategy by giving group mem-
bers the ability to "drag and drop" ideas into separate category "buckets." In addi-
tion, it provides a "drill down" feature into each category.
Another example of a GSS tool that supports the integration strategy is the Idea
Integrator. This is a cognitive mapping tool [17, 28] which, when combined with the
Dynamic Dimensionalizer, allows individual group members to draw the graphical
links between category buckets, thus supporting the development of integrative
frameworks. The effect of process enablers—facilitator techniques and GSS tools^is
to enable individuals to process more information at higher levels of integrative
complexity (see figure 4).

Process Variables

Individual Cognitive Processes

When group members in a face-to-face electronic meeting are given the task of orga-
nizing ideas, cognitive processes are the main "engines" used to achieve the task.
168 MARY-LIZ GRISfi AND R. BRENT GALLUPE

High B
Increased Complexity
Complexity Adjustment
Design
Strategies
Integrative
Complexity

Design
Sdategies
Low
Low High

Figure 5, Effect of Design Strategies on the Integrative Complexity Curve

There are many different kinds of cognitive processes, but the framework used in this
paper is Simon's Information Processing Model [49]. This model is concerned with
deci sion-making processes in the design stage and uses an in formation-processing
perspective. Design strategies are cognitive processes and techniques used during
the design stage of decision making. They help provide structure for the ideas gener-
ated in the previous information-gathering stage, and facilitate effective results in
the subsequent choice stage.
Design strategies may be conscious or unconscious processes. From within the ICT
theoretical framework, an individual begins the design task at some point on the IC
(integrative complexity) curve, dependent on the initial information load, for ex-
ample, point A in figure 5. Design adjustments occur when the individual attempts to
move along the curve towards the optimal point (point B). In this example, the
individual is attempting to avoid the effects of information overload by either de-
creasing load or increasing complexity, so as to move to point B. Using Miller's [36]
typology, some design strategies are listed in Table 2.

Outcome Variables and Propositions


A study of information overload, in the specific context of idea organization, requires
the development and use of some measures not previously used in a GSS study. One
of these, the mental workload measure (MWL) is a perceptual or subjective evalua-
tion. The other, outcome complexity, is a behavioral or objective measure.

Mental Workload

People are usually aware of the fact that they are working hard. Mental workload is a
perceptual construct that attempts to measure this sense of "working hard." When
INFORMATION OVERLOAD 169

Table 2. Design Strategies

Load adjustment design strategies Omitting


Queuing
Filtering
Complexity adjustment design strategies Approximating
Differentiating
Making errors
Integrating
Chunking

high demands are placed on information-processing capacity, individuals report high


mental workload (MWL) levels [18]. When comparing one task to another, they are
able to judge which of the two has the higher mental workload. Mental workload is
defined as the degree of processing capacity that is expended during task perfor-
mance [18] and is affected by the components of information load, namely, task
domain, the number of ideas, idea diversity, and time.

Output Complexity

Integrative complexity and information overload cannot be measured directly but


must be inferred from performance. This is often done by measuring the complexity
of output. For example, measuring the complexity of historical writings provides an
indication of the integrative ability of military leaders [53]. In the context of idea
organization, output complexity is broadly defined as the complexity of the fmished
output of organized ideas. For example, when the objective of idea organization is to
take a set of ideas, move or copy them into meaningful categories, label those catego-
ries, and use the ideas in more than one category, output has higher complexity if one
or more of the following has been achieved: (I) all of the ideas are categorized and
none is left over; (2) more, rather than fewer, categories are created; (3) ideas are
repeated in more than one category. Some people will do one of these, and some will
do all three, but they are independent items that do not necessarily correlate. By
combining the three components, one can gain an overall measure of output com-
plexity that captures the different ways in which people can demonstrate complexity
when organizing ideas. Output complexity is therefore defined in this context as the
sum of these three characteristics, namely, the number of ideas organized plus the num-
ber of categories created, plus the number of ideas repeated in separate categories.
As with mental workload, output complexity may be affected by information load,
that is, by task domain, number of ideas, idea diversity, and time. Information over-
load, which can result from the imposition of high information loads on individuals,
usually results in performance changes. It is common in the mental workload litera-
ture to find cases where task performance will stay constant or increase to the point of
capacity, after which it drops [18]. In the electronic group context, this could mean
lower levels of output complexity (see figure 6).
170 MARY-LIZ GRIS^ AND R. BRENT GALLUPE

High
Mental Workload

Integrative Complexity

Low
Low High
Information Load

Figure 6. Relationship Between Mental Workload and Integrative Complexity

Other Outcome Variables

Other outcome variables appropriate to a GSS study of information overload include


satisfaction, difficulty, comfort, level of understanding, and motivation. These are
used to help explain some of the main statistical results for mental workload and
output complexity.

Research Hypotheses
FROM THE INFORMATION OVERLOAD MODEL FOR GSS, THE INPUT VARIABLES that
seem most likely to affect outcome variables are the number of ideas, the task do-
main, and the GSS tool. Three main hypotheses are drawn and tested based on the
model.

HI: The number of ideas to be organized (information load) will affect informa-
tion overload outcomes.

H2: Task domain will affect information overload outcomes


H3: Use ofa GSS tool will affect information overload outcomes.
The main information overload outcomes are mental workload and output complex-
ity; therefore each of the above hypotheses can be expressed in terms of these two
variables.

HIa: The number of ideas to be organized will have a significant main effect on
perception of workload—that is, individuals in the high-load conditions will
INFORMATION OVERLOAD 171

report higher levels of mental workload than will individuals in the low-load
conditions.
Perception of mental workload seems to have a linear relationship to information
load [18]. This means that, as load increases, people will report higher levels of
mental workload. It is harder to organize many ideas than it is lo organize a few ideas.
The more ideas that must be organized, the higher the mental workload reported.
Hlb: The number of ideas to be organized will have a significant main effect on
levels of output complexity—that is. individuals in the high-load conditions will
have lower levels of output complexity than individuals in the low-load condi-
tions.
When individuals experience information overload, the levels of integrative com-
plexity decrease. The group members will organize the ideas, but they will do so with
less complexity, which could mean that relatively fewer categories will be created, or
fewer ideas will be repeated in more than one category, or relatively fewer ideas will
be organized.
H2a: Task domain will have a significant effect on mental workload—that is. in
different task domains, individuals will report different levels of mental workload.
If there are any differences between tasks, this will be due to either differences in
topic domain or idea diversity, both of which are aspects of information load. In this
study, the ideas that were organized in the task—the cheating task and the school
task—were chosen because they are both within domains that are familiar to the
subject population. Differences may exist in the complexity of the specific domains.
For example, the cheating task may seem more relevant and more controversial to
students in the short term than the implications of the unlikely event of extending the
school year. It is also possible that the ideas themselves vary in complexity {or
diversity). This could result in higher levels of mental workload.
H2b: Task domain will have a significant effect on output complexity—that is,
individuals in one task domain will report higher levels of output complexity
than individuals in the other task domain.
If there are any differences between the tasks in terms of topic domain or idea
diversity, this may affect the ability of subjects to organize the ideas. For example, a
familiar topic may lead a subject to create many categories fairly quickly; con-
versely, an unfamiliar topic may lead to fewer categories. A topic with high idea
diversity would also lend itself to the creation of many categories, as opposed to a
topic where ideas seem to cluster into fewer categories.
H3a: The use of a GSS tool that regulates the flow of ideas will have a signifi-
cant effect on mental workload—that is, individuals using cm Idea Regulator
will report lower levels of mental workload than individuals not using an Idea
Regulator

An Idea Regulator tool is an ICT-based GSS tool that evens out the presentation of
172 MARY-LIZ GRISfi AND R. BRENT GALLUPE

ideas to group members. Presenting ideas in groups of five at a time facilitates


chunking and integration strategies hy giving the impression that a relationship may
exist among those five ideas. Chunking and integrating reduces information load,
which makes it easier for group members to process the next group of ideas. Using
this tool should decrease the amount of mental workload reported.
H3b: The use ofa GSS tool that regulates the flow of ideas will have a signifi-
cant effect on output complexity—that is, individuals using an Idea Regulator
will report higher levels of output complexity than individuals not using an
Idea Regulator.

An Idea Regulator tool that facilitates chunking and integration should aid sub-
jects in the task of putting similar ideas together and identifying categories. Subjects
using an Idea Regulator should be able to organize ideas with a higher level of
complexity.

Experimental Design

A 2 X2 X2 (LOAD XTOOL XTASK) NONREPEATED MEASURES FACTORIAL experiment


was conducted at an individual level of analysis within a group context. This means
that individuals organized the ideas separately but with the implied intent that their
categorizations would be combined at the next stage. This process replicates what
our many nonexperimenta! groups had done during idea organization. The purpose
was to compare the mental workload and output complexity of group members under
conditions of high information load versus low information load for two idea-orga-
nization tasks. There were four experimental conditions (load ' tool) and within each
condition, the subjects were given two tasks (load ' tool ' task). Each subject was as-
signed randomly to one of these conditions and stayed in that condition for both tasks.
Pilot tests were conducted over three days during the week prior to the actual
experiment. The purpose of these tests was: (1) to determine what constituted high
load, (2) to build an appropriate set of ideas for the experimental task, (3) to test and
modify the perceptual measures, and (4) to fine-tune the procedures. Eighteen sub-
jects participated in the pilots.
Initially, it was thought that high and low load could be represented by 100 ideas
and 40 ideas, respectively, to be organized in a ten-minute period. In the first set of
pilots, eight subjects were each given 100 ideas, double-spaced in one long list. No
subject was able to organize even 40 ideas, so it seemed that high load could be
represented by 40 ideas. Experimental procedures were recorded during the first
session and gradually refined in subsequent sessions. The postsession questionnaire
and the mental workload instrument were also administered and later refined on the
basis of feedhack from the subjects.
In the second set of pilot studies, twelve tasks were performed by six subjects in two
different groups. In the first group, forty ideas were presented in the Low Regulator
condition. The task was varied, so that both the school and cheating tasks were
tested. In the second group, both task and load were varied. Two subjects were first
INFORMATION OVERLOAD 173

given 15 school ideas in the Low Regulator condition, and then 40 cheating ideas in
the Low Regulator condition. From this test, it was concluded that 15 and 40 ideas
would be used to manipulate low and high load.
The purpose of the third day of pilot-testing was to refine both the measures and the
regulator conditions. Four subjects were each given two tasks and forty ideas in each
task. All three regulator conditions were tested, and procedures recorded. In addition,
the perceptual measures were tested.

Independent Variables
Information load had two levels, low load and high load. Following a practice task,
half of the subjects were given fifteen ideas to organize in ten minutes (low load)
while the other half were given forty ideas to organize in ten minutes (high load).
The GSS tool used consisted of two variations of GroupSystem's Group Writer. The
tool in the "No Regulator" condition was self-paced. The ideas were presented as one
double-spaced list of ideas. Subjects were told they could use cursor keys and page-
down and page-up keys to move through the list at random. They were given len
minutes to organize the entire list. The tool in the "Regulator" condition was facili-
tator-paced and consisted of a list of ideas divided into batches of five ideas at a time,
which could be viewed by pressing the page-down key. This number of ideas seemed
to be a reasonable "chunk" and was visually able to fit on the screen. Within a ten-
minute period, subjects were given 100 seconds per batch in the low-load condition,
and 75 seconds per batch in the high-load condition. At the end of each batch period,
they were told to press the page-down key and to work on the next batch of ideas.
They were allowed to move back to a previous screen, but at no time could they move
down to a new screen unless told to do so.
Each group of subjects completed two tasks. To increase the external validity of the
task and internal control over load (number of ideas generated), the ideas used were
generated and randomly selected from previous idea-generation experiments con-
ducted with the same subject pool. The "cheating" problem ("What can be done to
reduce cheating at our university?") and the "school year" problem ("What would be
the consequences of a longer school year at our university?") had been used before
and the relevance to the subject population previously demonstrated [23].
Task order was varied. The GSS tool and the information load remained constant
for each group across tasks. The different treatments are listed in Table 3,

Measures of the Dependent Variables


AT THE END OF EACH TASK, SUBJECTS COMPLETED THE N A S A - T L X MEASURE (Na-
tional Aeronautics and Space Administration Task Load Index) of overall mental
workload [261. This procedure measures workload on a scale from 0 to 100. Subjects
were presented with six defined sources of mental workload—mental demand, physi-
cal demand, temporal demand, performance, effort, and frustration. These six dimen-
sions were chosen for the NASA-TLX instrument based on extensive multivariate
174 MARY-LIZ GRISE AND R. BRENT GALLUPE

Table 3. Experimental Treatments

Treatment First task Second task

Low-load/No Regulator School year Cheating


Cheating School year
Low-load/Regulator School year Cheating
Cheating School year
High-load/No Regulator School year Cheating
Cheating School year
High-load/Regulator School year Cheating
Cheating School year

analyses in varying task domains and have been found to be sensitive to variations in
levels of workload [44]. Pairwise comparisons for each dimension were also made,
with subjects indicating which dimension in each pair represented the greater source
of workload. These comparisons were used to derive a weighting for each dimension
depending upon the number of times subjects selected a particular dimension as
opposed to one of its five companions. After each condition, subjects scored the task
they had completed on a 0-100 scale for each dimension. These values were recorded
as the raw scores. The weights were used to multiply the raw score for each dimension
to give weighted workload values. The total of each weighted scale was added to-
gether and the sum divided by 15 (the total number of weightings) to give a workload
average for the condition.
While not a true test of test-retest reliability, an approximation was done by testing
the correlation between mental workload (MWL) for the first task and MWL for the
second task. Although the first task consisted of half school task and half cheating
task scores, there were no task order effects and TASK was not found to be significant
for MWL.
Output complexity (OUTCOMP) was used to measure how well subjects organized
the ideas in each task. OUTCOMP was defined as the combination of three types of
output—the number of ideas organized, the number of categories created, and the
number of times ideas were repeated in two or more categories. In the high-load
conditions (40 ideas), OUTCOMP was simply calculated as the sum of ideas orga-
nized, categories created, and ideas repeated. In the low-load conditions (fifteen
ideas), this total was then converted to a relative measure based on forty ideas. For
example, in the low-load condition, one subject might have organized eleven ideas,
created three categories, and repeated one idea. In this case, OUTCOMP is calculated
as (1 l-t-3-t-l)*40/15 = 40. This measure allows for several types of complex output,
such as organizing fewer ideas and creating more categories. Making OUTCOMP a
relative rather than an ahsoiute measure also makes it possible to compare the output
of low-load tasks (fifteen ideas) to the output of high-load tasks (40 ideas).- To test
the reliability of this method of calculating OUTCOMP, the raw data were also trans-
formed to 2-scores, based on the means and standard deviations for low-load and
high-load conditions. This transformation had no effect on the multivariate and
INFORMATION OVERLOAD 175

univariate results, so the relative measure was chosen rather than ::-scores to facilitate
interpretation of results.
A postsession questionnaire was administered to measure other perceptual vari-
ables, such as satisfaction, task difficulty, time to complete, comfort, understanding
of task instructions, advocacy, motivation, and perception of work finished.

Subjects
One hundred and seventy-three undergraduate students participated voluntarily in
this experiment for partial course credit. Subjects were randomly assigned to mixed-
gender groups of six or seven persons. Each subject organized the ideas in the task on
an individual basis, but within each group the subjects were in the same condition,
that is, they were organizing the same number of ideas using the same tool and the
same task. Individuals were told that they were performing an individual task but that
everyone physically present in the group was performing the same task.

Procedures
All experimental sessions were held in the Group Decision Lab of a large university.
Facilitators used a standard script to control for experimenter effects. Groups of sub-
jects were randomly assigned to the facilitators. One of two facilitators ran each
session. Procedures were consistent between conditions except where the particular
treatment condition necessitated special instructions, which were then repeated in all
groups for that treatment.

1. The subjects were welcomed by the facilitator and introduced to other mem-
bers of the group. They were asked to read and fill out a consent form, an
agreement to confidentiality, and a presession questionnaire.
2. Instructions were handed out and read aloud by the facilitator.
3. The facilitator conducted a warmup session to clarify any misunderstandings
about instructions, allow practice with the GSS tool, and administer the first
NASA-TLX measure.
4. The tasks were introduced and undertaken one after the other, with instruc-
tions and procedures varying according to condition.
5. At the end of each task, the subjects were given the Task 1 NASA-TLX instru-
ment and a posttask questionnaire to complete.
After subjects were thanked and dismissed, the facilitator collected and labeled the
questionnaire data generated for each subject. Electronic and paper backups were
made of all electronic data. I ;

Results
DATA WERE COLLECTED FROM ONE HUNDRED AND SEVENTY-TWO SUBJECTS (one
subject had to withdraw for health reasons) using the procedure described in the
176 MARY-LIZ GRISE AND R. BRENT GALLUPE

Table 4. Multivariate Analysis of Variance Results

Effect Pillai's F Noncentrality Power(0.01 level)

Load xtool xtask 0.17 0.353 0,02


Tool X task 0.50 1.002 0.04
Load xtask 0.19 0.387 ao2
Load xtool 1.07 2.138 0.09
Task 6.14"* 12.286 0.73
Tool 23.15*** 46.301 1.00
Load 246.06*** 492.129 tm
*p<0.05; **p<O.OU *** p< 0.005.

previous section. Each subject performed two tasks, so that the study group ulti-
mately generated 344 data points for analysis.
Data analysis was conducted using multivariate (MANOVA) and univariate (ANOVA)
analysis of variance. Testing of the assumptions for MANOVA was conducted and
criteria met. Task order would have been considered if there were a significant inter-
action between task order and the independent variables, LOAD, TOOL, and TASK.
There were no significant interactions at a 0.01 level of significance. Perception of
task difficulty and time to complete, both items in the postsession questionnaire,
were used as manipulation checks for the low- and high-load treatments. Analysis
revealed a significant difference between the two u-eatments for both task difficulty
(F = 19.59, p = 0.000), and time to complete (F = 97.87, p = 0.000), indicating a
successful manipulation. Given the large sample size, univariate and muttivariate
tests were considered significant if alpha (p) was found to be < O.OI. MANOVA tests
revealed main effects of all three independent variables: LOAD (Pillai's F = 246.06,
p = 0.000), TASK (Pillai's F=6.l4,p = 0.002), and TOOL (Pillai's F = 23.15,p =
0.000) (see Table 4). Power analysis also showed that one can have reasonable confi-
dence in these findings, in particular for the main effects of LOAD and TOOL. There
were no significant interaction effects.
To evaluate the impact of each main effect on the individual dependent variables,
ANOVA tests were also conducted. Significant univariate results were considered
only in the case where multivariate results were significant, that is, for LOAD, TOOL,
and TASK.
Univariate significance is suspect when there is a high correlation between depen-
dent variables 155]. Examination of within-cell correlations revealed acceptahle
ranges, indicating the independence of behavior of the dependent variables.
As expected, mental workload was higher in the high-load conditions than in the
low-load conditions (see Table 5 for ANOVA results and Table 6 for treatment means).
This difference was significant at /J = 0.000, supporting hypothesis la. Output com-
plexity was lower in the high-load conditions than in the low-load conditions. The
difference was significant at/? = 0.000. This supported Hypothesis lb. Mental workload
was not significantly different between the two tasks, at/j = 0.022. Hypothesis 2a was
INFORMATION OVERLOAD 177

Table 5. Univariate Analysis of Variance Results

Effect DV F Nonce ntral ity Power


(0.01 level)

Load xtool MWL 0.35 0.354 0.00000


xtask OUTCOMP 0.02 0.019 0.00187
Toolxtask MWL 0.19 0.188 0.00000
OUTCOMP 0.95 0.952 0.05644
Load xtask MWL 0.22 0.225 0.00000
OUTCOMP 0.25 0.248 0.00000
Load xtool MWL 1.64 1.638 0.09720
OUTCOMP 0.94 0.943 0.05567
Task MWL 5.33* 5.331 0.39239
OUTCOMP 9.49*" 9.488 0.68672
Tool MWL 19.32*" 19.319 0.96343
OUTCOMP 17.11*" 17.110 0.93762
Load MWL 102.68*" 102.679 1.00000
OUTCOMP 61.96"* 461.961 1.00000

* p < 0.05; • * p <: 0 . 0 1 ; * * * / ? < 0.005.

not supported. Output complexity was higher in the school task conditions than in
the cheating task conditions. This difference was significant at/? = 0.002. Hypothesis
2b was supported.
There were also significant main univariate effects for the GSS tool condition. Unex-
pectedly, mental workload was higher in the Regulator conditions than in the No Regu-
lator conditions. H3a was not supported. As expected, output complexity was higher
in the Regulator conditions than in the No Regulator conditions, supporting H3b.
The results support some of the stated hypotheses but not others. It was found that
the number of ideas presented to subjects affected both the perception of mental
workload (Hla) and the complexity with which those ideas were organized (Hlb).
With fewer ideas to organize, the suhjects felt that mental workload was lower and the
resulting output was more complex. Relatively more ideas were organized, the ideas
were divided into more categories, and more ideas were repeated.
Mental workload was not affected by task. It made no difference to subjects whether
they were working on a set of ideas about the school year or about cheating, both had
similar mental workloads. On the other hand, output complexity was affected by task.
When the two different sets of ideas were presented to subjects, the output from the
cheating task was less complex than the output from the school task. Mental workload
was affected by the GSS tool used, but not as expected. Using the Idea Regulator
actually increased mental workload. As expected, the Idea Regulator increased out-
put complexity. Regardless of the number of ideas to be organized, or the task as-
signed, group members who used the Regulator tool increased the complexity of
output, but also felt that the work was harder.
178 MARY-LIZ GRISE AND R. BRENT GALLUPE

Table 6. Mental Workload and Output Complexity Means


by Load, Task, and Tool

Condition MWL OUTCOMP

High load 58.2 35.6


Low load 41.7 60.3
School task 48.7 49.0
Cheating task 52.1 45.6
Regulator tool 56.0 49.3
No Regulator tool 47.4 46.3

Discussion
THERE ARE THREE MAIN RESULTS OF THIS EXPERIMENT. First, giving individuals a
large number of ideas to organize affects how they complete the task. Organizing a
large number of ideas results in lower complexity of output but bigher perceived
mental workload. Second, the task domain had an effect on outcomes, since output
from the cheating task was less complex. The third result relates to the Regulator
conditions. When the facilitator controlled the pace at which the subjects moved
from one batch to the next, more ideas were organized, more categories were created,
and more ideas were repeated. However, subjects found using the Regulator tool to be
harder, since mental workload was higher.
Responses to the postsession questionnaire (see Table 7) also support these find-
ings. Individuals in the high-load conditions were less satisfied and were less likely
to advocate tbe process to others. Subjects were less satisfied witb tbeir performance
in tbe cheating task compared with the school task. Finally, subjects using the Regu-
lator tool reported a lower level of comfort with the process than those not using the
Regulator tool.
As expected, individuals had a more difficult time organizing forty Ideas (high
load) than fifteen ideas (low load). It appears tbat subjects in the higb-load condition
had greater difficulty generating categories and slotting ideas into those categories.
It may have been that tbere were more categories possible witb the higb-load condi-
tion (for either task) and subjects were unable to think of additional categories, or to
keep track of those categories when they were organizing ideas.
There are at least two possible explanations for the differences between tbe cheat-
ing and school tasks. The first explanation is problem definition. The wording of the
cheating task and the school task may have been sufficiently dissimilar to trigger
different levels of integrative complexity in tbe subjects. The question for tbe cbeat-
ing task read, "What can be done to reduce cheating at this university?" Tbe question
for the school task read, "What would be the consequences of a longer scbool year at
tbis university?" Many subjects used the word "consequence" in tbeir categoriza-
tions. For example, some of tbe categories were "positive consequences," "negative
consequences," and "social consequences." Using tbis word in the question may
INFORMATION OVERLOAD 179

Table 7. Significant Effects for Postsession Questionnaire Items

Question no. Effect F (or Pillai's F for multivariate)

Satisfaction 1,5 Load 17.207*"


1.5 Task 7.848*"
1 Load 22.10"*
5 Load 33.67*"
1 Task 8.65*"
5 Task 15.68*"
Task difficulty 2.12 Load 19.59*"
2,12 Tool 9.36"*
2 Load 13.82'"
12 Load 34.87*"
2 Tool 18.51"*
Time to complete 3,8 Load 97.87*"
3,8 Tool 5.94*"
3.6 Load xtool 9.14"*
3 Load 129.64*"
e Load 195.34"*
3 Tool 6.86"
8 Tool 11.92*"
3 Load xtool 8.54"*
8 Load xtool 18.03"*
Comfort 4 Tool 14.94"*
Advocacy 7 Load 27.63"*
Perception of work finished 11 Load 47.88"*

•/j<0.05; **p<0.01; •**/?< 0.005.

have provided subjects with an integrative framework to begin with, whereas the
categories for the cheating task were less obvious. A second explanation could be
that the cheating task was more controversial and more real to the subjects than the
school task. Considering implications for extending the school year could have been
considered a frivolous task, since this was unlikely to happen. On the other hand, the
cheating issue is a very real one, and possibly a problem the subjects have had to deal
with in courses they have taken. The more controversial cheating task was far more
likely to evoke moral principles, guilty consciences, and strong opinions. This may
have made the task more difficult; one could argue that it also made the task more
realistic and more likely to engage the subjects' attention. If so, this may have in-
creased integrative complexity, not lessened it.
The results for the Regulator tools are interesting. It appears that information over-
load occurred in the No Regulator high-load condition, since mental workload in-
creased and output complexity dropped. On the other hand, information overload
may not have occurred in the Regulator high-load condition, since both mental
workload and output complexity increased. The Idea Regulator may have helped
180 MARY-LIZ GKlSt AND R. BRENT GALLUPE

High

Output Complexity Curve


(Human Information Processor)
+ Process Enabler)

Output Complexity Curve


(Human Information
Processor)
Low
Low A B High
Information Load

Figure 7. Relationship Between Mental Workload and Integrative Complexity Using Idea
Generator as Process Enabler

prevent information overload for those who used it, so that performance could in-
crease instead of decrease. One might expect that an effective tool would seem easier
to use, leading to lower mental workload. On the other hand, it is possible that using
the tool actually led to additional information load. This would make sense, since
subjects using the unfamiliar tool were obliged to pay attention to and follow the
instructions of the facilitator throughout the experiment. In other words, information
load was more than just the number of ideas to be organized. The intervention of the
facilitator appears to have also increased information load, yet it did not induce
information overload.
The Regulator tool may have actually prevented information overload from occur-
ring by changing the integrative complexity curve, as proposed in figure 6. The
increase in information load may also have led to a movement along the mental
workload slope, as shown in figure 7. The combination of increasing information
load (from point A to B) and raising the curve would lead to an increase in both output
complexity and mental workload {from point C to D). While the proposed theory
supported movement along this curve, the findings suggest that process enablers
may actually move the curve itself.
This study provides support for two aspects of the Information Overload Model for
GSS. First, the experiment supports the oft-repeated caution in the GSS literature that
research tasks can affect experimental outcomes. The number of ideas, the familiarity
of the task domain, the relative diversity of the ideas, idea controversiality—all of
these components of information load affect research outcomes and must be consid-
ered in experimental design.
INFORMATION OVERLOAD lSl

Second, it appears that ICT-based process enablers (facilitation techniques and


GSS tools) do bave an impact on bow individuals handle high loads of information
during idea organization. It seems tbat using tbe Idea Regulator tool prevented infor-
mation overload but increased mental workload by changing tbe integrative com-
plexity curve. Tbis allowed more information to be processed in spite of higher levels
of mental workload.

Limitations and Future Directions

ONE LIMITATION OF THIS STUDY IS THAT IT CONCENTRATES on a very small portion


of tbe meeting process—tbat of idea organization. Tbis limitation seems reasonable
since the main hypotbesis of tbis work is tbat information overload most often occurs
in this stage following tbe bigbly productive stage of idea generation. In addition,
the research was conducted in a face-to-face electronic meeting context, so one can-
not necessarily extend tbe findings to a distributed meeting context.
Another limiting assumption is that the idea-organization task in a meeting begins
with an individual-level organizing task, rather tban witb a group-level collabora-
tive task. In any group, individual cognitive processes occur simultaneously and
interactively with group-level processes of sense making, negotiation, and consen-
sus seeking. Information overload can also occur at both tbe individual level and tbe
group level. Just as one cannot necessarily extend an understanding of individual
cognitive processes to the group level, so, too, one cannot assume that information
overload experienced at the individual level would have the same characteristics (or
even the same definition) as information overload experienced at the group level.
The study of information overioad in face-to-face electronic meetings has just begun.
It seems reasonable in the beginning stages of tbis research to distinguish between
individual and group-level phenomena, control for group-level effects, and study
individual processes within tbe group context. One cannot assume tbat these find-
ings would extend to the group level in a collaborative organization task. This
extension must be left for future research.
As witb many laboratory experiments, anotber limitation lies in the subject ptiol.
The large number of subjects required for this experiment necessitated tbe use of
student subjects. Nevertheless, previous experiments determined tbat tbe particular
tasks cbosen for this research project were relevant to tbe subject group. Certainly,
cbeating is a very real and pressing problem for some students. In general, however,
tbe argument can be made tbat if information overload is as universal a pbenomenon
as Miller [36] suggests, it should surely extend to undergraduate students as well as
to employees in organizations. Future research migbt investigate actual idea-organi-
zation tasks by electronic brainstorming groups in organizations, as opposed to
universities.
Caution must also be used in interpreting the results because of tbe use of tbe new
unvalidated measure, output complexity. This measure is defined as the sum of three
components: (I) the number of ideas organized, (2) tbe number of categories created,
182 MARY-LIZ GRISE AND R. BRENT GALLUPE

and (3) the number of ideas repeated in separate categories. It is assumed that an
increase in any one of these three results in higher complexity of output. They are not
necessarily interrelated, and correlational analysis would not be meaningful. Cau-
tion is advised because this measure does not include an evaluation of quality. Some
of the output could be the result of errors and approximation. Some subjects may
have repeated ideas in more than one category because of a lack of clarity in the
categories created. The actual categories created were not evaluated in terms of qual-
ity and could themselves be subject to redundancy or irrelevance. Further research
could examine the relationship between output complexity and output quality.
A final limitation is that only the Idea Regulator tool was studied. Different results
might be obtained if other theory-based tools were used to support the idea-genera-
tion task. Future research might focus on the development and investigation of the
use of ICT-based process enablers to reduce the problem of information overload in
electronic group processes.

Conclusions
PEOPLE EXPERIENCE INFORMATION OVERLOAD IN ELECTRONIC MEETINGS. This
paper has focused on a particularly troublesome part of electronic meetings—the
idea-organization stage. The research indicates that using theory-based tools may
minimize the information overload problem during the idea-organization phase of
meetings. It also demonstrates that information overload is dependent on not just the
flow of information and ideas, but on the task domain itself.
One of the contributions of this research is the development of an Information
Overload Model and the use of the Mental Workload Instrument. It is hoped that as
information overload continues to pose a pressing problem for individuals in organi-
zations, researchers might be able to enhance and verify the model and use the
Mental Workload Instrument in a greater variety of information system contexts.
The concept of information overload crosses many disciplinary borders. As a prob-
lem, it goes beyond the human being as an information processor and has far-reach-
ing effects on groups, organizations, and society in genera!. Information overload is
a problem that deserves a greater level of research focus in many arenas.
From both a practical and theoretical perspective, the Information Overload Model
for GSS can provide a useful framework for studying the problem of information
overload in face-to-face electronic meetings. As McGrath and Hollingshead [35]
observed, groups using GSS "wi]] continually be on the verge of disastrous informa-
tion overload unless some active measures are taken to fend it off." This issue has
been recognized before, in both theory and practice. It is a problem that affects
organizations where meetings have become a way of life, and information overload
has become a seemingly necessary evil. This need not be the case. Particular atten-
tion to the problem of information overload, and research focused on finding theory-
based solutions, can lead to more effective meetings. In particular, GSS effectiveness
will improve when GSS researchers and practitioners undertake more comprehensive
research on the problem. The Information Overload Model for GSS is a beginning.
INFORMATION OVERLOAD 183

NOTES

Acknowledgments: The authors thank the reviewers and associate editor for helpful comments on
previous versions of this paper. They also acknowledge Ihe generous funding provided by the
Social Sciences atid Humanities Research Council of Canada for this work.
1. An electronic meeting is defined as a computer-supported meeting of organizational mem-
bers who cotnc together at the same time and ihe same place lo accomplish a specific organiza-
lional ohjective. Typically, "computer-supported" means ihat each grtiup member has access to a
networked computer A group support system (GSS) is defined as an "interactive compuier-based
syslcm that combines communication, computer and decision technologies lo support problem
formulation and soluiion in group meetings" 113]. A GSS provides the computer support for an
electronic mecling. In a group support system, each group member has a laptop or a recessed
desktop computer that is networked with the other computers in the rotim, and controlled from a
main facilitator station. Since the computers are connected, each person in the meeting has the
ability to contribute, often anonymously, and to see the contributions of others. GSS provide Ihe
potential for cffeciive electronic collaboration and successful meeting outcomes.
2. Increasing the number of ideas in ihe high-load condition is just one method of manipulat-
ing different levels of load. The extra twenty-five ideas can be considered a means of adding stress
to the human information processor, it is assumed that individuals in the low-load group would
work at the same pace and complexity if they had forty ideas, since they probably did not
experience inlormalion overload. Two other ways of manipulating load would he to decrease the
time available in one condition, or to increase the diversity of the ideas. Both of these would have
Ihe effect of increasing load, but with less experimental control.

REFERENCES

1. Anson. R.; Bostrom, R.; and Wynne. B. An experiment assessing group support system
and facililalor effects on meeting outcomes. Management Science.4l. 2 (1995), 189-208.
2. Boiney. L.G. Reaping Ihe benefits of information technology in organizations: a framework
guiding appropriation of group support systems. Journal of Applied Behavioral Science. 34, 3
(September 1998), 327^346.
3. Bostrom, R.P., and Anson. R. The face to face meeting: a tutorial. In R.P. Bostrom, R.T.
Watson, and S.T Kinney (eds.). Computer Augmented Teamwork: A GuidedTour. New York:
Van Noslrand Reinhold, 1992. pp. 16-23.
4. Briggs, R.O.; Nunamaker, J.F. Jr; and Sprague, R.H.. Jr. 1001 unanswered research
questions in GSS. Journal of Management Information Systems, 14. 3 (Winter 1997-98), 3-21.
5. Chen. H. An algorithmic approach lo building concept space for a scientific community. In
Proceedings of the Twenty-Seventh Annual Hawaii Intemaiional Conference on System Sciences,
vol.4. 1994, pp. 201-210.
6. Chen, H.; Hsu. P.; Orwig. R.; Hoopes. L.; and Nunamaker. J.F. Automatic concept
classification of text from electronic meetings. Communications of the ACM. 37. 10 (1994),
56-73.
7. Chin. W.W.; Hayne. S,; and Licker, P. The group consolidation problem during GSS usage:
a new approach and initial empirical evidence. Journal of Information Technology Management,
i, 2 (1992), 15-27.
8. Clawson. V.R.; Bostrom, R.; and Anson, R. The role of the facilitator in computer sup-
ported meetings. Journal of Smalt Group Research, 24, 4 (November, 1993), 547-565.
9. Cook. G.J. An empirical investigation of information search strategies with implications for
decision support system design. Decision Sciences, 24. 3 (1993), 683-697.
10. Dennis, A.R. Information processing in group decision making: you can lead a group to
information, bul you can't make it think. Working Paper. University of Georgia, 1994.
11. Dennis, A.R.: Hilmer, K.M.; and Taylor, N.J. Information exchange and use in GSS and
verbal group decision making: effects of minority influence. Journal of Management Information
Systems. 14, 3 (Winter 1997-98). 61-88.
12. Dennis. A.R.. and Valacich, J.S. Computer brainstorms: more heads are better than one.
Journal of Applied Psychology, 78,4 (\ 993), 531 -537.
184 MARY-LIZ GRISfi AND R. BRENT GALLUPE

13. DeSanctis, G,, and Gallupe, R.B. A foundation for the study of group decision support
systems. Management Science, 33, 5 (1987), 589-608.
14. DeSanctis, G., and Poole, M.S. Capturing the complexity in advanced technology use:
adaptivestmcturationtheory. Orgoniza(/on5c(e/ice,5.2(May 1994), 121-147.
15. Dickson. G.W.; Partridge. J.L.; and Robinson, LH. Exploring modes of facilitative support
for GDSS technology. MIS Quarterly, 17, 2 (1993), 173-194.
16. Drazin, R.; Glynn, M.A.; and Kazanjian, R.K. Multilevel theorizing about creativity in
organizations: a sensemaking perspective. Academy of Management Review, 24,2 (April 1999),
286-307.
17. Eden, C. Cognitive mapping. European Joumal of Operational Research, id (1988), 1-13.
18. Eden, C , and Ackermann. F. Strategy development and implementation—the role of a
group decision support system. In R.P. Bostrom, R.T. Watson, and S.T Kinney (eds.). Computer
Augmented Teamwork: A Guided Tour. New York; Van Nostrand Reinhold, 1992, pp. 325-343.
19. Eggemeier. F T Properties of workload assessment techniques. In A. Hancock and N.
Meshkati (eds.). Human Mental Workload. Amsterdam: North Holland, Elsevier Science Pub-
lishers, 1988, pp. 41-62.
20. Fjermestad, J., and Hiltz, S.R. An assessment of group support systems experimental
research: methodology and results. Journal of Management Information Systems, 15,3, (1999),
7-150.
21. Gallupe, R.B.; Bastianutti, L.M.; and Cooper, W.H. Unblocking brainstorms. Joumal of
Applied Psychology, 76 (1991), 137-142.
22. Gallupe, R.B., and Cooper, W.H. Brainstorming electronically. Shan Management Review,
35, 1 (1993), 27-36.
23. Gallupe, R.B.. and Cooper, W.H. Idea generation of controversial topics. Unpublished
data. Queen's Universily, 1996.
24. Gallupe, R.B.; Cooper, W.H.; Grise, M.L.; and Bastianulti, L.M. Blocking electronic
brainstorms. Journal ofApplied Psychology, 79, 1 (1994), 77-86.
25. Gallupe, R.B.; Dennis. A.R.; Cooper, W.H.; Valacich, J.S.; BasUanutti L.M.; and Nunamaker.
J.K, Jr. Electronic brainstorming and group size. Academy of Management Joumal, 3S,2{\ 992),
350-369.
26. Hart, S.G., and Staveland, L.E. Development of NASA-TLX (Task Load Index): results of
empirical and (heoretical research. In PA. Hancock and N. Meshkati (eds.). Human Mental
Workload. Amsterdam: North Holland, Elsevier Science Publishers, 1988, pp. 139-183.
27. Hiltz. S.R.. and Turoff, M. StniclurJng computer-mediated communication systems to
avoid information overload. Communications of the ACM, 28,1 (July 1985), 680-689.
28. Huff. A.S. Mapping Strategic Thought. New York: Wiley, 1990.
29. Iselin, E. The impact of information diversity on information overload effects in unstruc-
tured managerial decision making. Journal of Information Science, 15 (1989), 163-173.
30. Jessup, L., and Valacich, J., eds. Group Support Systems: New Perspectives. New York:
Macmillan, 1993.
31. Kelly, G.G., and Bostrom, R.B. Facilitating the socio-emotional dimension in group sup-
port systems environments. In L. Olnan (ed.). Proceedings of the 1995 ACM SIGCPR Confer-
ence, Apri\ 6-^, 1995.
32. Kelly, G.G., and Bostrom. R.B. A facilitator's general model for managing socio-emotional
issues in group support systems meeting environments, Joumal of Management Infomiation
Systems. 14, 3 (Winter 1997-98), 2 3 ^ 4 .
33. Malhoira, N.K. Information load and consumer decision making. Joumal of Consumer
Research. 809S2), 419-430.
34. Massey, A.P. and Clapper. D.L. Element finding: the impact of a group support system on
a crucial phase of sense making. Joumal of Management Infonnation Systems. 11,4 (Spring
1995), 149-176.
35. McGrath, J.E,, and Hoilingshead, A.B. Groups Interacting with Technology: Ideas, Evi-
dence. Issues, and an Agenda. Thousand Oaks, CA: Sage Publications. 1994.
36. Miller, J. Living Systems. New York: Wiley, 1978.
37. Nagasundaram, M., and Bostrom, R. P. The slnicluring of creative processes using GSS:
a framework for research. Joumal of Management Information Systems, 77,3(1995), 89-116.
38. Nagasundaram, M.. and Dennis. A.R. When a group is not a group: the cognitive founda-
INFORMATION OVERLOAD 185

tion of group idea generation. Small Group Research, 24,4 (November 1993), 463-489,
39. Newell, A., and Simon, H.A. Human Problem Solving. Englewood Cliffs, NJ: Prentice-
Hall, 1972.
40. Niederman, F.; Beise, C M . ; and Beranek. P.M. Issues and concerns about
computer-supported meetings: the facilitator's perspective. MIS Quarterly, 20, \ (March 1996),
1-22.
41. Nunamaker, J.F.; Dennis, A.R.; Valacich, J.S.; Vogel, D.R.; and George. J.F Electronic
meeting systems to support group work. Communications of the ACM, 34.7 (July 1991), 40-61.
42. Nygren, T.E. Psychometric properties of subjective workload measurement techniques:
implications for their use in the assessment of perceived mental workload. Human Factors, 33
(1991), 17-33.
43. O'Reilly, C.A, Individuals and information overload in organizations: is more necessarily
better? Academy of Management Journal. 23,4 (1980). 684-696.
44. Osbom, A. Applied Imagination. New York: Scrihner, 1957.
45. Payne. J.W., and Bettman, J.R. Behavioral decision research: a constructive processing
perspective. Annual Review of Psychology, 43 (1992), 87-131.
46. Schick, A.G.; Gordon, L.A.; and Haka, S. Information overload: a temporal approach.
Accounting, Organizations and Society, 15, 3 (1990), 199-220.
47. Schroder, H.M; Driver, J.J.; and Streufert, S. Human Information Processing. New York:
Holt, Rinehart, and Winston. 1967.
48. Silver, M.S. Decision support systems: directed and non-directed change. Information
Systems Research, 1,1 (1990). 47-70.
49. Simon, H. The New Science of Management Decision. New York: Harper and Row, 1960.
50. Stevens, J. Applied Multivariate Statistics for Social Science. Hillsdalc, NJ: Lawrence
Erlbaum. 1986.
51. Streufert, S.. and Streufert, S.C. Behavior in a Complex Environment. Washington, DC:
VH. Winston and Sons, 1978.
52. Streufert, S., and Swezey, R.W. Complexity, Managers, and Organizations. New York:
Academic Press, 1986.
53. Suedfeld. P. APA presidential addresses: the relation of integrative complexity to historical,
professional, and personal factors. Journal of Personality and Social Psychology, 49.6 {\ 985),
1643-1651.
54. Tabachnick, B.G., and Fidell, L.S. Using Multivariate Statistics. New York: Harper and
Row, 1983.
55.Zigurs, I., and Buckland, B.K., A theory of task/technology fit and group support systems
effectiveness. MIS Quarterly, 22,3 (September 1998), 313-334.

You might also like