Methods For IA The Challenges and Opportunities Ah

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 26

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/227285094

Methods for IA: The Challenges and Opportunities Ahead

Article  in  Environmental Modeling & Assessment · August 1998


DOI: 10.1023/A:1019019024003

CITATIONS READS

290 1,556

1 author:

Jan Rotmans
Erasmus University Rotterdam
165 PUBLICATIONS   19,007 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Integrated health modelling View project

PhD research project Beneficial Area Development View project

All content following this page was uploaded by Jan Rotmans on 24 December 2014.

The user has requested enhancement of the downloaded file.


Environmental Modeling and Assessment 3 (1998) 155–179 155

Methods for IA: The challenges and opportunities ahead


Jan Rotmans
ICIS, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands
E-mail: j.rotmans@icis.unimaas.nl

1. Introduction will improve the quality of IA by giving access to prac-


tical knowledge and experience, and to a wider range of
There is increasing recognition and credibility for the perspectives and options.
rapidly evolving field of Integrated Assessment (IA). Within However, the reality is that there are two groups of IA
the setting of the political arena it is accepted that IA can be researchers. First, the group using analytical methods, such
supportive in the long-term policy planning process, while as models. They combine knowledge elements from var-
in the scientific arena more and more scientists do realise ious disciplines in an analytical framework, to assess the
the complementary value of IA research. One of the best socio-economic and environmental consequences of human
indicators for this increased recognition is the establish- activities that affect the environment. This group is be-
ment of the European Forum on Integrated Environmental ing pushed into embracing stakeholder participation, but is
Assessment (EFIEA) by the European Commission DGXII. struggling with this because traditionally it has never been
In spite of this growing appreciation for IA, the method- a major concern for them. Second, there is a group of
ological basis of IA is still narrow, and lags behind the high social scientists, arguing that IA itself is a participatory
expectations from the outside world. Broadening the basis process. They are pushing the analytical group to make
of the methodologies underlying IA should therefore be their tools more relevant to decision makers. On the other
one of the top priorities of the IA community. This paper hand, the analytical group is trying to convince the partic-
deals with some ideas which could form a basis for an IA ipatory group of the value of using analytical, quantitative
research agenda for the next 5–10 years. tools such as models and scenarios.
One of the problems of IA is still the many definitions The complex, intuitive, and value-laden process of IA
and interpretations that circulate (Weyant et al. [118], Rot- operates at a variety of levels and scales, so researchers
mans and Dowlatabadi [100], Parson [85–87], Ravetz [91], cannot address the process by only one, unique approach.
Jaeger et al. [49]). Notwithstanding this diversity, these Multiple diverse approaches are needed, varying from ana-
definitions have two elements in common, i.e., interdis- lytical methods (such as models) to participatory methods
ciplinarity and decision support. These two common el- (such as policy exercises). The divergence of methods em-
ements make Integrated Assessment difficult to plan and ployed arises from the uneven state of scientific knowledge
even harder to conduct. Instead of coming up with another across different problem domains, from the differences in
definition of IA, we simply focus on the above commonal- problem perception, and from the different perspectives of
ities as points of departure for exploring challenges for the the scientists and stakeholders involved in the assessment
future. process.
Thus irrespective of whatever definition is taken, IA can The many methodological approaches used at this time
be described as to address complex issues from an integrated viewpoint are
all relatively immature. The ultimate selection of a specific
a structured process of dealing with complex issues, us-
method for performing an Integrated Assessment depends
ing knowledge from various scientific disciplines and/or
on the context of the assessment, and the requirements of
stakeholders, such that integrated insights are made
the IA clients. The simultaneous use of various methods
available to decision makers.
can improve the quality and adequacy of the Integrated As-
Integrated Assessment is an iterative, continuing process, sessment considerably, enabling analysis and interpretation
where integrated insights from the scientific and stakeholder of the impacts of a wide range of policy options and strate-
community are communicated to the decision-making com- gies.
munity, and experiences and learning effects from decision Integrated Assessment attempts to shed light on com-
makers form one input for scientific and social assessment. plex issues by illuminating different aspects of the issue
Although participation of stakeholders is not a necessary under concern: from causes to impacts, and from options to
prerequisite, the conviction in the IA community grows strategies. IA has been widely applied in the global change
that participation of stakeholders is a vital element in IA. research area. IA emerged as a new field in global change
The engagement of non-scientific knowledge, values and research because the traditional disciplinary approach to
preferences into the IA process through social discourse global change research has been unable to meet two signifi-

 Baltzer Science Publishers BV


156 J. Rotmans / Methods for IA

of the puzzle are fitting together, indicating priorities for


narrower disciplinary research.
Performing IAs has a number of advantages. In general
terms, IA can help to:
• put a complex problem in the broader context of other
problems, by exploring the interrelations of the specific
problem with other issues;
• assess potential response options to complex problems.
This may be, but not necessarily, in the form of cost–
benefit and cost–effectiveness analysis;
• identify, illuminate and clarify the different types and
sources of uncertainties in the cause–effect chain(s) of
a complex problem;
• translate the concept of uncertainty into the concept of
Figure 1. Relevant research areas. risk analysis, to assist in decision making under uncer-
tainty;
cant challenges central to understanding global phenomena. • set priorities for research topics, also by identifying and
The first challenge is the development of an adequate char- prioritising decision-relevant gaps in knowledge.
acterisation of the complex interactions and feedback mech- This paper first discusses a variety of IA methods. For
anisms among the various facets of global change. Such the most important analytical and participatory methods, the
feedbacks and interactions are defined away or treated para- state-of-the-art knowledge, a typology, strengths and weak-
metrically in traditional disciplinary research. The second nesses, and current trends are discussed. This forms the
challenge is that of providing support for public decision basis for a discussion of the challenges for the future for
making. IA offers an opportunity to develop a coherent the various IA methods. In addition, over-arching method-
framework for testing the effectiveness of various policy ological challenges are discussed. Then a research agenda
strategies, and estimating trade-offs among different policy for IA methods is presented. The paper ends with recom-
options. mendations how these methodological challenges could be
We should be aware that we are not re-inventing the materialised.
wheel. Integrated Assessment partly overlaps with exist-
ing research areas, especially technology assessment, risk
2. Methods for Integrated Assessment
analysis and policy analysis (see figure 1). These research
areas also address some kind of complex problem, however, In general, two types of Integrated Assessment meth-
from a specific point of view. The essential difference is ods can be distinguished: analytical methods and partici-
that IA aims to integrate knowledge from an a-priori inte- patory methods. While analytical methods are often rooted
grated point of view. Nevertheless, co-operation with the in the natural sciences, participatory methods, also labelled
research communities on technology assessment, risk analy- as interactive, deliberative or communicative methods, stem
sis and policy analysis should be sought in order to benefit from the social sciences. The group of analytical methods is
from gained experiences. reasonably well-defined and basically includes model analy-
From a practical perspective, interlinked social, eco- sis, scenario analysis and risk analysis. Their commonality
nomic, and environmental conditions demand integrated is that they provide analytical frameworks for represent-
policies. Despite an early history of isolated regulatory ing and structuring scientific knowledge in an integrated
initiatives related to for instance air, water, and soil over manner. The group of participatory methods, however, in-
the past two decades, environmental policies have become volves a plethora of methods, varying from expert pan-
increasingly integrated. Nevertheless, this is only a first els, delphi methods, to gaming, policy exercises and focus
step towards further integration of environmental, social groups. They have in common that they aim to involve
and economic policies. non-scientists as stakeholders in the process, where the as-
Opponents of Integrated Assessment argue that Inte- sessment effort is driven by stakeholder–scientist interac-
grated Assessment is premature, and can lead to inappro- tions.
priate confidence in questionable or misleading findings. The aim of the various methods is to facilitate the IA
Proponents of IA, however, argue that the complexity of process as sketched in figure 2. It should be noted that this
the issues demands an integrated approach to ensure that figure is a highly simplified representation of the complex
key interactions, feedbacks and effects are not inadvertently IA process. For example, it does not exclude the possibil-
omitted from the analysis. They argue that the various ity that societal actors also supply (delivering non-scientific
pieces of the complex puzzle can no longer be examined in expertise), and scientists also demand (e.g., in framing re-
isolation. IA endeavours to keep track of how the pieces search agendas).
J. Rotmans / Methods for IA 157

Model typology. Weyant et al. [117] classified IA models


in two categories: (i) policy optimisation models, which try
to optimise key policy variables such as carbon emission
control rates and carbon taxes, given certain policy goals
(such as maximising welfare or minimising costs of cli-
mate policy), and (ii) policy evaluation models, which try
to evaluate the environmental, economic and social con-
sequences of specific policy strategies. However, some
IA models characterised as policy evaluation models can
also be used for optimisation experiments. Therefore, an-
other possible classification of IA models is given by Rot-
mans and Dowlatabadi [100]: (i) macroeconomic-oriented
models, which represent relatively simple, parameterised
decision-analytic formulations of complex problems; and
(ii) biosphere-oriented models, which represent a more
comprehensive, process-oriented description of a complex
problem. Most macroeconomic-oriented models are neo-
Figure 2. Sketch of IA process. classical models based on an equilibrium framework, us-
ing traditional economic concepts regarding optimisation
and capital accumulation, largely ignoring environmental
Below we discuss concisely the analytical and partici-
dynamics. Biosphere-oriented models, however, focus on
patory methods, followed by an exploration of the com-
a systems-based description of the geophysical and bio-
plementary value of the various approaches by sketching a
geochemical processes and feedbacks, but do not ade-
tool kit for Integrated Assessment.
quately represent the socio-economic system. The DICE
model (Nordhaus [79,80]) is a well-known exponent of
2.1. Analytical methods the macroeconomic-oriented school, whereas the IMAGE
model (Rotmans [97], Alcamo [2]) is representative for the
2.1.1. Models biosphere-oriented school.
Integrated Assessment models are computer simulation Meanwhile, some attempts are underway to combine the
(including optimisation) frameworks that try to describe best of both worlds, yielding a hybrid of the two categories
quantitatively as much as possible of the cause–effect re- above. These hybrid models may contain a detailed general
lationships of a specific issue, and of the interlinkages and equilibrium model for the economy that interacts with a dy-
interactions among different issues. Current projects in IA namic environment, such as GCAM (Edmonds et al. [32]),
modeling build on a tradition started in the early seventies and the MIT model (Prinn et al. [90]). However, they
by the Club of Rome (Meadows et al. [71,72]). In this may also use complexity and uncertainty as guiding mod-
context the first global computer simulation model, i.e., the eling principles for both the human and natural system, of
World3 model, was developed, which described couplings which ICAM (Dowlatabadi and Morgan [25]) and TAR-
between the major social and physical components of the GETS (Rotmans and de Vries [99]) are examples. There-
world system. This model inspired the development of fore, the above two classes should be considered as the
numerous global models, focusing on resource depletion, polar ends of a continuum populated by many IA modeling
population and pollution. efforts.
The next generation of IA models explicitly addressed
environmental issues. The first among these models Strengths and weaknesses. Any attempt to fully represent
emerged in the late 1970s from earlier energy modeling a complex issue and its numerous interlinkages with other
(Nordhaus [78], Edmonds and Reilly [33]). Meanwhile, the issues in a quantitative model is doomed to failure. Never-
RAINS model of acidification in Europe was developed in theless, even a simplified but integrated model can provide
the 80s (Alcamo et al. [4]). a useful guide to complex issues and complement highly
The phenomenon of global climate change has prompted detailed models that cover only some parts of complex phe-
the development of a new class of IA models. The first nomena. Among the major strengths of IA models are:
steps to an integrated model of climate change were
taken by Mintzer [75], Lashof and Tirpack [63], and • exploration of interactions and feedbacks: explicit in-
Rotmans [97]. Since then approximately 40 IA mod- clusion of interactions and feedback mechanisms be-
els of climate change have been developed (van der tween subsystems can yield insights that disciplinary
Sluys [106]). Recent overviews of IA modeling activities studies cannot offer;
in the field of climate change can be found in Rotmans • flexible and rapid simulation tools: the simplified na-
and Dowlatabadi [100], Schneider [103] and Parson and ture and flexible structure of IA models permit rapid
Fischer-Vanden [88]. prototyping of new concepts and scientific insights;
158 J. Rotmans / Methods for IA

• consistent frameworks to structure scientific knowledge: to population growth, economic development, resource use
critical uncertainties, gaps in scientific knowledge and and environmental degradation, let alone regional interac-
weaknesses in discipline-oriented expert models can be tions through migration and trade.
identified; The trend towards greater and more complicated IA
• tools for communication: IA models can be useful tools models makes heavy demands on the documentation on the
in communicating complex scientific issues to decision one hand, and the scientific testing procedure on the other.
makers, disciplinary scientists, stakeholders, and the A potential weakness of complicated models is that they are
general public. often poorly documented, which makes them intangible for
other modelers, but, more importantly, incomprehensible
Obviously, IA models also have limitations and weak- for a wider audience. The scientific testing of complicated
nesses. Among the most important ones are: IA models should be a major part of the model building
• high level of integration: many processes occur at a mi- cycle. The more complicated the model, the higher the
cro level, far below the spatial and temporal aggregation possibility of errors and bugs. It requires thorough testing
of current IA models; to pick up most if not all errors and bugs, an activity which
is, unfortunately, heavily underrated.
• inadequate treatment of uncertainties: IA models are
prone to an accumulation of uncertainties, and to a va- 2.1.2. Scenarios
riety of types and sources of uncertainty; Thinking about the future is often done by the devel-
• absence of stochastic behaviour: most IA models de- opment of scenarios. Theoretically speaking, scenarios are
scribe processes in a continuous, deterministic manner, hypothetical sequences of events, constructed for the pur-
excluding extreme conditions that may significantly in- pose of focusing attention on causal processes and deci-
fluence the long-term systems behaviour; sion points (Kahn and Wiener [57]). In practice, scenarios
• limited calibration and validation: the high level of ag- are archetypal descriptions of alternative images of the fu-
gregation implies an inherent lack of empirical variables ture, created from mental maps or models that reflect differ-
and parameters, and current data sets are often too small ent perspectives on past, present and future developments.
and/or unreliable to apply. Ideally, they should be internally consistent, plausible and
recognisable stories exploring a path into the future (see
Current trends. The trend in current IA modeling is to also Anastasi [5]). Although many scenario definitions ex-
move toward greater and greater disaggregation, assuming ist, most of them share the following characteristics:
that yields better models. In general, it is difficult to know • scenarios are hypothetical, describing possible future
when to stop building more detail into an Integrated As- pathways;
sessment model. With each incremented level of sophis- • scenarios describe processes, representing sequences of
tication from the interlinkages of more processes in more events over a period of time;
detail, comes new insight. However, past decades of model
building have shown that small and transparent models are • scenarios consist of states, events, actions and conse-
often superior in that they provide similar results to large quences which are causally related;
models faster and offer ease of use. In this respect, it • scenarios start from an initial state (usually the present),
is useful to distinguish between complicated and complex depicting a final state at a fixed time horizon.
models. Complicated models are models that include a va- Scenarios are perhaps most effective when seen as a
riety of processes, many of which may be interlinked. If powerful tool to broaden perspectives, raise questions, and
incremental changes in these processes generally lead to challenge conventional thinking. However, that is not the
incremental changes in model output, the dynamics of the way they have been used. Most scenarios developed in the
model is almost linear, and not complex at all. Complex past were dominated by engineers, economists, and plan-
models, however, may contain relatively few processes, but ners, while contributions from the social sciences have been
incremental changes in these processes may result in con- rare (Jungermann [55]). That could possibly be an expla-
siderable changes in the results of the overall model. This nation for the rather dogmatic and predictive character of
non-linear behaviour may be due to the inclusion of feed- many scenario exercises in the past. Looking back at de-
backs, adaptation, self-learning, and chaotic behaviour, and velopments of the past 25 years, however, one clear lesson
is often unpredictable. can be learned from scenario-based assessments made in
Practically speaking, this means that disaggregation of the 1970s: dogmatic predictions regarding the Earth’s fu-
IA models has profound consequences for the dynamics ture are unreliable and can be politically counterproductive
of the model. Breaking down a global model into vari- (UN [112]).
ous regions, implies that the regional dynamics should be Nowadays, it is generally accepted that scenarios do not
dealt with in an adequate manner. Regional IA models use predict, but that they paint pictures of possible futures and
grid cells or classes for representing geographical differ- explore the different outcomes that might result if basic
ences and heterogeneities in regional IA models. They do assumptions are changed, for example, regarding policy in-
not capture, however, the regional dynamics with regard terventions. So the only relevant question that scenarios can
J. Rotmans / Methods for IA 159

address is not whether an event will happen in the future, set of possible events irrespective of their (un)desirability,
but what we could do if it did happen. Also, the crucial role whereas normative scenarios take values and interests into
of uncertainty is increasingly recognised, which has led to account. Current scenario studies mostly have a descriptive
the conclusion that scenario building should not be a purely character or are implicitly normative.
deterministic scientific activity. This increased understand- Finally, trend and peripheral scenarios can be set out
ing has resulted in a number of well-thought-out scenario against each other. A trend scenario represents the ex-
exercises during the last decade that have proven their value trapolation of past and current trends, while a peripheral
in a particular context. For example, the IPCC has devel- scenario includes trend-breaking and/or surprising devel-
oped two sets of scenarios, which illustrate the impacts of opments. Many recent scenario studies are hybrid, in the
specific developments in population growth, energy use and sense that they are neither trend nor peripheral scenarios.
technology in terms of resulting emissions and atmospheric This has to do with the fact that the anticipated changes in
concentrations of greenhouse gases and its associated pat- these scenarios are merely incremental.
terns of climate change (IPCC [47,48]). The World En- Although in many long-term scenario studies surprises
ergy Council (WEC [116]) developed global scenarios that and bifurcations are not taken into account, history arguably
underlined the importance of diversity of supply, assuming shows that trends are characterised by strong fluctuations
shifts in the energy system due to autonomous technological rather than smooth curves, often triggered by unexpected
developments and market forces, and not through govern- changes. Surprises can be categorised in the following way:
ment interference. A widely used scenario developed by the
• unimaginable surprises (like a journey to the Earth’s cen-
Global Scenario Group (Gallopin et al. [40]) was the Con- tre in the times of Jules Verne);
ventional Development Scenario. This scenario assumes
demographic, economic and energy assumptions based on • imaginable surprises that are improbable (like a global
mid-range IPCC choices, supplemented with assumptions nuclear war);
for water and land use and food intake. The Global Sce- • imaginable surprises that are probable (like an oil price
nario Group has recently produced three classes of global shock, or massive migration due to ecological disasters);
scenarios, based on alternative social visions: Conventional • certain surprises or events (like earthquakes).
Worlds, Barbarisation, and Great Transitions, described in
terms of economic scale, equity, environmental quality, Scenario method classification. Scenarios can be classi-
technological change, and degree of social and geopolitical fied along the lines of: (i) the type of tool or method used
conflict. Shell (Kassler [61]) recently produced a number for designing the scenario, and (ii) the functionality and fo-
of world scenarios, where three alternative societal visions cus of the scenario exercise. Regarding the first group, we
have been formulated and weighed against their economic distinguish between quantitative scenarios based on the use
robustness. of models, qualitative scenarios based on narratives, and
In the field of sustainable development, one of the more hybrid scenarios based on participatory methods. Model-
comprehensive scenario studies has been undertaken by ing methods involve the use of computer models, either as a
Robinson [93], sketching a desirable, sustainable future for central means to explore the future consequences of sets of
Canada in 2030, and the changes that would need to oc- assumptions or as a tool to check the consistency of the de-
cur till 2030 to arrive at this future vision. Also at the veloped scenarios. Narrative scenarios are usually deployed
European scale a number of scenario studies have been in cases where data is weak or missing. Pathways into the
performed: EC-DGXII [29], in the field of energy by the future are then qualitatively described. These two methods
European Commission DGXVII [30], and on political and are usually adopted in scenario projects involving a small
institutional developments by Smith [107]. “core” group of experts. In contrast, participatory methods
refer to approaches in which stakeholders (non-scientists),
Scenario classification. In the scenario literature different such as decision makers, business people, and lay people
subdivisions of scenarios exist (van Asselt et al. [8]). First, play an active role: the participants are all co-designers of
a subdivision can be made between forecasting and back- the scenarios. Anastasi [5] advocates the use of partici-
casting scenarios, or exploratory versus anticipatory sce- patory approaches with the argument that complementary
narios. Forecasting or exploratory scenarios are forward- heterogeneity in perspectives, expertise and knowledge is
directed, i.e., they explore future consequences of a se- needed to guarantee sufficient richness. He uses the focus
quence of assumptions, with or without expected/desired and functionality of scenarios as distinguishing character-
policy efforts. The majority of recent scenario studies can istics for classifying scenarios: global scenarios, focused
be characterised as forecasting exercises. Backcasting or scenarios and single issue scenarios. Global methods result
anticipatory scenarios on the other hand, are backward di- in scenarios that depict the future of the world as a whole,
rected, i.e., they start from some assumed final state, and focused methods concentrate on a particular geographical
explore the preconditions that could lead to this state, in- area, and single-issue methods focus on a particular issue
cluding a palette of strategies to reach this situation. like energy or transport. Anastasi [5] explored the differ-
Secondly, we can distinguish between descriptive and ences between the various approaches and their use in sce-
normative scenarios. Descriptive scenarios state an ordered nario building. He concluded that up till now model-based
160 J. Rotmans / Methods for IA

Table 1
Scenario methods.
Modeling Narrative Participatory

Global Rotmans and de Vries [99] UN-DPCSD [112] no examples


WEC/IIASA [116]
Focused BCI [10] VROM [115] TCPA/CER [110]
EC-DGXI [29]
Single issue ECN [31] Button [14] EC-DGXVII [30]
IPCC [47,48]

scenario studies, as well as global narratives, and energy Current trends. There is an increasing tendency towards
and technology-oriented scenarios enjoyed most exposure the use of participatory methods for designing scenarios.
in science and decision-making communities. Another trend is to develop scenarios with both a strong
Table 1 provides some examples of scenario studies as- narrative and a quantitative component. These trends can
sociated with the various approaches. For a detailed de- complement and reinforce each other: a group of stakehold-
scription of the scenario studies listed in this table, the ers with different knowledge, experiences and perspectives
reader is referred to van Asselt et al. [8]. provides a greater richness to scenarios, while appealing
narratives underpinned by numbers are able to engage key
Strengths and weaknesses. Scenarios can be useful tools stakeholders. One of the major lessons drawn from the last
to: decades is that active dissemination is needed to ensure an
adequate usage of scenarios by the stakeholders involved.
• articulate our key considerations and assumptions: sce- In order to fulfil their communicative and decision-
narios can help to imagine a range of possible futures if supportive role, scenarios need to be coherent and consis-
we follow a key set of assumptions and considerations; tent. The aspect of coherency refers to the inclusion of all
• blend quantitative and qualitative knowledge: scenar- relevant dimensions and all relevant interlinkages between
ios are powerful frameworks for using both data and the various processes considered. Coherent scenarios, how-
model-produced output in combination with qualitative ever, should not be too complicated, to avoid an overload
knowledge elements; of information communicated to decision makers. Consis-
tency implies that key assumptions done in the scenarios are
• identify constraints and dilemmas: exploring the future
checked among different scales, sectors and issues. For ex-
often yields indications for constraints in future devel-
ample, assumptions with regard to future energy use should
opments and dilemmas for strategic choices to be made;
be checked with assumptions regarding future land and wa-
• expand our thinking beyond the conventional paradigm: ter use.
exploring future possibilities that go beyond our conven- Unfortunately, many scenario studies do lack such a con-
tional thinking may result in surprising and innovative sistency check and are far from coherent, which makes them
insights. highly vulnerable in the assessment process. This also indi-
cates the need for IA scenarios, i.e., scenarios that condense
Current scenarios are also characterised by a number of critical developments into a consistent set of assumptions
deficiencies: along different temporal and spatial scales and different
aggregation levels, which ultimately results into coherent
• lack of diversity: scenarios are often developed from
narratives useful for decision making.
a narrow, disciplinary based perspective, resulting in a
limited set of standard economic, technological, and to
2.1.3. Risk analysis1
a lesser extent environmental assumptions;
Risk can be defined as the possibility that certain losses
• extrapolations of current trends: many scenarios do or damages occur as the result of a particular event or series
have a “business-as-usual” character, assuming that cur- of events. It usually refers to the possibility of something
rent conditions will continue for decades, excluding sur- going wrong. The notion of “risk” traces back to the 13th
prises; century (Jungermann [56]). It circulated in Italy in the
• inconsistent: sets of assumptions made for different sec- context of insurance against the loss of shiploads coming
tors, regions, or issues, are often not consistent with each from the Orient through, for example, storms or piracy. The
other; notion of “risk” is still used in the context of insurance, but
• not transparent: key assumptions and underlying im- apart from that it has become a widely applied, often used
plicit judgements and preferences are often not made and ambiguous notion. It is used in the field of business,
explicit. Also implicit is which factors are exogenous technology, health, politics, and sport.
and which ones are endogenous, and to what extent soci- Apart from this common use, risk is also used as a con-
etal processes are autonomous or influenced by concrete cept in scientific analysis for decision support. Since the
policies. 1 This section builds upon the ongoing PhD research of van Asselt.
J. Rotmans / Methods for IA 161

early 1970s, risk analysis has become an established field integrated risk analysis. Out of the scholarly literature, two
of research. It was originally rooted in the engineering key issues emerge that underlie the controversy in the risk
and decisions sciences. Since the 80s, risk analysis has community, i.e.:
gained importance as a concept in assessment activities, es-
pecially in technology assessment (O’Riordan [82]). In the • How to define risk?
last decade, risk analysis is increasingly used in environ- • How risky are modern hazards?
mental science, and applied to environmental hazards, like
With regard to both questions, roughly two extreme posi-
stratospheric ozone depletion and climate change. Since
tions can be identified. In the discussions pertaining to
the beginning of the 90s, risk analysis is gaining a broader
defining risk, the objectivists (e.g., Starr [108], Crouch
meaning as confined to general aspects of decision making
and Wilson [21]) and constructivists can be distinguished.2
in modern society (Kämper [58]).
Objectivists distinguish between objective and perceived
Risk analysis has now become an active field involving
risks. They consider perceived risks to be inherently
disciplines as diverse as mathematics, statistics, epidemiol-
wrong and have it that objective risks can be adequately
ogy, toxicology, biology, engineering, system analysis, law,
measured in terms of probability and utility. Construc-
psychology, decision sciences, sociology, anthropology, ge-
tivists deny the possibility of pure scientific, objective es-
ography, history, economics, management science, policy
timates (see, for example, Douglas and Wildavsky [23],
sciences and philosophy.
Otway and von Winterfeldt [83,84], Fischhoff et al. [34],
The concept of “risk” is closely connected with the
Slovic et al. [105], Johnson and Covello [54], Morgan
prevalence of uncertainty. Risk deals with uncertain situa-
and Henrion [76], Mayo and Hollander [69], Funtow-
tions in which a number of possible outcomes might occur,
icz and Ravetz [36,37], Krimsky and Golding [62], and
of which at least one, but possibly more, is undesirable
Adams [1]). Concerning the second issue, we find scholars
(Merkhofer [74]). Personal, collective, societal and envi-
(e.g., Beck [11] and later publications) holding the position
ronmental risks thus arise from the uncertainty surrounding
that hazards have increased and are increasing versus those
the occurrence of particular events or series of events. Al-
(e.g., Whipple [119] and Douglas and Wildavsky [23])
though uncertainty is central to risk analysis, the paradox is
who argue that modern society is only more sensitive to
that risk analysis is also complicated by the omnipresence
risk.
of uncertainty. Scientific uncertainties surrounding com-
plex societal phenomena make it very difficult to assess Notwithstanding this weakness in terms of lack of a com-
what is likely to happen if a particular event or a concur- mon interdisciplinary framework, the risk community pro-
rence of circumstances comes about. Furthermore, social vides concepts, tools, and methods that can be of help in
and behavioural research on risk perceptions indicates that the effort to conduct integrated assessments of complex is-
in case the underlying mechanisms and processes are not sues. Risk is a concept that is deployed in everyday life.
well understood, lay estimates concerning involuntary risks So it is a concept familiar to decision makers, stakeholders
tend to be higher (Covello [19]). So there is not only the and the broad public. As argued by Luhmann [65], risk is
problem that uncertainty violates the exactness of risk esti- the general form in which society describes and assesses
mates, but also that the level of uncertainty co-determines its future. In other words, two-way communication of fu-
the risk itself. ture outlooks in terms of risk is therefore likely to enhance
The above reveals that uncertainty and risk are closely understanding and co-operation between science and soci-
interlinked: the one cannot be adequately studied without ety.
considering the other, and vice versa. The challenge is to Three decades of risk analysis furthermore provide expe-
address both in an integrative manner. riences that are relevant to Integrated Assessment. Lessons
can be learned with regard to the following issues:
Strengths and weaknesses. Notwithstanding the involve- • the role of uncertainty and subjectivity in science for
ment of so many disciplines, we can doubt whether the policy;
risk community is in fact interdisciplinary. There is a clear
tendency to treat the issue of risk from the perspective of • the use of probabilistic analysis;
the own discipline and speciality (Conrad [16]). It is there- • risk communication;
fore perhaps not surprising that there is no common cog- • participation of non-experts in assessment efforts.
nitive action-guiding paradigm that integrates the different
aspects of risk. In contrast, the risk community suffers from In addition, research on risk perception and communica-
fundamental controversy. tion yields insights about what society wants to know about
The dilemma the risk community faces is that uncer- complex issues. Awareness of what is societally relevant
tainty interferes with facts and that conflicts are inherent may improve IA efforts in terms of usefulness for societal
to risk issues. The risk debate is rooted in the failure to decision-making processes.
reconcile the limitations of scientific analysis, the special 2 Terminology as used by Linnerooth-Bayer and Thompson [64]. This
characteristics of the human mind, and social and cultural debate is also described in terms of the quantitative versus the qualitative
factors into a common cognitive framework that can guide approach (Jasanoff [53]).
162 J. Rotmans / Methods for IA

Current trends. The risk debate has severe implications 2.2.1. Participatory methods for IA
for risk analysis activities. Over the last 25 years, both Participatory methods for IA differ with regard to the
schools had fervent advocates and fierce proponents in and profile of the participants, the goal of participation and the
outside the risk community, with the result that there is degree of participation. We distinguish three classes of
neither agreement over appropriate methods for risk analy- methods (Rotmans and van Asselt [101]):
sis nor acceptance of the outcomes of public processes.
Till the 90s the “objectivist” perspective has dominated risk • dialogue methods;
analysis. The objectivist view has come under increasing • policy exercises;
attack and its supremacy is weakening. The majority of • mutual learning methods.
risk analysts now seems to agree on the following premises
(Jasanoff [53]): The dialogue method is applied in cases where the in-
tended users are considered as a source of information nec-
• experts perceive risk differently from lay people; essary for the analysts to perform the assessment. The goal
• risk analysis is not a purely objective process: facts and of participation is to extract this essential information from
values merge frequently in dealing with issues of high the intended users. Within this class, we can discriminate
uncertainty; between continuous dialogues, and cases in which the di-
• cultural factors affect the way people assess risk. alogue takes place in a specific phase of the assessment.
In the first case, the role of the intended users can be de-
This shift, however, does not imply that the construc- scribed as that of co-designer. In case the dialogue takes
tivists are close to “take-over”. Criticism on the construc- place in the design phase, the role of the intended users
tivists school is that it notes that different parties hold dif- is to contribute to the design principles by sketching their
ferent risk perceptions, without making clear what to do wants and needs. This type of dialogue can be described as
about this (Vlek and Stallen) [113]. They do not provide a user-platform. If the dialogue is part of the final stage, the
an operational alternative for performing risk analysis. participants can be described as a test-group that helps to
The state-of-the-art in the risk community is thus char- determine how the results of the assessment will be brought
acterised by two trends, namely: (i) the objectivists’ par- to the fore. Van Asselt [6] sketches an example in which
adigm is dominating actual risk analysis, and (ii) the con- a “dialogue”3 was intended to serve as a user-platform. In
structivists win ground in epistemological matters. With adaptive ecological modeling (e.g., Holling [45], Costanza
regard to an integrated approach towards risk the follow- and Ruth [17]), the explicit aim is that the crucial design
ing trend is observed: individual researchers signalise the choices are made by the user community. The Delft process
necessity for a synthesis approach to risk analysis (e.g.,
around the IMAGE model (van Daalen et al. [22]) can be
Funtowicz and Ravetz [35], Jasanoff [53]), start building
considered as an example of a test group. In the RAINS
relevant networks and search for research commitments and
project the epistemic community involving scientists and
funds (Linnerooth-Bayer and Thompson [64]).
policy makers realised that the involved stakeholders got
the role of co-designer (Hordijk [46]).
2.2. Participatory methods for IA
Policy exercises build upon the tradition of simulation
“Participatory methods” is an umbrella term describing games. A policy exercise can be described as a flexibly
approaches for assessment in which non-scientists, such as structured process designed as an interface between sci-
policy people, stakeholders or even lay people, play an entists and policy makers (Toth [111]), in which a com-
active role. Brewer [13] first referred to the idea of using plex system is represented by a simpler one with relevant
participatory approaches in assessment efforts, however, IA behavioural similarity, and of which decision making is
projects only recently seek to apply such methods for IA part through human participants (Parson [85]). In gen-
purposes. eral, a game is set up that represents a negotiation process
Participatory methods are not new. Focus groups, for in which the different teams are responsible for a certain
example, have been applied in market research since the country or region. Policy exercises quite often make use
seventies (Cox et al. [20]), while consensus conferences of computer support. A computer model can be used as
have been applied in technology assessment (for example, a consulting device or as tool to convert the negotiated
on the issue of predictive genetic research (Platform for agreements into a new “state of the world”.
Science and Ethics [89])). Gaming approaches have been Policy exercises differ from the other participatory ap-
prominently in use for training purposes (e.g., military and proaches in the following ways: (i) the participants play
business games (Hausrath [44])). Finally, traditional con- roles, and (ii) the participants do not explicitly take part
tract research is a participatory process between clients and in the assessment process. A policy exercise is a way to
researchers. For instance, in developing a model, clients get information on human behaviour and policy preferences
may play an active role in: defining the model objectives necessary for the assessment the analysts produce. Exam-
and model output, making decisions about model features 3 Analysts and intended users in the policy community did not meet di-
involving various trade-offs, and choosing the method of in- rectly, but got informed about each others’ perspectives and attitudes via
teraction between the model and user (MacCallaway [66]). interviews.
J. Rotmans / Methods for IA 163

ples of policy exercises in IA are the climate policy ex- considered to be desirable in the eyes of stakeholders and
ercises (Parson [85–87]) and RIVM’s SusClime exercises citizens?
(de Vries [114]). Both efforts made use of IA models, The major weakness pertaining to participatory methods
respectively, a “tailor-made” version of the GCAM model is that the use of it in Integrated Assessment is in its in-
(Edmonds et al. [32]) and a preliminary version of the TAR- fancy, with the result that there are not (yet) established
GETS model (Rotmans and de Vries [99]). Experiences procedures and work-packages for setting up participatory
with policy exercises thus far are discussed in Parson [85]. processes.
The principle behind mutual learning approaches is that
participation of stakeholders and citizens enriches the as- 2.2.3. Current trends
sessment by a multiplicity of perspectives, skills and com- IA’s complexity, multi-disciplinary character, and wide
petence. The participants are considered as co-producers variety of societal actors present special challenges to the
of knowledge. They are not selected dependent on whether use of participatory methods. The use of such methods
the assessment is of use to them, but whether their per- in IA, therefore, requires substantial adjustment of existing
spectives, skills or competence complement the scientists’ procedures and the development of new tools and methods.
expertise and knowledge. Projects like the ULYSSES project (Jaeger et al. [50]), the
We distinguish two forms of mutual learning: IIASA policy exercise effort (Parson [85–87], Parson and
Fisher-Vanden [88]) and the Mackenzie Basin Impact Study
• the focus group approach, in which scientists play the (Cohen [15]) currently explore the necessary features of
role of facilitator and observer, and participatory methods for IA.
• the interactive approach, in which scientists are actively In addition, there is an increasing interest in participa-
involved as participants. tory methods, also in the modeling community. See, for
example, Steve Schneider’s plea (Schneider [103]) for the
In the first case, special groups are composed, typically
involvement of decision makers and stakeholders in IA ef-
consisting of citizens, policy makers or other stakeholders
forts.
who are provided with scientific input. The assessments
of these groups are then used in a broader IA process.
In interactive analysis (Guba and Lincoln [43], Grin and 2.3. Towards an IA tool kit
Hoppe [42]), problem definitions and proposed solutions
are integrated into a joint assessment that makes sense to Models, scenarios, risk analysis, and participatory meth-
both stakeholders and scientists. The ULYSSES project ods are by no means mutually exclusive methods for IA.
involves IA focus groups with lay people (Dürrenberger In general, they can complement and reinforce each other
et al. [28], Kasemir et al. [59]). The VISIONS project in different ways. This yields a tool kit for IA as sketched
(Rotmans [98]), starting in 1998, can be considered as a in figure 3. Models, for instance, can be used as a tool
first example of the interactive approach. for developing scenarios. On the other hand, models and
scenarios could be the result of participatory processes. Di-
2.2.2. Strengths and weaknesses alogue approaches can be oriented towards the development
Participatory methods are able to integrate quantitative and/or usage of IA models. Within policy exercises models
with qualitative knowledge (Parson [85]). IA models can- are often used as reference frameworks. In the ULYSSES
not (yet) grasp qualitative knowledge. In this way, partici- project (Jaeger et al. [50]), IA models are used in focus
patory methods can complement IA models and scenarios. groups as expert input into group discussions. In mutual
It is furthermore argued that involvement of non-sci- learning approaches, models might be helpful to compare
entists is needed to ensure the relevance and later accep- scenarios developed by participants with scientific insights.
tance of the analytical (modeling) approach (Hordijk [46],
van Asselt [6], Costanza and Ruth [17], Alcamo et al. [3]).
Participatory methods are needed to set up such a dialogue
between analysts and societal actors. A widely used ar-
gument in favour of using participatory methods is that it
helps structure the IA process, and thus also the analytical
methods. After a period of building up thrust and support,
they will become supportive to those methods.
Participatory methods are also justified on arguments in-
spired by the nature of democracy (e.g., Kasemir et al. [59]).
Policies for managing sustainability will be effective only
if they have the moral support of a great mass of people
(Ravetz [91]). It is therefore argued that assessments should
comprise the opinions and attitudes of stakeholders and cit-
izens. Are the strategies proposed by the assessment study Figure 3. IA tool kit.
164 J. Rotmans / Methods for IA

pressure for research toward unattainable goals or may mire


development efforts in the struggle among conflicting value
systems. For example, decision makers want to have geo-
graphically explicit impacts, which has led many research
groups to eschew accurate characterisation of what may be
projected in favour of a precise characterisation of impacts
with a much higher probability of error. Under these cir-
cumstances, the modelers need to communicate to decision
makers the difficulty of attaining such lofty goals. Failing
that, they should make clear the magnitude of uncertainties
in geographically explicit projections.
Figure 4. Methods cycle (source: Rotmans and van Asselt [101]). The next generation of IA models should make use of
enhanced IA concepts, as a result of a continuous consul-
A major methodological challenge for IA is to take ad- tative process and interfacing with representative groups
vantage of the existing multitude of methods. Figure 4 of decision makers. This means, among other things, de-
suggests how the IA models, scenarios and participatory termining the appropriate level of aggregation, making the
methods could be used in a complementary manner. models actor-oriented by introducing behavioural rules and
The cyclic scheme can serve as a heuristic device to de- decisions, deriving consistent and integrated scenarios, and
sign a variety of sensible procedures for IA. Starting with including explicit treatment of uncertainties. The interac-
participatory methods and then turning clockwise, sketches tion process among model development groups and deci-
a procedure in which participatory methods provide the ba- sion makers should curb the demand for the unattainable
sis for the development of an IA model that, in turn, is used and prepare decision makers for the difficulties they face in
to develop scenarios. These scenarios are then input to the setting strategies when large uncertainties abound.
next round of participatory meetings, where the developed
models are adjusted, etc. Starting with IA models and turn-
ing anti-clockwise, results in the following iterative process: 3. Methodological challenges
an IA model is developed as a multidisciplinary effort, and
is then used as scientific input to participatory processes As the Integrated Assessment process develops, more
in which a variety of scenarios is developed. The consis- sophisticated methods and tools are needed. Therefore, we
tency and plausibility of these scenarios are then checked will face challenges in developing a balanced tool kit in
using IA models. These scenarios in turn serve as input for order to achieve an adequate knowledge base. In this sec-
the participatory process, etc. For example, the VISIONS tion we will discuss some of the methodological challenges,
project (Rotmans [98]) envisages an anti-clockwise com- which largely follow from the strength/weakness analysis
bined usage of scenarios, participatory approaches and IA of the IA methods as presented in the former section. The
models, starting with the design of scenarios. challenges are subdivided into over-arching methodological
The practice, however, differs from this ideal picture. focal points and method-specific challenges.
None of the current generation of IA models of climate
change is being developed in close conjunction with deci- 3.1. Over-arching methodological challenges
sion makers. This explains why many of the current IA
models are only of limited use to decision makers. For ex- Over-arching methodological challenges are those that
ample, single-region economic optimisation models offer are relevant to all methods discussed before. We identify
the comparative efficiencies of various strategies, but offer the following three key issues:
little insight of political value because they do not deal with • Aggregation versus disaggregation;
the distributional aspects of climate policies. This means • Treatment of uncertainty;
that this type of IA model serves only one purpose, i.e., cal-
culating optimal control strategies under uncertainty, which • Blending qualitative and quantitative knowledge.
is more difficult in multi-region, multi-agent models. This In the following, these methodological challenges are
also illustrates that IA models should function within the discussed in more detail.
setting of the political arena, where efficiency often takes a
back seat to equity. In such a setting, different groups ex- 3.1.1. Aggregation versus disaggregation
hibit different levels of risk perception, they adopt different One of the most critical issues in Integrated Assessment
decision rules, and their utility bundle is composed of dif- is that of aggregation versus disaggregation. The level of
ferent indicators. Only a few of the current IA models are aggregation of an IA refers to the spatial and temporal res-
flexible enough to represent the interests of various parties olution and the level of complexity used in the assessment.
(through space and time). The problem of IA is that it has to cope with a variety
Having decision makers as early stakeholders in an IA of processes that operate on different temporal and spatial
research program is desirable, but may lead to irresistible levels, and differ in complexity.
J. Rotmans / Methods for IA 165

In general, there is no optimal spatial and temporal scale • reciprocity: to test the reciprocal relationship between
at which IA has to operate: that strongly depends on the local/regional processes and global processes.
nature and scope of the problem to be tackled by IA.
First, IA has to connect disciplinary processes that differ Third, IA is faced with a multitude of temporal scales.
by nature: physical processes, monetary processes, infor- Short-term needs and interests of stakeholders have to be
mation processes, and policy processes. Because of the considered. However, biogeochemical processes usually
multitude of disciplinary processes to be combined, an as operate on a long time scale, whereas economic processes
simple as possible representation of disciplinary knowledge operate on short- to medium-time scales. Another chal-
is preferable. There is, however, no unifying theory for IA lenging aspect of IA is to interconnect long-term targets
how to do this. In addition, the processes to be linked are as specified as a result of analysing processes operating on
usually studied in isolation from other disciplinary fields. longer-term time scales, with short-term goals for concrete
This isolation is needed as part and parcel of the classic policy actions (Jaeger et al. [49]). Unfortunately, there is
not yet a sound scientific method how to do this; thus so
model of scientific progress and discovery. However, when
the constraints of isolation are removed, there is a variety far only heuristic methods have been used.
of ways in which to connect the reduced pieces of disci-
3.1.2. Treatment of uncertainty
plinary knowledge. This manifold of possible integration
Any exploration of future developments inevitably in-
routes, for which there is again no unifying theory, is one
volves a considerable degree of uncertainty and Inte-
of the reasons why quality control is so difficult to achieve
grated Assessment is no exception. Because of the cross-
in IA.
disciplinary character of IA, it includes many different types
For instance, in order to link the reduced pieces of dis-
and sources of uncertainty. Because IAs are end-to-end ap-
ciplinary knowledge in a systemic way, one can use ele-
proaches, they also contain an accumulation of uncertain-
ments from classical systems analysis, or the method of
ties. Uncertainties may arise from incomplete knowledge
system dynamics, or a sequential input–output analysis, or
of key physiological, chemical and biological processes.
a correlation-based approach, or a pressure–state–impact–
Many uncertainties are of a socio-economic nature – related
response approach.
to people’s behaviour – and reflect inadequate knowledge
Second, IA has to deal with different spatial scale levels.
with respect to the driving forces of human behaviour.
One of the ultimate challenges in IA is to connect higher
Various attempts have been made to classify the differ-
scale assessments with lower scale ones. So far, there has
ent types and sources of uncertainty. Morgan and Hen-
been hardly any experience with playing around with scale
rion [76] distinguish uncertainty about empirical quanti-
levels in IA. Down-scaling or up-scaling the spatial level
ties and uncertainty about the functional form of mod-
of analysis has profound consequences. This is related to
els, which may arise from: subjective judgement; dis-
the question to what extent the processes considered are
agreement among experts; systematic errors; approxima-
generic, or distinctly spatially bound in character. In other
tion; and inherent randomness. Funtowicz and Ravetz [35]
words: does a relationship hold at larger or lower scale
classified uncertainties in three categories: (i) techni-
levels? For instance, land and water issues seem to be
cal uncertainties (concerning observations versus measure-
local/regional by nature, whereas atmospheric issues often
ments); (ii) methodological uncertainties (concerning the
appear to be trans-regional or even global.
right choice of analytical tools); and (iii) epistemological
In this context, Root and Schneider [95] propose an in-
uncertainties (concerning the conception of a phenomenon).
teresting approach: the so-called Strategic Cyclical Scaling
Whatever classification is chosen, the various types and
(SCS) method. This method involves continuous cycling
sources of uncertainties in IAs need to be addressed in an
between large and small-scale assessments. In modeling or
adequate manner. Unfortunately, the classical methods of
scenario terms such an iterative scaling procedure implies
handling uncertainty, such as Monte Carlo sampling, and
that a specific global model or scenario is disaggregated and the usage of probability density functions, are merely useful
adjusted for a specific region, country or river basin. The to address technical uncertainties. Methodological and epis-
new insights are then used to improve the global version, temological uncertainties, however, which primarily arise
after which implementation for another region, country or from subjective judgements and fundamental disagreement
river basin follows. This SCS method can be used for con- among experts, cannot be addressed by the classical meth-
ceptual validation of models or scenarios. ods. Applying these methods usually results in estimates of
Methodologically, a local or regional approach for IA minimum, maximum and best guess values that are often
is more attractive than a global approach, for a number of erroneous and misleading.
reasons (Costanza and Tognetti [18]) in terms of: An alternative and simple way of presenting uncertain-
ties is by specifying a set of future scenarios, where scenar-
• data quality: more reliable data is available locally;
ios then span a range of plausible, representative futures.
• manageability: fewer interactions, and fewer causes and This, however, does not give an indication of the cumula-
impacts have to be considered; tive uncertainty as well as the origin and meaning of the
• communication: greater political interest; uncertainty range.
166 J. Rotmans / Methods for IA

In IA models uncertainties are often reduced to technical validity. IA therefore needs combinations of quantitative
artefacts. By attaching deterministic intervals or stochastic and qualitative approaches, from the perspective that they
probability distribution functions to uncertain model para- complement each other. To this end, analytical and partic-
meters, it is suggested that variations in parameter values ipatory methods are not, and should not be, mutually ex-
do yield estimates of the uncertainty in the model outcome. clusive alternative methods for integrated assessment. The
However, that may be true in the mathematical sense (al- major methodological challenge for IA is to take advan-
though only partly), but it does not reflect the nature and tage of the complementary value. In order to combine and
source of the real world uncertainties. One may compare display qualitative and quantitative information in a sys-
one type of uncertainty with the other in mathematical tematic, coherent framework, various traditional methods
terms, but in physical terms that could lead to compar- could be used (Morgan and Henrion [76]). However, we
ing apples and pears. It is not allowed to simply compare will single out only two of the “new” methods: fuzzy logic
the uncertainty of the climate sensitivity (representing un- and knowledge patterns.
certainties in geophysical feedbacks) with the uncertainty One promising way to blend quantitative and qualitative
of the fertility rate (representing uncertainties in triggering information could be to incorporate vague and qualitative
factors behind fertility behaviour). While the geophysical knowledge in IA models. A useful tool to implement vague
uncertainty might be reduced by future research, the demo- information is fuzzy logic. Contrary to Boolean logic, fuzzy
graphic uncertainty might be structural in the sense that it logic makes use of a continuum of values, which represent
cannot be reduced in the longer term. elements of fuzzy knowledge. This technique has been
In sum, current methods to analyse uncertainties in IAs successfully applied in control and optimisation problems.
are unable to: (i) produce an order of magnitude of the cu- In the field of IA, the technique is, amongst others, be-
mulative uncertainty; (ii) render various sources and types ing applied in the Syndromes project, where syndromes are
of uncertainties explicit; (iii) provide systematic and coher- the symptoms of global change as proposed by the Ger-
ent clusters of uncertainties; and (iv) explain and clarify man Advisory Council of Global Change (Schellnhuber et
uncertainties in a manner understandable to decision mak- al. [102]).
ers. A more radical approach would be to leave the pathway
In order to meet the above requirements, we need new of strict cause–effect chains, and enter the road of knowl-
methods. An interesting manner to make various types edge patterns. Knowledge patterns are sequences of knowl-
of uncertainties explicit might be to link uncertainty aris- edge elements which consist of facts (certain, quantitative
ing from subjective judgement and fundamental disagree- knowledge), hypotheses (fuzzy, qualitative knowledge), and
ment among experts to the different perspectives that people fundamental uncertainties (perspective-based knowledge).
have adopted. A typology of perspectives is then necessary The basic idea is to consider complex issues as controver-
to arrive at a limited set of perspective-based interpreta- sies, where each controversy is analysed as a knowledge
tions of uncertainties. The method of multiple perspective- pattern that consists of a sequence of facts, hypotheses
based model routes (van Asselt and Rotmans [7]) helps to and uncertainties. The knowledge pattern could then be
understand the differences in projections as the outcome coloured according to different perspectives. This helps to
of the divergent views and valuations, instead of merely develop a consistent and coherent picture of the underlying,
low, medium and high values. The first application of this complex dynamics of the controversy in question.
method in IA modeling is extensively described in Rotmans
and de Vries [37]. 3.2. Method-specific challenges

3.1.3. Blending qualitative and quantitative knowledge 3.2.1. Models


In most frameworks for IA, quantitative and qualitative During the last decades, IA models have proven them-
knowledge are considered and treated as mutually exclu- selves as legitimate and powerful approaches to com-
sive. For instance, usually those aspects of a problem plex issues. In particular to the climate change debate,
under concern that are not well known, or about which IA models have contributed by exploring impacts of climate
there is only vague and qualitative knowledge, are left out change and evaluating mitigation and abatement strategies.
in the modeling process. This means, however, that we IA models have also provided useful information on bal-
miss crucial links in the causal chains that form archetypal ancing the carbon budget, sulphate aerosols, and on various
patterns of human–environment interactions. Quantitative integrated aspects of land use.
rigour therefore prevents IA from being comprehensive, in However, there is still a long way to go before IA models
the sense of studying all relevant aspects of a complex prob- will be fully accepted by the scientific modeling commu-
lem. It is therefore illusory to think that the full complexity nity on the one hand, and by the decision making com-
of human–environment interactions could be integrated into munity on the other. Therefore, an ultimate challenge for
a formal, quantitative modeling framework. IA modeling is to build up scientific and political credi-
On the other hand, purely qualitative assessments in the bility. To improve both scientific and political credibility
form of narratives that are not underpinned by quantitative of IA models, IA scholars should enhance IA models, im-
estimates cannot be checked in terms of consistency and prove the communication with disciplinary scientific and
J. Rotmans / Methods for IA 167

policy-making communities, and enrich and augment com- context. So far, the knowledge available has hardly been
munication techniques. These strategies will be discussed used in IA models. Among the more pressing questions in
in this section. social dynamics are the following:
Enhancing IA models is a continuous process, and vari-
ous strategies for further improving IA models can be fol- • What brings about the demographic and epidemiological
lowed. First, realising that an IA framework is as good transition?
as its weakest part of the whole model chain, it is much • What drives environmentally related human behaviour?
more effective to improve the rather weak or poorly defined • What are the roots and dynamics of technological inno-
model parts, rather than refining and disaggregating the al- vation and diffusion?
ready adequate submodels. Second, a no-regrets modeling • What drives processes such as urbanisation and migra-
strategy is to bring in, where necessary and where possible, tion?
region-specific information for submodels, but maintain as
overall modeling strategy the ability to run the model at Although the above questions cannot be answered unam-
a fairly high level of aggregation. Third, building generic biguously, sound scientific hypotheses do exist. Using these
submodels (models independent of regional or temporal dif- hypotheses in combination with data and expert models
ferences) has many disadvantages. This means that the the- available, forms the basis upon which these socio-economic
ories and assumptions must be applicable at different levels processes can be represented in IA models. Here, we will
of spatial aggregation and for different regions in differ- only present IA modeling examples in the fields of demog-
ent periods (Rotmans and Dowlatabadi [100]). And finally, raphy and health, and human behaviour. Both cases are
empirical testing is a crucial element in enhancing the qual- examples of generic model building strategies, i.e., mod-
ity of IA models. To this end, various model components els that can be applied to different scale levels. As far
could be tested against disciplinary expert models, not only as the remaining socio-economic issues are concerned, the
to compare the outcomes of both types of models, but also reader is referred to other sources (e.g., Schneider [103],
to validate the scientific hypotheses, process formulation, Dowlatabadi [24]).
boundary conditions, and model structure with each other.
For brevity’s sake we will not discuss here all modeling IA modeling of consumption behaviour. Human consump-
strategies, but instead focus on improving the weak parts tion of goods and services is a major factor in the process
of IA models. Let us consider some of the weaker parts of global change. So far, IA models only implicitly incor-
of IA models. In general, knowledge of key dynamics in porate the behaviour of consumers. Two main reasons for
both social and natural systems is limited. But there is no modeling consumer behaviour more explicitly are: (i) to
doubt that IA models are unbalanced with regard to the enhance and validate future projections generated by IA
representation of both interacting systems. models; (ii) to offer the possibility to simulate effective
While in many macroeconomic-oriented IA models the and efficient behavioural strategies in the context of IA.
socio-economic aspects are adequately represented, but the The growing concern about the environment has stim-
physical aspects poorly, in most biosphere-oriented IA mod- ulated the social sciences in developing conceptual multi-
els the natural system is reasonably well represented, but theoretical models describing various environmentally detri-
the socio-economic system only caricaturally. For example, mental behaviours. Multi-theoretical here indicates that
the dynamics of the climate–atmosphere–biosphere system various theories on specific types of behaviour are com-
continues to be far from well understood, mainly because of bined. These conceptual models are useful for the post-
a dearth of knowledge and data about the internal dynamics hoc explanation of environmentally relevant behaviour, and
of this complex system. A continuing stream of surprise they indicate policy strategies that might be effective in
findings (e.g., missing carbon and nitrogen sinks, aerosols, changing such behaviour. However, because these con-
and solar spots) typifies this. Still, there is a solid scientific ceptual models generally involve a poor specification of
knowledge base for disciplinary modeling of many biogeo- the relations among relevant variables and lack a system-
physical processes. Disciplinary expert models exist, from dynamical approach, they do not allow for the translation
which knowledge can be extracted and represented in re- into mathematical models, which would fit into the frame-
duced form. work of IA models. A useful behavioural model within the
With regard to the socio-economic system, the situa- framework of IA should enable us to develop projections of
tion differs materially. The social scientific knowledge base future behaviour, thus providing a tool for the simulation of
is smaller, resulting in less disciplinary models to be rep- behaviour. Such a generic behaviour model provides means
resented in reduced form in IA models. In addition, the to generate possible answers to the following questions:
modeling of socio-economic aspects suffers from a dearth
of basic data, in particular outside industrialised countries. • How do various kinds of consumptive behaviour interact
Therefore, the models we have developed to describe so- with projected developments in various domains, e.g.,
cial dynamics have not been calibrated and validated ad- population dynamics, water, food, and energy resources?
equately. There is, however, still sufficient knowledge of • Which response strategies provide effective means to
vital socio-economic processes to use in the IA modeling change particular kinds of environmentally detrimental
168 J. Rotmans / Methods for IA

one’s motivation to use this opportunity. Opportunities and


consumptive behaviour in the context of the search for
abilities are characterised in terms of physical means, finan-
a sustainable future? cial means, knowledge, authorisation, and one’s social and
organisational support. Dependent on the level of motiva-
In order to be able to address these questions, the dynam- tion, opportunities and abilities, the behavioural process may
ics of the behavioural process have to be modeled in a for- include more or less deliberation or cognitive processing. The
mal, quantitative way. In the field of economics, modeling degree of cognitive processing before using or rejecting any
consumer behaviour has a long history. Demand theory is opportunity depends on the strength of a consumer’s motiva-
tion to use this opportunity and on his or her behavioural con-
well developed, and allows for the specification of demand
trol. The underlying paradigms are those of reasoned versus
as a function of, e.g., prices and income. Models based automatic behaviour processes, in combination with individ-
on this theory suppose that there is a general equilibrium ual versus social behaviour processes, spanning eight major
between demand and supply (which is also price-driven). theories on consumptive behaviour.
The dynamics then arise from: (i) providing for market In translating the conceptual model into a behaviour simu-
lation model, consumers can be represented as computerised
clearing; (ii) adding more market levels; (iii) adding a time
agents, so-called “consumats”. The basic design of every
dimension; (iv) adding an optimal allocation of resources; consumat comprises a set of rules, which are rooted in the
and (v) changing consumer preferences. used theories of human behaviour. However, the consumats
However, the overall problem with the neo-classical gen- differ with respect to their motivations (e.g., cultural perspec-
eral equilibrium theory is that it does not allow for multiple tive), and personal abilities, thus representing various types
of consumers. By including interactions and feedbacks be-
equilibria (such as the transition from one economic equi-
tween different consumats, various social processes can be
librium to another), and the complex dynamics of price simulated: e.g., status-related consumption, the dispersion of
adjustments. Further, it does not take into account positive new consumer opportunities (“trickling down”) and the man-
feedbacks such as “learning by doing”. Finally, equilib- ifestation of scarcity.
rium dynamics are only valid on a certain domain, for in-
cremental changes from a given situation. The attempts
IA modeling of population and health. Fertility modeling
to model consumer behaviour in current IA models are
approaches are widely accepted and, only recently, existing
largely based on the general equilibrium model. For ex-
mathematical techniques have been introduced in the health
ample, many macroeconomic-oriented IA models feature a
area (Weinstein et al. [117], WHO [120]). Usually, regres-
Ramsey type of growth model in which a consumer allo-
sion techniques are used to explore the relations between
cates income between consumption and investment with a
broad health determinants, like literacy, income status, nu-
view to intertemporal utility maximisation.
tritional status, water supply and sanitation, education and
The above obviously illustrates that there is a need for
medical services, and the health status measured in healthy
an integrated, more realistic approach towards modeling
life expectancy. However, these regression techniques can
of consumer behaviour. This means a systems-based ap-
only give some suggestive evidence on the causes of pop-
proach of consumer choice (like a physically based model ulation and health changes. Statistical models to estimate
of ecosystem succession). An example of such an inte- future fertility and health levels are based on extrapolation
grated, conceptual model of consumer behaviour is pre- of past and current data. They operate on a short time
sented in box 1. horizon, and are static in terms of specifying the dynamics
behind changing fertility and health patterns.
Box 1 Therefore, there is a need for integrated approaches that
A conceptual model of consumer behaviour.
take account of the simultaneous occurrence of multiple
A conceptual model of consumer behaviour is under devel- risk factors and diseases, as well as cause–effect relation-
opment, and described in Jager et al. [51,52]. As a basis for ships. Such an integrated approach cannot be used in the
the model, a systems approach is chosen, using the Pressure– clinical area on an individual basis, but is appropriate at the
State–Impact–Response framework. The system-dynamical population level.
structure allows for the inclusion of interactions and feed-
In Rotmans and de Vries [99], a Pressure–State–Impact–
back loops. The Pressure system describes the pressures on
consumptive behaviour in terms of people’s Motivation, Op- Response systems approach to population and health is pre-
portunities and Abilities (MOA). Then, the relations between sented. A generic model has been designed that simulates
the MOA variables and behavioural processing are described the driving forces (socio-economic and environmental fac-
in the State system, based on eight main theories on behav- tors), the fertility behaviour and disease-specific mortality,
iour. The Impact system describes the impacts of human
the burden of disease and life expectancy as well as the size
behaviour in terms of consumption (amount of product use)
and need–satisfaction. Further, four types of policy strategies and structure of the population, and a number of fertility
are adopted in the Response system: (i) provision of physi- and health policies. The major objective of the population
cal alternatives; (ii) regulation; (iii) financial stimulation; and and health model is to simulate changes in morbidity and
(iv) social and cognitive stimulation. mortality levels under varying social, economic and envi-
The first principles on which consumer behaviour is based
ronmental conditions.
in this model are the following. The concept of motivation
is concerned with the satisfaction of human needs by means A second but related example is the MIASMA frame-
of “opportunity consumption”. The more one perceives an work, which is designed to describe the major cause and
opportunity as a possibility to satisfy certain needs, the higher effects relationships between atmospheric changes and hu-
J. Rotmans / Methods for IA 169

man population health (Martens [67,68]). The model fo- The basic procedure of using cellular automata models,
cuses on climate change (in terms of changes in temper- GIS models, and IA models in an integrative way is as fol-
ature and precipitation) and ozone depletion (in terms of lows: macroscopic patterns indicate changes on the macro-
changes in UV-B radiation). Under varying climate and scale for a region with regard to: physical, demographic
ozone regimes, changes in the dynamics and distribution and socio-economic changes, which are generated by sce-
of vector-borne diseases are simulated (malaria, schistoso- narios in combination with IA models. These macroscopic
miasis, and dengue), changes in cardiovascular diseases, trends will be passed on to a region-specific cellular au-
changing patterns of skin cancer incidences, and changing tomata model, which will allocate the detailed localisation
mortality levels as a result of thermal stress. Although the of these changes on the basis of its own spatial dynam-
model operates on the global level, it is being validated ics. Using site-specific information in combination with
for specific developing regions. It should be noted that this GIS information, cellular automata allow for representation
model does not take into account changes in the health care of the microscopic dynamic patterns. Then the cellular au-
system. A further integration step would then imply esti- tomata model will feedback local-specific information to
mates of future human health conditions as a result of both the macroscopic trends to see what kind of consequences
environmental and socio-economic changes. this may have. An example of dynamic spatially-explicit
IA modeling is presented in box 2.
Regional IA modeling. The majority of current IA models
consists of global models. Many of them are subdivided Box 2
Region-specific modeling using cellular automata.
into world regions, varying from 2 till 16. Notwithstanding
the dominant orientation towards global modeling, there is Let us consider a nature area that is surrounded by a num-
a growing need and interest in regional assessments, and ber of urban agglomerations. These cities have as a com-
thus regional IA models. At the regional (or local) level mon characteristic that they want to expand, which brings
the tensions between different forms of development can the nature area under intensifying pressure because of macro-
scopic socio-economic trends. There is an increasing need for
be made spatially-explicit. Regional development is char- new houses, for drastic infra-structural changes. On the other
acterised by increasing competition for space, resources and hand, there is increasing opposition against those interven-
finances between a variety of stakeholders. An Integrated tions in favour of preservation of the nature and recreational
Assessment of regional development then involves the es- function. According to the different perspectives of the vari-
timation of trade-offs between space and resources claimed ous stakeholders, the nature area could be filled in differently.
The anthropocentric perspective would lay the emphasis on
by economic development (industry, services, public trans- the extension of social and economic activities, clearing the
port), social development (houses, recreation, and private way for far-reaching infra-structural changes (road transport,
transport), and environmental development (nature preser- air traffic, harbour function and housing). The ecocentric per-
vation, quality of water, soil, air, noise pollution). spective will emphasise the important ecological and recre-
So far, spatially explicit modeling has been dominated by ational function the area has. Whereas the holistic perspective
will favour the dynamic equilibrium between social, economic
so-called GIS approaches. GIS models describe the spatial and ecological needs and interests (more houses with space
changes in a landscape in a detailed way, but they are static for infra-structural changes in combination with a nature and
and thus do not give insights into the dynamic mechanisms recreational function).
behind past and future changes. On the other hand, models Macroscopic trends could indicate that in the nature area
that endeavour to describe the dynamics of a number of 50,000 new houses would be needed. The cellular automata
model will allocate these houses on the basis of availability
relevant social, economic and ecological processes, system- of land, suitability of land, accessibility of the location, the
dynamic models, do not adequately (or not at all) describe immediate vicinity of economic and industrial activities, etc.
spatial changes. Next, the cellular automata will indicate what and where the
Recently, however, the technique of cellular automata amount of available land is, what the price and suitability
has been developed, derived from Artificial Intelligence, are, etc. In addition, information can be provided on how
to fill up space in the sense of: How will land use change?
that enables the linkage between system-dynamic and GIS What kind of ecologically valuable land will be lost? What
models. Cellular automata are cells (automata) in a grid- is the risk of surface water pollution? What type of activity
field, where all cells in the grid-field communicate with clusters will arise? Is there a tendency towards overconcentra-
each other through simple decision rules or algorithms. In tion or ghetto formation? Will economically weak activities
this way these protocols determine the dynamic spatial be- be threatened by economically strong ones? Based on this
information macroscopic trends could be adjusted, using sim-
haviour of all cells. The protocols may be based on differ- ulation models, to reconsider the suitability and attractiveness
ent paradigms. For instance, in the neo-classical economic of the specific region for new houses.
paradigm the problem can be translated as an optimisation
problem: how to use the land to achieve the highest possi- The advantage of using cellular automata is that it en-
ble level of human welfare, given differences in soil suit- ables visualisation of complex, spatial dynamics of regional
ability, climate, land use opportunities, and capital endow- development in a transparent way. The disadvantage may
ments. Other paradigms, however, take notions as equity be that it still has to be explored which processes deter-
or ecological quality as guiding principles for determining mining regional development can be captured by cellular
the decision protocols. automata, and which cannot. While for physical processes
170 J. Rotmans / Methods for IA

such as infrastructural changes the cellular approach seems signers and planners constructed future worlds behind their
appropriate, for mental processes representing human be- desks, where those future development paths were dom-
haviour (e.g., consumptive patterns), application of cellular inated by rather technocratic world views. Above all, it
automata seems more difficult. needs to be stated that these “engineering type” of scenar-
ios are often well thought and have proven their value in
Multi-agent modeling. The introduction of actors (agents) a particular context for a particular purpose. The argument
allows to model self-organisation phenomena in complex made here, however, is that these scenarios are of limited
systems. Agent-based models focus on the micro-level in- or no value if used in a broader setting, and if we want
teractions in a system, which may lead to unexpected emer- them to be actively used in the policy or societal context.
gent behaviour of the system. These models specify rules
Taking into account these lessons, in combination with
that determine agent behaviour (individual organisms, peo-
the observed shortcomings as indicated before, future chal-
ple or firms), where evolution of agents is feasible. The ad-
lenges for scenario analysis in the context of IA can be
vantages of agent-based modeling are that: (i) the agents to
identified. The ultimate challenge for IA is to develop in-
be modeled are not identical and homogeneous; (ii) the low
tegrated, balanced and “rich” scenarios that can help pru-
level interactions between agents are important; (iii) spatial
dent decision makers to understand past and current de-
location is important; and (iv) it is easier to specify rules
velopments, to anticipate the future and to evaluate policy
for the actions of individual agents than it is to specify the
options and strategies. What are the characteristic require-
system behaviour. An example of agent-based modeling is
presented in box 3. ments for these scenarios in order to fulfil a valuable role
in the decision supportive process?
Box 3 First, it is of vital importance that integrated scenar-
Region-specific modeling using cellular automata and multi- ios should be developed in a participatory way, involv-
agents. ing a representative but heterogeneous group of stakehold-
An interesting experiment is to supplement a cellular automata
ers. Second, integrated scenarios should display diversity,
model by a multi-agent approach, which allows to model self- which means that they should contain multiple perspectives,
organisation phenomena in complex systems. Linking the de- and should not be developed along the lines of narrow,
cision rules underlying cellular automata models to an agent- disciplinary-based perspectives. Economists, for instance,
based model, enables to model the regional implications of tend to think about the future solely in terms of economic
human behaviour in a spatially-explicit way.
Let us consider the same case study as in box 2. However, we costs and benefits, defining away those societal processes
now place agents in the form of decision makers in control that cannot be expressed in monetary terms.
of the nature area that may evolve in any kind of direction. Integrated scenarios also need to be comprehensive: in-
The decision makers are responsible for infrastructure, land cluding economic, environmental, institutional, and social
use, energy services and taxation. Based on the outline given
by the decision makers, the regional population, representing
dimensions. Anastasi [5] noted that especially social di-
the second type of agents, decide whether or not to build, mensions are generally overlooked in scenario activities.
where to live (so long as there are houses for them), what Because of the multitude of choices and assumptions in sce-
routes to take to travel to work, etc. The fiscal environment narios, many of which are hidden, it is important to make
encourages business to locate or move out, affecting employ- them transparent. By nature, comprehensiveness is at odds
ment. The result could be a rich and absorbing simulation
of a developing nature area, where the decision makers take with transparency, which makes it even more challenging
“top–down” decisions, and the regional population make their to reconcile them. In addition, the underlying values, bi-
“bottom–up” responses. ases, and preferences, which are often hidden, need to be
made explicit.
Integrated scenarios are consistent with regard to as-
3.2.2. Scenarios sumptions for different sectors, different problems, and
The following lessons can be learned from two decades scale levels. Playing around with different spatial and tem-
of scenario analysis (Anastasi [5]): poral scale levels is essential. Usually, the higher the spatial
• Scenarios with a strong narrative as well as a quantita- scale level, the more ambitious the policies as formulated in
tive component are able to engage stakeholders. scenarios. However, implementing those policies at a lower
• A group of individuals with different, yet complemen- scale level is another story. So exercises that make the ten-
tary, knowledge, experiences, and perspectives is needed sions explicit between the different scale levels in terms of
to provide scenario “richness”. policy formulation versus realisation, are very useful. Sim-
ilarly, many policy strategies in scenarios are formulated
• Active dissemination is needed to ensure that stakehold-
in the long-term. If it is not clearly indicated what those
ers make best use of scenarios.
long-term strategies mean in terms of short-term, concrete
• Models can play an important underpinning role, pro- policy actions, the scenario exercise is of limited value.
viding a valuable consistency check. A balanced scenario balances the narrative and the quan-
Many scenario exercises performed during the last decades titative component. Challenging scenarios do appeal to the
had an “engineering type” of character. Engineers, de- user’s imagination, and are no simple extrapolations of past
J. Rotmans / Methods for IA 171

or current trends, but include surprises (events) and turning- the various methods can be used in a complementary
points (e.g., bifurcations). way;
In sum, the challenge with regard to the development of • making the disciplinary sciences more interdisciplinary,
future generations of scenarios in the context of IA (inte- and aware of the limitations of their knowledge base
grated scenarios) is twofold: when applied in an integrated fashion to problems faced
(i) to consider the scenario process as a creative and par- by humanity.
ticipatory process in which a diversity of stakeholders
is involved; and
4. Towards a research agenda for IA methods
(ii) to consider the scenario products as helpful tools to
imagine the future, and to explore those imaginable The main challenge for the Integrated Assessment com-
futures. munity is to overcome the scientific and political credibility
barrier. The requirements imposed on IA from disciplinary
3.2.3. Risk analysis scientists on the one hand, and from policy makers on the
Risk analysis seems to be a promising tool for Inte- other, seem to be contradictory. Scientists want their disci-
grated Assessment. Notwithstanding the involvement of plinary knowledge represented in as adequate a manner as
some scholars experienced in risk analysis in the IA com- possible, and they want to use disciplinary-based standards
munity, we cannot conclude that there is a stable bridge to evaluate the quality of the building blocks of IA. Policy
between IA and risk analysis. The major challenge for makers want IA tools that are as transparent and easy to
the IA community is to reach out to the risk community use as possible (i.e., capable of exploring many possible
to learn what they have to offer in methodological terms. alternatives quickly), while retaining as much complexity
In the long term, using risk analysis’ concepts, tools and as possible.
methods may eventually inspire risk analysts to use the The resulting task is then to create the right balance be-
available disciplinary knowledge and experience on risk in tween simplicity and complexity, aggregation and disaggre-
a complementary, more integrated way. In this way, pure gation, stochastic and deterministic behaviour, quantitative
risk analysis efforts may benefit from IA. and qualitative linkages, exogenous and endogenous fac-
tors, social and natural sciences, and, last but not least,
3.3. Participatory methods between analytical and participatory IA methods.

The major challenge is to gain more experience with 4.1. Codes of practice
participatory methods in IA. The evaluation of these
exploratory experiences should lead to an overview of Unmistakably, there is a need for guidelines for good IA
strengths and weaknesses of the various approaches. Such practice to evaluate the IA process, and quality rules for IA
an overview indicates the requirements of participatory to evaluate the products delivered by IA studies. The spe-
methods for IA. Such insights should then lead to the fur- cial character of IA implies that in a certain way scientists
ther development of appropriate tools, procedures and ide- will be judged as witnesses giving testimony (Ravetz [91]).
ally, manuals that allow to test and to improve participatory The evidence for integrity is transparency. Guidelines for
methods for IA. good practice aim to make the process transparent, so that
the methods and processes of integration can be judged by
3.4. Methodological opportunities ahead an “extended peer community”, involving other IA prac-
titioners, disciplinary scientists, users and affected publics
The following items are perhaps critical to the future (Funtowicz and Ravetz [36–38]).
utility of IA: Different authors in IA have recently addressed the issue
of developing guidelines for good IA practice. Ravetz [91]
• further development, usage, and improvement of scien- states that the quality of IA is assured only by the quality
tific methods for participatory processes; of its production. Risbey et al. [92] emphasise that process-
• further development of quality assessment procedures; based criteria should be developed. Bailey et al. [9] seek
• development of protocols for aggregation and disaggre- to identify key aspects of good practice in IA. Morgan and
gation; Dowlatabadi [77] propose hallmarks of a good IA. Oth-
ers like Parson [86] and Schneider [103] in turn have in-
• making the link between different spatial and temporal
terpreted and extended these contributions. Elements that
scales to identify the role of individuals and institutions,
are commonly shared are plurality, incorporation of val-
and their systems of governance;
ues, treatment of uncertainty, participatory process, and
• procedures for active dissemination of insights to soci- dialogue.
etal stakeholders and the general public; As a matter of fact, the quality of the products is de-
• an IA tool kit that comprises a wide variety of well- pendent on the purpose and context of the study in ques-
documented tools and methods, and suggestions for how tion. It is thus by definition impossible to come up with
172 J. Rotmans / Methods for IA

general quality standards that measure IAs with the same change, which was dictated by the (multi-)disciplinary re-
yardsticks. Nevertheless, it is useful to start thinking about search agenda, rather than the policy agenda. In order
rough guidelines for measuring the quality of IA studies. to increase the number of “demand-driven” IA studies,
Building upon Robinson [94], we make a distinction be- amongst others advocated by the post-normal school of
tween three types of quality criteria: (i) analytical criteria; thought (Funtowicz and Ravetz [39]), the strength of an-
(ii) methodological criteria; and (iii) quality in terms of us- alytical IA methods should be combined with the vigour
ability. Rotmans and van Asselt [101] elaborate on these of participatory methods. As stated before, both types of
quality criteria, but within the scope of this paper we fo- methods should be used in a complementary way. There is
cus on the methodological criteria for IA. Methodological a need to develop analytical methods as models and scenar-
criteria address the quality of IA methods, tools and ap- ios in close conjunction with decision makers. This is more
proaches. Questions that are relevant with regard to these easily said than done. For instance, having decision makers
criteria are: Are the chosen methods, tools and approaches as stakeholders in the IA model design and building process
sound, appropriate and legitimate? What framework is used is desirable indeed, but may lead to high expectations that
for conducting the IA, and to what extent is this framework cannot be fulfilled by IA. If the state of scientific knowledge
integrated? How many disciplines are involved, and is the does not permit reliable estimates of geographically-explicit
study multi- or interdisciplinary? Are essential processes impacts of global climate change, scientists need to make
left out, defined away or parameterised? Is there a balance clear to decision makers the difficulty of attaining such an
between aggregation and disaggregation and is it based on unrealistic goal.
a sound aggregation protocol? How are uncertainties dealt As regards IA models, this means in practice that, al-
with? Are values, preferences and biases made explicit? ready in its conceptual phase, the IA model should be co-
What is the level of participation by what kind of stake- designed by modelers, scientists, and decision makers (and
holders? possibly by other stakeholders as well). This means in-
We should be prudent in applying these quality criteria to volvement of the community of stakeholders to establish
past and existing IA studies. Many IA studies had a rather the model’s credibility and authority during its develop-
heuristic and explorative character. They can be compared ment. Involvement includes discussions on the model’s in-
to pioneering activities such as experimental gardens where puts and outputs, the temporal and spatial scale level used,
new research methods are being tried out. Such typical the level of aggregation (detail) needed for developing pol-
R&D studies rarely work from a well-thought-out method- icy strategies, issues and processes to be included or left
ological setting and research strategy, and often lack well- out, and presentation of the model set-up and model results
defined research goals. These “supply-driven” IA studies in terms of transparency. This culminates in a so-called user
should not be measured too stringently against the above model, which helps tailor the model to the user’s needs (van
quality criteria. On the other hand, if IA wants to evolve Asselt [6]). Since this is a continuing process, an interme-
to an independent, respected research field, the sheep have diary layer of project managers may be established between
to be separated from the goats, which can only be realised the group of modelers and decision makers. Except for the
if quality criteria for IA are applied. While on its way of RAINS model (Alcamo et al. [4], Hordijk [46]), which was
developing into a mature research field, which might take partly co-designed by scientists and potential users, such
another decade or so, a compromise would be not to apply a demand-driven IA modeling process would be unprece-
specific quality rules, but a check list of “codes of prac- dented.
tice”. “Codes of practice” are rough guidelines for good In the field of risk analysis there is also a mismatch be-
practice and quality rules, which have also been success- tween demand and supply. In terms of risk demand, there
fully applied in the interdisciplinary research field of Life- is a basic need for relevant information on risks of complex
Cycle Assessment (SETAC [104]). These codes of practice problems. However, different actors do have different per-
should not be applied rigorously, but in a rather malleable spectives and thus different risk perceptions. This means
way, in order to avoid any kind of guillotine effect. that the need for information is dependent on the context
Concrete recommendation for EFIEA: within which decision makers are functioning. That also
means that what is relevant for one actor could be irrele-
We therefore propose here to establish a European task vant for another. On the other hand, the risk supply side
force that embarks on the formulation of a check list of represents the risk assessments as provided by the scien-
codes of practice that can be applied to former, current and tists. However, a plurality of scientific approaches is avail-
future IA projects in a malleable way, which will ultimately able, primarily caused by the large uncertainties surround-
result in a definitive list of quality rules and rules for good ing complex problems. Various methods are needed to cap-
practice of IA. ture different aspects of complex problems, and to produce
anything like a comprehensive understanding of the nature
4.2. Demand- and supply-driven and extent of risks.
To bring together the risk demand and supply side, par-
We have observed that the majority of IA studies is ticipatory methods could be used. Although there has been
“supply-driven”. The most prominent example is climate some experience with participatory risk analysis, the in-
J. Rotmans / Methods for IA 173

creasing complexity of problems the world faces today re-


quires a more integrated approach, balancing the demand
and supply side as much as possible.
The field of research where probably most experience
has been gained with participatory processes is that of
scenario analysis. However, much of that experience
arises from long-term planning in the business community,
within the context of strategic management. Multinationals
like Shell, for instance, have tremendous experience with
stakeholder-oriented development of scenarios. Within the
research community much can be learned from these stake-
holder approaches, but up till now there has been relatively
little exchange of scenario expertise between the research
and business community.

Concrete recommendation for EFIEA:

How can we realise a synthesis between the analytical


and participatory side of IA? It is proposed here to start up
two pilot IA projects, demand-driven and supply-driven:
• In the supply-driven IA endeavour, a group of scien-
tists anticipates the societal relevance of a complex Figure 5. Demand- and supply-driven IA.
theme, using analytical IA tools to arrive at new insights.
Through participatory methods, the first assessment can
be enriched with societal aspirations, constraints and In spite of this disadvantage, regional Integrated As-
public knowledge, after which the process can be it- sessments may form fundamentally important examples of
erated. the usefulness and applicability of IA. Unfortunately, there
have been only a few examples of integrated regional case
• A demand-driven IA pilot project is a participatory en-
deavour in which scientists, decision makers, and stake- studies, most of which are impact studies, such as the
holders explore which complex issues are highly rele- MacKenzie Basin impact study (Cohen [15]), the MINK
vant to a future society. The resulting common agenda impact study (Rosenberg et al. [96]), and the neo-classical
and explorative assessment then determine the analytical regional U.S. impact study of Mendelsohn et al. [73].
methods to be either used or developed. The scientific While many of the past regional case studies focused
assessment then informs the next round of the partici- on regional impacts of climate change, there is a need for
patory process, etc. a broader scope of regional Integrated Assessments. An
example is planning for sustainable development, which is
The two approaches are visualised in figure 5. far more complex and will require different methods and
a changed paradigm from more traditional planning. New
4.3. Regional Integrated Assessment methods will therefore have to be developed and applied,
Integrated Assessment is still on the way upwards of on the interface between the short- and long-term, the ob-
building up authority. In this credibility process, there is jective and value-laden, the quantitative and the qualita-
an urgent need for integrative case studies. Thus far, many tive, and the certain and the uncertain. Concrete ideas
IAs had a fairly abstract character, which is partly due to the for how to implement sustainable development at the re-
global scale level at which IAs often operate. Case studies, gional level are urgently needed. For Europe, two vital
however, can convincingly demonstrate the added value of entities can be considered for implementing regional sus-
IA, in particular case studies at the local or regional level. tainable development. First, the European agglomerations,
Working at a lower scale level has two advantages: (i) IAs in particular the cities and their peripheries, which hold
can focus on concrete societal problems, issues which are the key to the future spatial organisation of social, eco-
already at stake in a specific region; and (ii) IAs can be nomic and ecological changes in a sustainable way. And
made geographically-explicit, which allows for visualis- second, the European catchments, where we may expect
ing the trade-offs between problems, sectors, resources and a concentration of water-related problems. Catchments
scale levels to be made. On the other hand, a major disad- are trans-regional “hot spots” that embrace many devel-
vantage of working at a lower scale level is that the region opments: population growth, economic growth, increasing
concerned has to be “isolated” from the outside world. In- food demand, inefficient water use, water pollution, land
teraction with the outside world can be realised by speci- cover changes, climate change, and lacking infrastructure.
fying exogenously important region-transcending processes This makes catchments pre-eminently suitable for an inte-
that impact on the specific region. grated approach.
174 J. Rotmans / Methods for IA

Concrete recommendation for EFIEA: • Tools like models and scenarios can be provided with in-
teractive communication shells or user interfaces, which
Integrated case studies are needed as practical examples enhance the communication between designers and users
of Integrated Assessment. It is therefore recommended to of those tools and methods. In general, visualisation
initiate two kinds of case studies in Europe: techniques facilitate the presentation of information in a
way that corresponds to the way people intuitively per-
• One for some representative European agglomerations,
ceive images. Innovative visualisation techniques can
to analyse the competition for space and resources in a
help to make models and scenarios as transparent as
geographically-explicit way. A tool kit of new, promis-
possible, to make the underlying theories as clear as
ing IA methods should be tested for those agglomer-
possible, to provide easy-to-use interactive interfaces,
ations, such as: cellular automata, genetic algorithms,
multi-agent models, in combination with existing IA and to display uncertainties and complexity of the sys-
tems behaviour comprehensibly. This all helps to make
methods: simulation models, GIS models, and scenar-
ios. the model and scenario outcomes, as well as the com-
plex messages they want to convey, more immediately
• The other for a representative European catchment. available and accessible to a wide range of users, who
A promising example seems to be the Mediterranean may vary from the general public to decision makers.
basin, for which previous prospective studies have Examples of user interfaces are the interactive scenario
been done, e.g., the Blue Plan study (Grenon and scanner built around the IMAGE 2.0 model (Berk and
Batisse [41]), upon which we can build. For such a Janssen [12]), and the interactive user interface around
basin, an integrated water assessment is to be developed, the TARGETS model (Rotmans and de Vries [99]), the
in combination with specific basin scenarios, through latter of which has appeared on CD-ROM.
a participatory process involving stakeholders from the
Mediterranean countries. • Indicators can also be useful tools in improving the com-
munication with decision makers. Indicators can be de-
4.4. Improving IA tools and methods fined as pieces of information that help measure changes
and progress in decision making. Decision makers al-
Currently used IA tools and methods are all relatively ready have indicators in their lexicon, using them to
immature (Rotmans and van Asselt [101]). To meet the formulate policies and resulting targets to be achieved.
high expectations that surround IA, the quality of the IA Models, scenarios and risks can be linked systematically
tool kit needs to be enhanced. To this end, two strategies are to indicators. In this way, indicators serve as vehicles
proposed here. A short-term strategy, aimed at improving to communicate model, scenario and risk analysis re-
the existing tools and methods, and a long-term strategy, sults, and can be used as a basis for mapping response
directed towards the creation of new tools and methods strategies.
that supplement the current tools and methods. • A common, unambiguous language could be developed
for external communication. IA researchers need to es-
4.4.1. Short-term strategy tablish common meanings and definitions of often used
As already stated by many Integrated Assessors (e.g., but also misused notions, such as horizontal versus verti-
Parson [85–87], Schneider [103], and Rotmans and Dow- cal integration, exogenous versus endogenous processes,
latabadi [100]) there is much room for improvement of cur- aggregation versus disaggregation, various types and
rent IA tools and methods. Much of the criticism against sources of uncertainty, calibration and validation, re-
IA tools and methods has to do with their lack of trans- duced form or meta-modeling, etc.
parency, and the rather technocratic abstraction of reality
they display. The following suggestions might be helpful Concrete recommendation for EFIEA:
in overcoming this communication barrier:
Current IA tools and methods can be relatively easily im-
• Analytical methods should be documented (most of
proved, through making them more transparent to a wide
them are now poorly documented), and descriptions and
audience, varying from decision makers to the general pub-
findings should be published regularly in the open sci-
lic. Although full transparency of current IA tools and
entific and policy literature. So not only in economics-
methods is illusory, it can be enhanced through:
oriented journals, but also in environment-related jour-
nals. (i) better documentation and more frequent publication;
• Using established and well-defined research methods (ii) using established and well-defined research methods
and goals can enhance the communication with the fun- and goals;
damental research community and with the decision
(iii) developing interactive user interfaces (communication
making world. Presenting these research methods in
shells);
an open and clear way makes, for instance, models and
scenarios as well as their results both comprehensible (iv) using indicators that are linked to IA tools and meth-
and defensible. ods; and
J. Rotmans / Methods for IA 175

(v) developing a common, unambiguous IA language for ments of themselves to later generations in an effort to find
external communication. the most successful solutions. A crucial assumption is that
the degree to which the agents’ expectations meet measure-
4.4.2. Long-term strategy ments determines the fitness of the agents’ perspective.
The next generation of IA tools and methods should meet Some preliminary exercises are already underway, test-
the demands of scientists and decision makers adequately, ing various elements of this new research paradigm (Rot-
probably resulting in an even greater variety of IA tools mans and Dowlatabadi [100]). It would be extremely useful
and methods addressing the scientific issues and the policy to expand these experiments, and apply these experimental
debate at different levels. methods to a limited number of concrete case studies. Obvi-
A guiding principle for the development of new tools ously, in these experiments, the agents’ representations used
and methods for IA could be a more realistic representa- are quite abstract images of decision makers and stakehold-
tion of the social and natural system. Still, the majority of ers. Only a weak isomorphism may exist between the real-
the tools and methods we use follow the research paradigm world adaptation, and the way in which simulated agents
that is based on equilibria, linearity, and determinism. The adapt to their changing environment. Therefore, it may take
result is that incremental changes in parts of the subsystems a while before the application of complex, adaptive systems
we analyse will cause gradual and incremental changes in theory has matured to real world problems. Nevertheless,
the system as a whole. Unfortunately, the world does not the first preliminary experiments, which deal with the no-
function in such a simple, linear way. The real world shows tion of surprises and flexible strategies to respond to them,
strongly non-linear, stochastic, complex, and chaotic behav- are encouraging and already produce interesting results.
iour. This implies that incremental changes in conditions
of subsystems may result in considerable changes in the Concrete recommendation for EFIEA:
results of the overall system, which may not be predicted
beforehand. Therefore the overall challenge for the next generation
Neo-classical economics, for instance, is based on the of IA tools and methods is to incorporate insights emerging
prevailing theory of general equilibrium. It supposes that from new scientific streams such as the areas of complex
the free market brings about an efficient resource allocation, systems, adaptive behaviour, bifurcations, and ignorance
except for a few “market failures” which can be remedied about systems. Incorporation of these insights will broaden
by corrective taxes or subsidies. But without taking into the scope of IA tools and methods and considerably im-
account the possibility of multiple equilibria, the complex prove the quality of IA. In addition, the incorporation of
dynamics of price adjustments, and the role of positive feed- cultural values, biases and preferences, which determine
backs such as “learning by doing” which makes established the manner in which complex problems are perceived and
technologies considerably cheaper over time, complex is- formalised, will allow the IA community to inform the de-
sues such as the role of induced technological change can cision making process in a more balanced way.
hardly be discussed. An alternative, more realistic way
of describing the contemporary world economy would be 4.5. Entering new problem areas
to consider multiple equilibria, in the form of a transition
of one economic equilibrium to another. New approaches IA has contributed to the climate change debate in ex-
are underway that take into account multiple equilibria and ploring impacts of climate change, mitigation and abate-
positive feedbacks. However, they do not yet provide a ment strategies, issues in co-operative implementation, the
coherent, comprehensive picture of the economy, still pro- likely equity effects of candidate policies, and complicat-
ducing equivocal results. ing factors such as aerosols. IA models have also provided
A new and promising research paradigm that has information on balancing the carbon budget and on vari-
emerged during the last decade is that of the complex, adap- ous integrated aspects of land use. In the past, IA has also
tive systems approach. Complex adaptive systems are com- proven to be useful in the field of acid rain, where the IA
posed of many agents who interact with their environment model RAINS played a guiding and supportive role in the
and can adapt to changes. These systems organise them- negotiating process of emission reductions of SO2 .
selves, learn and remember, evolve and adapt. Evolutionary Now the time seems ripe for entering new problem ar-
models, for instance, focus on social and natural systems as eas from the perspective of IA. In doing this, the notion of
dynamically interwoven, where impacts on natural systems transitions could be a helpful guide. A transition may be
feed back to the social system, and vice versa. Both systems defined as a gradual, continuous shift in society from one
continuously adapt to human-induced perturbations of the “mode of operation” to another. Transitions are described
system as a whole. This evolutionary approach is applied by the following phases: (i) a take-off phase; (ii) an acceler-
in various disciplines and uses a whole range of methods ation phase, characterised by instability due to rapid techno-
and tools, for instance, genetic algorithms. A genetic al- logical, social and environmental change; during this stage
gorithm is a robust problem-solving approach based on the societies and environmental quality are highly vulnerable
mechanics of the survival of the fittest. After reproduction, to damage; and (iii) a stabilisation phase, in which the pace
crossover and mutation, the fittest solutions pass on ele- of change slows and a new form of equilibrium is reached.
176 J. Rotmans / Methods for IA

Thus in system-dynamic terms, transitions suppose shifts decades), but also at different spatial levels: at the level of
from one relative equilibrium to another, moving from slow urban agglomeration (local sustainable development), at the
to rapid change, before reaching a different level of stability. national level (contribution to national emission profiles),
Using a transitional approach requires an analysis which and at the European level (contribution to the European
goes beyond the traditional cost–benefit approach. Such a emission targets as formulated in protocols for acidification
cost–benefit approach may be useful, but only for a certain and climate change).
domain, for relatively small departures from a given situ- Similarly, other fields that seem promising from an inte-
ation. Transitions, however, require major shifts from one grative perspective are: technological development, health
equilibrium to another, which makes cost–benefit analyses care, and infrastructure.
inappropriate (apart from the general methodological prob-
lems with measuring intangible impacts in monetary terms). Concrete recommendation for EFIEA:
One of the ultimate challenges for IA is to describe
complex societal dynamics in terms of a family of inter- We recommend to enter the following new problem ar-
connected transitions: social, economic, environmental and eas: water, technological development, health care, trans-
institutional transitions. Such a transitional approach opens port, and infrastructure.
the door to new problem areas, of which only a few will
be given here.
The water problem, for instance, lends itself to an inte- 5. Summary
grated transitional approach. The water transition can be
described in terms of increasing exploitation of resources Agenda for IA methodologies.
and increasing water pollution from industrialisation and
Activity Time span
urbanisation during the early stages of development. As
countries industrialise, the share of water used for agricul- Establishment of European task force Short-term
ture declines and an increasing proportion is taken by the IA pilots: Short-term
industrial, commercial and household sectors – domestic • supply-driven
consumption in particular increases greatly with more afflu- • demand-driven
ent lifestyles. This is followed by a period of more efficient
water use and water quality improvement, also induced by Integrated case studies: Medium-term
sophisticated water technologies. • European agglomerations
The water problem can be methodologically tackled by a • European catchments
transitional approach, considering the water transition in the
Improvement of current tools and methods Short-term
light of socio-economic and institutional transitions, start-
New generation of IA tools and methods Long-term
ing at a global level (through the influence of global cli-
(keywords: complex systems, adaptive behav-
mate change on water availability), descending to the re- iour, surprises, ignorance, cultural values, bias
gional level, where “hot spots” can be defined, down to the and preferences)
river basin or catchment level, which is a suitable method- New problem areas: Medium- and long-term
ological entity for practical implementation of a transitional
• water
approach.
• technological development
Another issue that seems appropriate for an integrated • health care
approach is transport. The transport problem comprises • transport
many dimensions, i.e., spatial, economic, ecological, in- • infrastructure
stitutional, and social/psychological. All dimensions are
heavily interrelated. The economic dimension concerns the
financial management of the problem: subsidising transport References
or following the market principles. This is related to the
institutional dimension, involving the co-ordinating aspects [1] J. Adams, Risk (UCL Press, London, 1995).
of the problem: public management versus market-based [2] J. Alcamo, IMAGE 2.0: Integrated Modeling of Global Climate
intervention. This relates to the social/psychological di- Change (Kluwer, Dordrecht, 1994).
[3] J. Alcamo, E. Kreileman and R. Leemans, Global models meet
mension that deals with equity and social cohesion versus
global policy, Global Environmental Change 6(4) (1996).
individualisation. This depends on the spatial organisation [4] J. Alcamo, R. Shaw and L. Hordijk, The RAINS Model of Acid-
of transport, creating compact (concentration) versus dif- ification: Science and Strategies in Europe (Kluwer, Dordrecht,
fuse (zones) cities, which is related to the ecological com- 1990).
ponent, addressing environmental pollution and quality of [5] C. Anastasi, Lessons learned from two decades of scenario de-
life at different spatial levels. velopment, Open meeting of the Human Dimensions of Global
Environmental Change Research Community, Laxenburg, Austria.
The transport problem operates at different temporal lev- [6] M.B.A. van Asselt, Global integrated assessment models as pol-
els: short-term (congestion, air pollution) versus long-term icy support tools: A triangular approach, University of Twente,
(investments in transport infrastructure over a period of Enschede, The Netherlands (1994).
J. Rotmans / Methods for IA 177

[7] M.B.A. van Asselt and J. Rotmans, Uncertainty in perspective, Forum, Brussels (1996).
Global Environmental Change 6(2) (1996) 121–157. [30] EC-DGXVII, Energy in Europe: European energy to 2020 – A sce-
[8] M.B.A. van Asselt, C.A.M.H. Storms, N. Rijkens-Klomp and nario approach, European Commission, Luxembourg (1996).
J. Rotmans, An overview and assessment of the last decade of [31] ECN, Energy scenarios for a changing Europe: Integration versus
European scenario studies, ICIS Working Paper 98-01 (1998). fragmentation, ECN, Petten, The Netherlands (1995).
[9] P. Bailey, C. Gough, M. Chadwick and G. McGranahan, Meth- [32] J. Edmonds, H. Pitcher, N. Rosenberg and T. Wigley, Design for the
ods for integrated environmental assessment: Research directions Global Change Assessment Model, in: Integrative Assessment of
for the European Union, Stockholm Environment Institute (SEI), Mitigation, Impacts and Adaptation to Climate Change, Laxenburg,
Stockholm, Sweden (1996). Austria (1994).
[10] BCI, Economic vision on logistic and industrial junction Venlo, [33] J. Edmonds and J.M. Reilly, Global Energy: Assessing the Future
Buck Consultants International, Nijmegen, The Netherlands (1996) (Oxford University Press, Oxford, 1985).
(in Dutch). [34] B. Fischhoff, S.R. Watson and C. Hope, Defining risk, Policy Sci-
[11] U. Beck, Risk Society: Towards a New Modernity (Sage, London, ences 17 (1984) 123–139.
1986). [35] S.O. Funtowicz and J.R. Ravetz, Uncertainty and Quality in Science
[12] M.M. Berk and M.A. Janssen, The interactive scenario-scanner for Policy (Kluwer, Dordrecht, 1990).
(ISS), RIVM report 981508005, Bilthoven, The Netherlands (1977). [36] S.O. Funtowicz and J.R. Ravetz, Risk management as a post-normal
[13] G.D. Brewer, Methods for synthesis: Policy exercises, in: Sustain- science, Risk Analysis 12(1) (1992) 95–97.
able Development of the Biosphere, eds. W.C. Clark and R.E. Munn [37] S.O. Funtowicz and J.R. Ravetz, Three types of risk assessment and
(Cambridge University Press, Cambridge, 1986) pp. 455–473. the emergence of post-normal science, in: Social Theories of Risk,
[14] K. Button, The future of European transport, in: The Future of eds. S. Krimsky and D. Golding (Greenwood, Westport, 1992).
Transportation and Communication: Visions and Perspectives from [38] S.O. Funtowicz and J.R. Ravetz, Science for the post-normal age,
Europe, Japan and the USA, ed. R. Thord (Springer, Berlin, 1993). Futures 25(7) (1993) 739–755.
[15] S.J. Cohen, Scientist–stakeholder collaboration in integrated assess- [39] S.O. Funtowicz and J.R. Ravetz, The worth of a songbird: Ecolog-
ment of climate change: Lessons from a case study of Northwest ical economics as a post-normal science, Ecological Economics 10
Canada, Environmental Modeling and Assessment 2(4) (1997) 281– (1994) 197–207.
293. [40] G. Gallopin, A. Hammond, P. Raskin and R. Swart, Branch points:
[16] J. Conrad, Society and problem-oriented research: On the socio- Global scenarios and human choice – A resource paper of the
political functions of risk assessment, in: Risk: A Seminar Series, Global Scenario Group, Polestar Series Report no. 7, Stockholm
ed. H. Kunreuther, IIASA, Laxenburg, Austria (1981). Environment Institute, Sweden (1997).
[17] R. Costanza and M. Ruth, Dynamic systems modeling for scoping [41] M. Grenon and M. Batisse, Futures for the Mediterranean Basin
and consensus building, in: European Chapter of the International (Oxford University Press, Oxford, 1991).
Society for Ecological Economics, University of Versailles, Paris, [42] J. Grin and R. Hoppe, Toward a comparative framework for learn-
France. ing from experiences with interactive technology assessment, In-
[18] R. Costanza and S.S. Tognetti, eds., Ecological Economics and dustrial and Environmental Crisis Quarterly 9(1) (1995) 99–120.
Integrated Assessment: Participatory Process for Including Equity, [43] E.G. Guba and Y.S. Lincoln, Fourth Generation Evaluation (Sage,
Efficiency and Scale in Decision Making for Sustainability (SCOPE, Newbury Park, 1989).
Paris, 1998, in press). [44] A. Hausrath, Venture Simulation in War, Business and Politics
[19] V. Covello, Uses of social and behavioural research on risk, Envi- (McGraw-Hill, New York, 1971).
ronment International 4 (June 1984) 17–26. [45] C.S. Holling, Adaptive Environmental Assessment and Management
[20] K.K. Cox and J.B. Higginbotham, Application of focus group in- (Wiley, London, 1978, revision 1990).
terview in marketing, Journal of Marketing 40 (1976) 77–80. [46] L. Hordijk, Use of the RAINS model in acid rain negotiations in
[21] E.A.C. Crouch and R. Wilson, Risk/Benefit Analysis (Ballinger, Europe, Environmental Science and Technology 25(4) (1991) 596–
Cambridge, 1982). 603.
[22] C.E. van Daalen, W.A.H. Thissen and M.M. Berk, in: The Delft [47] IPCC, Climate Change: The IPCC Scientific Assessment (Cam-
Process: Experiences with a Dialogue between Policy Makers and bridge University Press, Cambridge, 1990).
Global Modelers, eds. R. Leemans and J. Alcamo (1997). [48] IPCC, Climate Change 1992 – The Supplementary Report to The
[23] M. Douglas and A. Wildavsky, Risk and Culture: Essays on the IPCC Scientific Assessment (Cambridge University Press, Cam-
Selection of Technical and Environmental Dangers (University of bridge, 1992).
California Press, Berkeley, 1982). [49] C.C. Jaeger, T. Barker, O. Edenhofer, S. Faucheux, J.-C. Hourcade,
[24] H. Dowlatabadi, Sensitivity of climate change mitigation estimates B. Kasemir, M. O’Connor, M. Parry, I. Peters, J. Ravetz and J. Rot-
to assumptions about technical change, Energy Economics (1998, mans, Procedural leadership in climate policy: a European task,
accepted). Global Environmental Change 7 (September 1997).
[25] H. Dowlatabadi and M.G. Morgan, Integrated assessment of climate [50] C.C. Jaeger, M. Chadwick, B. Wynne, S. Funtowicz, M. Giaoutzi,
change, Science 259 (1993) 1813–1814. S. Giner, F. Toth, J. Jäger, G. Dürrenberger, J.R. Ravetz and C. Ca-
[26] H. Dowlatabadi and M.G. Morgan, A model framework for inte- silli, ULYSSES – Urban lifestyles, sustainability and integrated en-
grated studies of the climate problem, Energy Policy (March 1993) vironmental assessment – RTD proposal for the European Commu-
209–221. nity Framework Programme IV, Darmstadt Technical University,
[27] C. Ducot and H.J. Lubben, A typology for scenarios, Futures 12 Darmstadt, Germany (1995).
(1980) 15–57. [51] W. Jager, M.B.A. van Asselt, J. Rotmans, C.A.J. Vlek and C. Boodt,
[28] D. Dürrenberger, J. Behringer, U. Dahinden, A. Gerger, B. Kasemir, Consumer behaviour: A modeling perspective in the context of inte-
C. Querol, R. Schüle, D. Tabara, F. Toth, M. van Asselt, D. Vas- grated assessment of global change, 461502017, RIVM, Bilthoven,
silarou, N. Willi and C. Jaeger, Focus groups in integrated as- The Netherlands (1997).
sessment: A manual for a participatory tool, ULYSSES working [52] W. Jager, M.B.A. van Asselt, J. Rotmans and C.A.J. Vlek, Con-
paper WP-97-2, Darmstadt University of Technology, Darmstadt, sumer behaviour, a modeling perspective in the context of inte-
Germany (1997). grated assessment of global change, in: Sustainability in Question:
[29] EC-DGXI, Vision 2020: Scenarios for a sustainable Europe, XI- The Search for a Conceptual Framework, eds. J. Köhn, J. Gowdy,
120-97, European Commission, Directorate General XI, Environ- F. Hinterberger and J. van der Straaten (Edward Elgar Publishers,
ment, Nuclear Safety and Civil Protection, General Consultative Abingdon, 1998).
178 J. Rotmans / Methods for IA

[53] S. Jasanoff, Bridging the two cultures of risk analysis, Risk Analy- [77] M.G. Morgan and H. Dowlatabadi, Learning from integrated as-
sis 13(2) (1993) 123–129. sessment of climate change, Climatic Change 34 (1996) 337–368.
[54] B.B. Johnson and V.T. Covello, The Social and Cultural Construc- [78] W.D. Nordhaus, The Efficient Use of Energy Resources (Yale Uni-
tion of Risk: Essays on Risk Selection and Perception (D. Reidel, versity Press, New Haven, 1979).
Dordrecht, 1987). [79] W.D. Nordhaus, The DICE model: Background and structure of a
[55] H. Jungermann, Psychological aspects of scenarios, in: Environ- dynamic integrated climate economy, Yale University, USA (1992).
mental Impact Assessment, Technology Assessment, and Risk Analy- [80] W.D. Nordhaus, Managing the Global Commons: The Economics
sis, eds. Covello et al., NATO ASI Series, Vol. G4 (Springer, Berlin, of Climate Change (MIT Press, Cambridge, MA, 1994).
1985). [81] T. O’Riordan, Risk perception studies and policy priorities, Risk
[56] H. Jungermann, Risk, in: Technology Controversy. Actual Key No- Analysis 2(2) (1982) 95–100.
tions for Public Debates, eds. H. Schütz and W.P.M. (IMK, Frank- [82] T. O’Riordan, The impact of EIA on decision making, in: En-
furt am Main, 1993) (in German). vironmental Impact Assessment, Technology Assessment and Risk
[57] H. Kahn and A. Wiener, The Year 2000 (MacMillan, New York Analysis, eds. V.T. Covello, J.L. Mumpower, P.J.M. Stallen and
1967). V.R.R. Uppuluri (Springer, Berlin, 1985).
[58] E. Kämper, Sociology of risk: Implications for the analysis of [83] H.J. Otway and D. von Winterfeldt, Beyond acceptable risk: On
environmental policy, European University Institute, Florence, Italy the social acceptability of technologies, Policy Sciences 14 (1982)
(1998). http://www.iue.it/WGES/ISS17/kaemper.h. 247–256.
[59] B. Kasemir, J. Behringer, B. de Marchi, C. Deuker, D. Durren- [84] H.J. Otway and D. von Winterfeldt, Risk management and accept-
berger, S. Funtowicz, A. Gerger, M. Giaoutzi, Y. Haffner, M. Nill- able risk criteria, in: Risk: A Seminar Series, ed. H. Kunreuther,
son, C. Querol, R. Schule, D. Tabara, M. van Asselt, D. Vassilarou, IIASA, Laxenburg, Austria (1982).
N. Willi and C. Jaeger, Focus groups in integrated assessment: The [85] E.A. Parson, A global climate change policy exercise: Results of
ULYSSES pilot experience, ULYSSES WP-97-4, Darmstadt Uni- a test run, July 27–29, 1995, WP-96-90, International Institute for
versity of Technology, Darmstadt, Germany (1997). Applied Systems Analysis (IIASA), Laxenburg, Austria (1996).
[60] B. Kasemir, M.B.A. van Asselt, G. Dürrenberger and C.C. Jaeger, [86] E.A. Parson, How should we study global environmental problems:
Integrated assessment: Multiple perspectives in interaction, Inter- A plea for unconventional methods of assessment and synthesis,
national Journal of Environment and Pollution (special issue “Meth- WP-96-157, International Institute for Applied Systems Analysis
ods and Models for Decision Support”) (1998, accepted). (IIASA), Laxenburg, Austria (1996).
[87] E.A. Parson, Three dilemmas in the integrated assessment of cli-
[61] P. Kassler, Energy for development, in: Shell Selected Paper (Shell,
mate change, Climatic Change 34 (1996) 315–326.
London, 1994).
[88] E.A. Parson and K. Fisher-Vanden, Integrated assessment of global
[62] S. Krimsky and D. Golding, Social Theories of Risk (Praeger, West-
climate change, Annual Review of Energy and the Environment 22
port, 1992).
(1997, submitted).
[63] D.A. Lashof and D.A. Tirpak, Policy options for stabilizing global
[89] Platform on Science and Ethics, Workshop reports on the societal
climate, U.S. Environmental Protection Agency, Washington, USA
debate on genetic research, W43, Rathenau Institute, The Hague,
(1989).
The Netherlands (1995).
[64] J. Linnerooth-Bayer and M. Thompson, Risk and governance: Pro-
[90] R.H.J. Prinn, A. Sokolov, C. Wand, X. Xiao, Z. Yang, R. Eckaus, P.
posal to the European Science Foundation, IIASA, Laxenburg, Aus-
Stone, D. Ellerman, J. Melillo, J. Fitzmaurice, D. Kicklighter and
tria (1997).
Y. Liu, Integrated global system model for climate policy analysis:
[65] N. Luhmann, Risk: A Sociological Theory (De Gruyter, Berlin,
model framework and sensitivity studies, Global Change Center,
1993).
Massachusetts Institute of Technology, USA (1996).
[66] MacCallaway, Personal communication (1998).
[91] J.R. Ravetz, Integrated Environmental Assessment Forum: Devel-
[67] W.J.M. Martens, Health impacts of climate change and ozone de- oping guidelines for good practise, ULYSSES WP-97-1, Darmstadt
pletion: An eco-epidemiological approach, Ph.D. thesis, Maastricht University of Technology, Darmstadt, Germany (1997).
University, Maastricht, The Netherlands (1997). [92] J. Risbey, M. Kandlikar and A. Patwardhan, Assessing integrated
[68] W.J.M. Martens, Health and Climate Change: Modeling the Im- assessments, Climatic Change 34 (1996) 369–395.
pacts of Global Warming and Ozone Depletion (Earthscan Publi- [93] J.B. Robinson, Life in 2030: Exploring a Sustainable Future for
cations, London, 1998). Canada (UBC Press, Vancouver, 1996).
[69] D.G. Mayo and R.D. Hollander, Acceptable Evidence: Science and [94] J.B. Robinson, Of maps and territories: The use and abuse of socio-
Values in Risk Management (Oxford University Press, New York, economic modelling in support of decision making, Technological
1991). Forecasting and Social Change 42 (1992) 147–164.
[70] A. Mazur, Societal and scientific causes of the historical develop- [95] Root and S. Schneider, Ecology and climate: Research strategies
ment of risk assessment, in: Society, Technology and Risk Assess- and implications, Science 269(52) (1995) 334–341.
ment, ed. J. Conrad (London, 1980). [96] N.J. Rosenberg, P.R. Crosson, K.D. Frederick, W.E. Easter-
[71] D.H. Meadows, D.L. Meadows, J. Randers and W.W. Behrens, The ling, M.S. McKenny, M.D. Bowes, R.A. Sedjo, J. Darmstadter,
Limits to Growth (Universe Books, New York, 1972). L.A. Katz and K.M. Lemon, The MINK methodology: background
[72] D.L. Meadows, D.H. Meadows and J. Randers, Dynamics of and baseline, in: Towards an Integrated Impact Assessment of Cli-
Growth in a Finite World (Wright-Allen Press, Cambridge, 1972). mate Change: The MINK Study, ed. N.J. Rosenberg (Kluwer, Dor-
[73] R. Mendelsohn, W. Morrison, M. Schlesinger and N. Andronova, drecht, 1993).
Country-specific market impacts of climate change, unpublished [97] J. Rotmans, IMAGE: An Integrated Model to Assess the Greenhouse
manuscript (1996). Effect (Kluwer, Dordrecht, 1990).
[74] M.W. Merkhofer, Decision Science and Societal Risk Management [98] J. Rotmans, Integrated visions for a sustainable Europe: An inte-
(D. Reidel, Boston, MA, 1987). grated assessment proposal, European Union, DG XII, Work Pro-
[75] I. Mintzer, A matter of degrees: The potential for controlling the gramme Environment and Climate (second phase), Brussels, Bel-
greenhouse effect, Research report No. 5, World Resource Institute, gium (1997).
Washington, USA (1987). [99] J. Rotmans and H.J.M. de Vries, Perspectives on Global Change:
[76] G.M. Morgan and M. Henrion, Uncertainty – A Guide to Dealing The TARGETS Approach (Cambridge University Press, Cambridge,
with Uncertainty in Quantitative Risk and Policy Analysis (Cam- 1997).
bridge University Press, New York, 1990). [100] J. Rotmans and H. Dowlatabadi, Integrated assessment of climate
J. Rotmans / Methods for IA 179

change: Evaluation of methods and strategies, in: Human Choice in the city region, Town country planning association (TCPA),
and Climate Change: An International Social Science Assess- Manchester, UK (1996).
ment, eds. S. Rayner and E. Malone (Battelle Press, Washington, [111] F.L. Toth, Policy exercises, Simulation and Games 19 (September
1998). 1988) 235–276.
[101] J. Rotmans and M.B.A. van Asselt, Integrated assessment: Current [112] UN-DPCSD, Critical trends: Global change and sustainable de-
practises and challenges for the future, in: Ecological Economics velopment, Department for Policy Co-ordination and Sustainable
and Integrated Assessment: A Participatory Process for Including Development, United Nations, New York, USA (1997).
Equity, Efficiency and Scale in Decision Making for Sustainability, [113] C.A.J. Vlek and P.J. Stallen, Rational and personal aspects of risk,
eds. R. Costanza and S. Tognetti (SCOPE, Paris, 1998, in press). Acta Psychologica (1980) 273–300.
[102] H.-J. Schellnhuber, A. Block, M. Cassel-Gintz, J. Kropp, G. Lam- [114] H.J.M. de Vries, SusClime, Globo report series 11, RIVM, Bilt-
mel, W. Lass, R. Lienenkamp, C. Loose, M.K.B. Lüdeke, O. Mold- hoven, The Netherlands (1995).
enhauer, G. Petschel-Held, M. Plöchl and F. Reusswig, Syndromes [115] VROM, The Netherlands in 2030 – Exploration of spatial per-
of global change, Gaia 6(1) (1997) 19–34. spectives, Ministry of Housing, Planning and the Environment
[103] S. Schneider, Integrated assessment modeling of climate change: (VROM), The Hague, The Netherlands (1997) (in Dutch).
Transparent rational tool for policy making or opaque screen hid- [116] WEC and IIASA, Global energy perspectives to 2050 and beyond,
ing value-laden assumptions?, Environmental Modeling and As- WEC, London, UK (1995).
sessment 2(4) (1997) 229–250. [117] M.C. Weinstein, P.G. Coxson, L.W. Williams, T.M. Pass, W.B. Sta-
[104] SETAC, Guidelines for life-cycle assessment: A “code of practice”, son and L. Goldman, Forecasting coronary heart disease incidence,
Report of SETAC Workshop held at Sesimbra, Portugal, 31 March– mortality and cost: the coronary heart policy model, American
3 April 1993, Society of Environmental Toxicology and Chemistry Journal of Public Health 77 (1987) 1417–1426.
(SETAC) (1993). [118] J. Weyant, O. Davidson, H. Dowlatabadi, J. Edmonds, M. Grubb,
[105] P. Slovic, B. Fischhoff and S. Lichtenstein, Rating the risks: The E.A. Parson, R. Richels, J. Rotmans, P. Shukla, R.S.J. Tol, W. Cline
structure of expert and lay perception, in: Environmental Im- and S. Fankhauser, Integrated assessment of climate change: An
pact Assessment, Technology Assessment and Risk Analysis, eds. overview and comparison of approaches and results, in: Economic
V.T. Covello, J.L. Mumpower, P.J.M. Stallen and V.R.R. Uppuluri and Social Dimensions of Climate Change, eds. J.P. Bruce, H. Lee
(Springer, Berlin, 1985). and E.F. Haites (IPCC, Cambridge University Press, Cambridge,
[106] J.P. van der Sluijs, Anchoring amid uncertainty, Utrecht University, 1996).
Utrecht, The Netherlands (1997). [119] C. Whipple, Opportunities for the social sciences in risk analysis,
[107] D. Smith, Eurofutures (Capstone, Oxford, 1997). in: Environmental Impact Assessment, Technology Assessment and
[108] C. Starr, Social benefit versus technological risk, Science 165 Risk Analysis, eds. V.T. Covello, J.L. Mumpower, P.J.M. Stallen
(1969) 1232–1238. and V.R.R. Uppuluri (Springer, Berlin, 1985).
[120] WHO, Health future research, World Health Statics Quarterly
[109] R. Swart, Climate change: Managing the risks, Free University,
47(3/4) (1994).
Amsterdam, The Netherlands (1994).
[110] TCPA and CER, Manchester 2020: Sustainable development

View publication stats

You might also like