ARTICLE IN PRESS Global Environmental Change 16 (2006) 170–181 www.elsevier.com/locate/gloenvcha Science-based stakeholder dialogues: Theories and tools Martin Welpa,, Anne de la Vega-Leinerta, Susanne Stoll-Kleemannb, Carlo C. Jaegera a Potsdam Institute for Climate Impact Research (PIK), Department of Global Change & Social Systems, P.O. Box 601203, 14412 Potsdam, Germany b Humboldt University of Berlin, Institute of Agricultural Economics and Social Sciences, Luisenstr. 53, 10099 Berlin, Germany Received 19 October 2004; received in revised form 12 November 2005; accepted 11 December 2005 Abstract Science-based stakeholder dialogues are structured communication processes linking scientists with societal actors, such as representatives of companies, NGOs, governments, and the wider public. Stakeholders possess knowledge needed by scientists to better comprehend, represent and analyse global change problems as well as decision-makers’, managers’ and other stakeholders’ mental models. We will examine the relevance of three theoretical frameworks for science-based stakeholder dialogues in the context of sustainability science. These are Rational Actor Paradigm, Bayesian Learning and Organisational Learning. All three contribute to a better theoretical framework for dialogue practice and the understanding of stakeholders as actors in society and in research in particular. Furthermore, these theories are important for tool development. A combination of analytical and communication tools is recommended to facilitate stakeholder dialogues. The paper refers to examples of dialogue practice gained in the European Climate Forum (ECF). r 2006 Elsevier Ltd. All rights reserved. Keywords: Stakeholder dialogues; Participation; Global change research 1. Introduction Today’s economic, social and environmental problems are increasingly complex and global in nature. Climate change, loss of biodiversity and poverty in the South, illustrate problems where causes and effects are often distant in time and space. This complexity challenges the capacity of humankind to learn from past experiences and, maybe most importantly, to create a shared vision of a desired world. Sustainability science seeks to understand the dynamics of global change, i.e., the fundamental character of interactions between nature and society. It also seeks to explore ways to collectively create a sustainable world (Kates et al., 2001; Senge, 2003). Science has an important role to play in a sustainability transition. Meeting the needs of the future world populaCorresponding author. Tel.: +49 331 288 2619; fax: +49 331 288 2600. E-mail addresses: martin.welp@pik-potsdam.de (M. Welp), delavega@pik-potsdam.de (A. de la Vega-Leinert), susanne.stoll-kleemann@agrar.hu-berlin.de (S. Stoll-Kleemann), carlo.jaeger@pik-potsdam.de (C.C. Jaeger). 0959-3780/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.gloenvcha.2005.12.002 tion (with emphasis on reducing hunger and poverty), while maintaining the planet’s life-support systems is at the core of such a transition. Science can identify critical drivers of global change and plausible risks, such as the shutdown of the North Atlantic Thermohaline Circulation (e.g. Rahmstorf and Zickfeld, 2005). Furthermore, science can analyse and model impacts and vulnerabilities, such as extreme weather events (e.g. Christensen and Christensen, 2003). Although uncertainties of various kinds remain, we already know a substantial amount about the dynamics between nature and society. In terms of solutions to global change problems, however, much remains to be done. We simply do not know how world-wide species extinction could be halted or how to foster a transition of global energy systems. Sustainability science aims at exploring potential solutions to such problems (Kates et al., 2001). When embedded in a transdisciplinary context, sustainability science can play an important role in finding workable solutions for mitigating, and adapting to, global change. However, when detached from the ‘real world’ (e.g. from lifestyles, technological innovations, expectations and mental models of actors), it may remain a purely academic ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 endeavour with little social relevance. Therefore, science needs to have access to the insights and expertise of different societal actors and incorporate their knowledge bases. On the other hand, scientists need to communicate the results of their inquiries in a comprehensible way. In recent years, research institutions have become increasingly involved in science-based stakeholder dialogues. This has been partly driven by researchers themselves, but also to a great extent by funding agencies and the general public’s demand for greater accountability in science. The objectives of scientific dialogues have, however, often remained unclear. Furthermore, the absence of a theoretical framework has hampered the practice of scientific dialogues and the development of appropriate tools, including both communication and analytical tools. The objectives of this paper are: (a) to describe the specifics of science-based stakeholder dialogues and the differences to other types of dialogues, and (b) to examine the relevance of three theoretical frameworks for sciencebased stakeholder dialogues. The three theoretical frameworks are the Rational Actor Paradigm, Bayesian Learning, and Organisational Learning. We proceed in this paper as follows: in Section 2, we discuss the motivation for science-based stakeholder dialogues: Why are such dialogues needed from the point of view of the general public, funders and researchers themselves. In Section 3, we compare scientific dialogues with other types of stakeholder dialogues, in particular policy dialogues, multistakeholder dialogues and corporate dialogues. Although the methods applied in these different types of dialogues may be similar, there are differences in the objectives of stakeholder involvement. The relevance of theory in general and of the three selected theoretical frameworks in particular for science-based dialogues is discussed in Section 4. In the discussion part (Section 5), we link the theoretical frameworks with each other. The paper is largely based on practical experience gained in the process of establishing the European Climate Forum (ECF). ECF1 is a non-profit organisation committed to facilitating dialogue between scientists in the field of climate change, energy and integrated assessment on the one hand and various stakeholders including corporations, small and medium-sized enterprises (SMEs), non-governmental organisations (NGOs), policy makers and citizens. Thus the paper is, besides a theoretical inquiry, also a result of action research, in which observations have nurtured conceptual thinking and vice versa2 (Reason, 2002). 2. The need for science-based stakeholder dialogues The practice of science-based stakeholder dialogues can be seen as an effort to link different domains of discourse. 1 European Climate Forum. See www.european-climate-forum.net We are grateful for the inspiring discussions with our colleagues, who also participated in the PIK ‘stakeholder task force’, Antonella Battaglini and Diana Runge. 2 171 Domains of discourse are contexts needed for reasonably coherent exchange of arguments (Jaeger, 2003). The ones used by scientific communities and by society at large are, and should be, diverse and pluralistic. Too often, however, the domains of discourse are also disconnected from each other. Although the ‘scientific method’ is often seen as a guarantee for the quality of arguments, the pragmatist philosopher Rorty sees science just as one manner of talking among others not distinguished or privileged in any way (Rorty, 1991; Pihlström, 2001). Accordingly, ‘science talk’ cannot claim to have a monopoly on the right way of knowing. Other ways of talking, such as reflection on personal intuition, art, etc. are equally ‘true’. We do not defend a solely relativistic view on knowing, such as that of radical constructivists (Maturana and Varela, 1998). More useful seems to be Putnam’s (2002) view, proposing that it is sometimes useful to distinguish between factual claims and value judgements. However, a sharp ‘fact/value’ dichotomy becomes harmful when identified with a dichotomy between the objective and the purely subjective. Putnam argues that we can share rational arguments about values. Building on Putnam’s views we believe that conversation and the exchange of arguments become crucial at the interface of science and society, in particular, when dealing with the complex problems related to global environmental change. The need for science-based stakeholder dialogues is on the one side motivated by the general public’s call for greater accountability of science. For example, climate change research in general and the Intergovernmental Panel on Climate Change (IPCC) in particular have been criticised for their narrow professional milieu. The Economist called for an extended peer review and stronger stakeholder involvement especially when developing CO2 emission scenarios (Economist, 2003). Increasingly funding agencies also require that stakeholders are consulted in different phases of research projects. From the point of view of researchers, science-based stakeholder dialogues are needed for at least four reasons (Table 1). First, stakeholders can play an important role in identifying socially relevant and scientifically challenging research questions. Surveying stakeholders’ needs can be done systematically by means of questionnaires (e.g. Bärlund and Carter, 2002). Alternatively, new research questions may emerge in brainstorming sessions and small groups, where group dynamics and reflection boost the creative process. Second, scientists need a ‘reality check’ for the research they are doing. Dialogues with stakeholders can provide such a check. Stakeholders may be actively involved in evaluating the research methodology and the (conceptual and computer) models that are used in research, as well as in evaluating the final results (de la Vega-Leinert et al., in preparation). Early and regular involvement creates a sense of ownership of the research process so that the results are more likely to be used by stakeholders. Beyond ‘model calibration’, this reality check may be an inspiring as well ARTICLE IN PRESS 172 M. Welp et al. / Global Environmental Change 16 (2006) 170–181 Table 1 Science-based dialogues in comparison with policy, governance and corporate dialogues Type of dialogue Initiator/principal coordinator Objectives Who are the stakeholders? Science-based stakeholder dialogues Researchers, scientific institutions or networks thereof Deepening our understanding, combining knowledge bases, checking social relevance Corporations, SMEs, NGOs, policy-makers, citizens Policy dialogues Policy-makers, bureaucrats Creating support for policies, ultimately passing laws and regulations, creating a monitoring structure Various organized interest groups, corporations, SMEs, NGOs, researchers, citizens Multi-stakeholder dialogues for governance Intergovernmental organisations, International non-profit organizations Creating multi-stakeholder partnerships, influencing policy and business practices Governments, corporations, international and national NGOs, researchers Corporate dialogues Corporations Taking society’s expectations into account in business practices and in the transformation of business strategies Government, NGOs, customers, employees, suppliers, communities, researchers as difficult experience for model developers. Indeed, all models are based on ‘worldviews’, which during stakeholder dialogues can be seriously questioned. This provides an opportunity to improve the models. One of the early examples of a science-policy dialogues in sustainability science was the so-called Delft Process. Its aim was to investigate the usefulness of the IMAGE global model in climate negotiations and in their preparations (in particular, COP3 and the Ad Hoc Meeting on the Berlin Mandate) (van Daalen et al., 1996). The scope, type and scale of the IMAGE analysis as well as its results were iteratively considered and adjusted to better suit policy needs. To this end ‘the safe (carbon emission) landing’ method (a backcasting scenario approach) was successfully developed and later used in the UK as well as at EU level. Third, social science research on global change faces limits to scientific reasoning and requires the incorporation of ethical considerations. An example of a question that requires the incorporation of value judgements is: ‘What is dangerous climate change according to article 2 of the United Nations Framework Convention on Climate Change (UNFCCC)?’ An assessment exercise with scientists and stakeholders from selected world regions, organised by the European Climate Forum, showed that much research is needed in this field and that it is necessary to involve a broader range of stakeholders (not only experts) in such dialogues (European Climate Forum, 2004). The terms stakeholder dialogues and public participation in science are sometimes used ambiguously. A sciencebased stakeholder dialogue is here defined as a structured communicative process of linking scientists with selected actors that are relevant for the research problem at hand. These actors possess specialized knowledge and have insights which are of relevance to the scientific process. Stakeholder dialogue processes do not always aim at being representative of the full spectrum of interests at hand. The focus is on securing certain competencies. Public participation on the other hand refers to the participation of ordinary citizens in debates on controversial issues such as genetically modified organisms or climate change. Here the focus is on allowing all interests to voice their concerns (cf. Webler, 1995). Dunkerley and Glasner (1998) have discussed how citizen juries can be used to support policy making in relation to new genetic technologies. An example, where not only experts were engaged in the assessment of values, expectations and risk perception related to climate change was the EU research project ULYSSES.3 Several hundred citizens were engaged in Integrated Assessment Focus Group sessions to learn and debate about the climate change problem (Kasemir et al., 2003b; Welp, in preparation; Stoll-Kleemann et al., 2001). Participants were confronted with the latest knowledge on climate change and synthesised their newly gained understanding in citizen assessments of the causes and impacts of climate change, as well as possible solutions. These included suggestions on mitigation and adaptation measures (e.g. within the transport, energy and household sectors) as well on who should act, where and when. The fourth reason for the need of stakeholder dialogues is that scientists need to have access to data and knowledge that otherwise would remain unknown or at least very difficult to access. For example, during the ATEAM project, scientists gained much insight on local and sectoral drivers of change, implementation and feasibility of management and adaptive measures, which were not extensively considered in state-of-the-art ecosystem modelling (de la Vega-Leinert et al., in preparation). Relevant 3 ULYSSES (Urban lifestyles, sustainability and integrated environmental assessment) explored new channels of communication between scientists and lay persons. Integrated Assessment (IA) computer models were used in small groups to communicate scientific results and to explore how citizens perceive climate change. ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 knowledge that scientists may obtain from stakeholders ranges from quantitative (e.g. datasets) to qualitative knowledge on sectoral specificities and stakeholders’ understanding of global change. Science-based stakeholder dialogues are needed as an interface to combine different knowledge domains, and as such may play a key role in Integrated Assessments (e.g. Lemos and Morehouse, 2005). For further description of the expert-lay knowledge divide and emphasis on the role of local knowledge in decision-making, see Wynne (1996). 3. Comparison with other types of dialogues A stakeholder is usually defined as a person or a group who has a stake or special interest in an issue, policy, company, etc. A distinction can be made between individuals and groups who affect (determine) a decision or action and those who are affected by it (Freeman, 1984; Harrison and Qureshi, 2000). The concept originates from management literature where a distinction is made between shareholders, i.e. those who own the company, and stakeholders, i.e. individuals or groups, which are impacted by business activities or can influence the business environment. Besides science-based stakeholder dialogues, we can identify three main types of stakeholder-oriented dialogues: ‘policy dialogues’, ‘multi-stakeholder dialogues for governance’, and ‘corporate dialogues’. Table 1 summarizes the key features of science-based stakeholder dialogues, policy dialogues, multi-stakeholder dialogues, and corporate dialogues. All four types of dialogues share the basic concept of learning and exchange of knowledge and opinions. The intention is to create a safe space for the exchange of arguments, which is based on mutual trust. In such a setting, participants can learn from each other and as a group. What are the specifics of science-based stakeholder dialogues? Science-based stakeholder dialogues usually have different aims than those conducted by policymakers, international organisations or corporations. The main objective of ‘policy dialogues’ is to create support for policies and new pieces of legislation. Although collaborative policy dialogues are far from being the dominant way of policy-making (Innes and Booher, 2003), this approach is applied in many different sectors, including water policies, conservation policies and many others. ‘Multi-stakeholder dialogues for governance’ are international efforts to create partnerships and voluntary commitments among a broad range of international actors (cf. Hemmati, 2002). For example, the Stakeholder Forum for Sustainable Development (www.unedforum.org), which recently became a free-standing organisation in its own right but still closely linked with UN organisations, supports the increased involvement of stakeholders in international and national governance processes. Another example is the Forest Stewardship Council, an interna- 173 tional network promoting sustainable management of the world’s forests (www.fsc.org). The members, including the forest industry and environmental NGOs have developed an international label for sustainable forest products (Vallejo and Hauselman, 2004). The objectives of ‘corporate dialogues’ are to demonstrate openness and the will for a critical exchange of views. A key objective is to learn about the expectations of different stakeholder groups with regard to the company’s business ethics and practices. The gained insights can be important on different levels of corporate decisionmaking. For example, Shell International organised dialogues in several countries after the dispute over ‘Brent Spar’ (Jesper, 1998). The increasing importance of defining and communicating corporate responsibility has in recent years boosted an interest in stakeholder dialogues in the business world (e.g. SustainAbility, 1996a, b). By now, stakeholder dialogues are a key element in the effort of many corporations to pursue Corporate Social Responsibility (CSR) (see for example van den Hove et al., 2002). In such dialogues private companies reflect on society’s views and expectations through consultation with various groups such as consumer associations, suppliers, environmental NGOs, religious organizations, etc. As opposite to the other types of dialogues, sciencebased dialogues may not primarily aim at reaching a consensus. Effectively, dissent can be as valuable as consensus, since it reveals areas in which more research is needed. For example, the ECF has served as a platform for controversial discussions rather than for negotiating consensual positions (Hasselmann et al., 2003). The framing of a particular problem itself may be difficult, but this implies that there is need for more discourse and exchange of arguments. Controversial issues, such as Carbon Capturing and Sequestration (CCS) have within the ECF shown the need for reflecting on the framing issue (see Section 4.3). However, dealing with dissent may not always be easy for scientists who may question the competence of stakeholders and emphasise their own ‘scientifically’ based judgement. Conversely stakeholders may end up confirmed in their views that scientists live in a different reality—often in the so-called ‘ivory tower’ (Welp, 2001). Besides explicitly encouraging mutual learning, science-based dialogues may have other implicit goals, which should not be ignored. The latter can include seeking legitimacy and financial support for the research as well as acceptance for the produced results. These are valid objectives for scientists because stakeholder scrutiny of the research strategy is likely to increase the societal relevance of research; however, they should not be the main objective of the dialogue exercise. A distinction between five different types of stakeholder participation in science is presented in Table 2. The roles stakeholders may have range from commenting concepts to substantial involvement in the scientific inquiry. The selected methods vary accordingly. ARTICLE IN PRESS 174 M. Welp et al. / Global Environmental Change 16 (2006) 170–181 Table 2 Types of stakeholder participation in science Role of the stakeholder Methods When? 1. Commenting concepts, drafts and results Written reviews Interviews Group discussions Throughout the research process 2. Eliciting quantitative data, stylised facts and expert judgements Data mining Expert elicitations Beginning of a research process 3. Exploring values, preferences, expectations and risk perception Interviews Focus groups Beginning and end (if focus is on changed values and preferences) 4. Identifying research questions Workshops Surveys Beginning/end of a research process (open/new questions) 5. Substantial involvement in the process of modelling and generating scientific insights Research visits Team work Throughout the research process on a regular basis 4.2. Rational actor paradigm It has been widely applied primarily due to the fact that theorems and hypotheses on actors’ behaviour can be formulated in mathematical terms. According to RAP, individual persons or ‘agents’ try to maximise given utility functions under given constraints. RAP does not give answers to the question: where do the utility functions come from? Actors’ preferences are thus seen as given and assumed not to change over time. A key feature of RAP is social atomism. Individual persons try to maximise their utility functions without communicating with others. For example, consumer behaviour, one manifestation of preferences, is not influenced by family, friends, colleagues, media or the outside world. The assumptions of atomistic rationality and constant preferences appear in contrast with any research orientation that deals with dialogues and social learning4 (see Section 4.4 on organisational learning). The philosopher and economist Amartya Sen has challenged the standard neoclassical view that there can be no rational discussion (argument) about ends in economics. Since intersubjective comparison of utilities was declared meaningless, there was no room in neoclassical economy for such considerations. This relates closely to the distinction between facts and values as discussed earlier and has had great consequences for welfare economics (Putnam, 2002). RAP has not lost its importance in social sciences, despite abundant and diverse critique. This includes bounded rationality (Simon, 1955) and the literature on new institutionalism, which usually provides qualitative approaches explaining political, historical, economic and social institutions, with no explicit mathematical representation. Jaeger et al. (1998) furthermore argue that RAP seems to be successful in combining models of human actors with models of their interactions. Jaeger (1998) further suggests how to link RAP and social The RAP has been extremely influential in neoclassical economy and social sciences in general (Jaeger et al., 2001). 4 This atomistic rationality is also in strong contrast to marketing, behavioural and educational science orientations. 4. Relevance of selected theoretical frameworks 4.1. Why do we need theory? Science-based stakeholder dialogues have been driven by the practical need to link scientific inquiry with different knowledge bases and to take into account value and risk judgements of individuals and groups. The theoretical framework for such exercises has however remained rudimentary. Although theoretical overloading should be avoided, a solid theoretical foundation can be helpful for the art and practice of dialogues. A key requirement for a practical theory relevant for stakeholder dialogues is that it integrates the different domains and layers of a dialogue (Jaeger, 2003). Firstly, a dialogue is about exchanging arguments and creating common meaning. Secondly, dialogues also have a layer of personal relationships where trust building, empathy, antipathy, etc. play a big role. In science we also face the challenge of building a bridge between an individual’s mental model and conceptual/computer models, which may be used to create and test arguments. A major step in theory development is needed to enable the formal representation of stakeholders’ mental models, preferences and expectations. To achieve this we believe that links between formal modelling and dialogues should be developed substantially. In the following, we will examine the relevance of three theoretical frameworks for stakeholder dialogues. The selected theoretical foundations are Rational Actor Paradigm (RAP), Bayesian Learning, and Organisational Learning. This selection of theoretical approaches is by no means comprehensive, but appears the most promising and relevant to the authors. ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 rationality in a pragmatic way: such a procedure would involve lay-persons and invited experts in a series of focus groups or similar arrangements. Such participatory processes closely linked with representative democratic institutions are believed to preclude that, for example, risk management becomes a solely expert exercise.5 In global change research RAP is inherent in the idea of the ‘benevolent planner’, a single actor who aims at maximising global welfare. In reality, a multitude of actors make decisions based on their expected utility, which makes it impossible to find an ‘optimal’ solution in the technical sense of RAP (Jaeger et al., 1998). A basic question remains, namely, whether a global welfare function can be defined at all and whether this would be a sum of the welfare functions of each actor. Real-world actors do not have perfect knowledge of the costs and benefits of different alternatives. Thus actors are bound to make decisions under uncertainty. Including the notion of probability into RAP led to the concept of lottery as a metaphor for decision theory (i.e. a lottery is a set of possibilities which represent feasible courses of action). The second step was the combination of probability and utility in the expected utility model. A body of literature has been developed to explain how individuals make decisions under uncertainty. Most of this literature focuses on exposing von Neumann’s and Morgenstern’s (1947) expected utility axioms. In fact, the model was developed to describe how actors should behave if they were about to act rationally. According to the Expected Utility Theory (EUT), by assessing the probability of different outcomes actors try to maximise the expected utility, taking into account that some are risk averse while others are risk seeking. 175 data/piece of information.’6 (source: http://kiew.cs.unidortmund.de:8001/mlnet/instances/81d91eaae441d87555) Box 1 Bayes’ theorem is a result in probability theory. Bayes’ theorem gives the probability of a random event A occurring given that we know a related event B occurred. This probability is noted P(A|B), and is read ‘probability of A given B’. This measure is sometimes called the ‘posterior’, since it is computed after it is known whether B is the case or not. Bayesian belief network: a graphical tool to help make decisions under uncertainty. It can be used to build a Decision Support System (e.g. a Bayesian Expert System). Bayesian networks are composed of three elements: a set of nodes representing system variables, a set of links representing causal relationships between the nodes, and a set of probabilities, for each node, specifying the belief that a node will be in a particular state given the states of those nodes that can affect it directly. Bayesian learning: the process by which a Baysian belief network updates its set of probabilities (so-called conditional probability tables) as a result of receiving case data about variables in the table. Adapted from: Cain (2001), Wikipedia Bayesian learning seems partly to be a departure from RAP in its original version (i.e. in RAP there is no place for learning since, as mentioned above, actors have complete information and preferences do not change). Models based on Bayesian learning may however better represent true human behaviour, primarily since agents have limited information storage capacity. Similar to Game theory, Bayesian learning acknowledges uncertainty and operates with probabilities. One definition of Bayesian learning reads as follows: ‘Bayesian learning constitutes a probabilistic view of learning, based on Bayes’ Theorem. The underlying assumption is, that there is a set of hypotheses, each having a certain probability of being correct. Receiving more information changes the probabilities from a learner’s point of view. For instance an observation might contradict a hypothesis, or strengthen the belief in it. The aim in this setting is to be able to find a hypothesis with highest probability of being correct, given a specific set of Bayesian learning is represented in mathematical terms in the following way. In a simple example, suppose there are two states of the world s and s0 . Agents are unsure which of them is the actual or true state of the world but at time t, the ith agent attaches probability zi(t) to s0 being the true state of the world and thus believes s to be true state with the probability 1-zi(t). Beliefs are thus captured in the single parameter zi(t). In the light of their beliefs, the agents choose a particular course of action. Having acted they observe a result, which is called X. Based on this they update the probabilities s0 being the true state of the world (Breen, 1999). The Bayesian mechanism provides a plausible way in which beliefs can change over time, a process called belief updating. Developing RAP further and especially applying the concept of Bayesian learning seem to be promising paths for advancing the analysis of stakeholders’ assessment in global change research. Three main areas of relevance can be found: (a) framing problems, (b) finding differences and inconsistencies, (c) addressing the question of how actors learn. RAP, Bayesian networks and stakeholder dialogue can be linked with each other by developing and using appropriate analytical and communication tools. 5 See Ulrich Beck’s comments on risk industry and risk professionals. Both deal with the rational management of dangerous uncertainty. 6 Rational choice theory in its current version has little to say about the question of how to explain collective beliefs (Boudon, 1996, p. 147). 4.3. Bayesian learning ARTICLE IN PRESS 176 M. Welp et al. / Global Environmental Change 16 (2006) 170–181 4.3.1. Framing problems Environmental policy making is often faced with factual uncertainty and political controversy. In conflict literature this is described as issues being at dispute and values being subject to conflict (Hellström, 2001). Because environmental problems tend to be complex and subject to both factual uncertainty and conflicts over values, they are not easy to frame in a meaningful way (Garrison and Greer-Wootten, 2000). The inability to construct well-formed problems hampers efforts to find mutually acceptable solutions. Empirical studies have shown that the framing of an issue by using a positive or negative description (e.g. would you invest in a medicine, which saves 70% of the patients? vs. would you invest in a medicine when still 30% would die?) has a strong influence on the answers people give (Gardner and Stern, 1996). Other studies have attempted to show how citizen perceive certain complex issues (are there wrong, imprecise or irrelevant beliefs?), and how risk communication can take these insights into account when aiding the public’s understanding about complex issues (Bostrom et al., 1992). Wynne (2005) on the other hand turned the problem upside down and argued that public misunderstanding, mistrust or skepticism to scientific discourse on risk may in fact relate to the way risk issues are defined and the risk discourse constructed, which excludes citizens’ views and perceptions. The author further believes that participation processes and framing methods, developed to deal with the resistance of the public or to educate citizens solely focus on downstream risk issues (e.g. risk and impacts of a new technology), and deny citizens the ability and the possibility to address essential social debates (upstream issues—which human purposes drive science and innovation?). In this context, one application of Bayesian learning is the use of Bayesian belief networks to visualise the structure of our present knowledge and thus come up with an accepted problem definition. The Bayesian formalism allows for subjective probabilities, which is of interest in stakeholder dialogue processes. Imprecise information on complex systems can be presented by proceeding from a simple influence diagram to a causal network containing system components (nodes) and causal dependencies (links or arcs). The probabilistic concept underlying a Bayesian approach acknowledges the uncertainty of data and of the conceptualisation of problems and is more likely to be accepted by stakeholders than single predicted results. 4.3.2. Finding differences and inconsistencies Finding an agreement about an issue may be easier if subjective probabilities and assessments are made explicit. Here Bayesian learning can also be useful, since it helps to find inconsistencies in people’s thinking. Key experts and decision-makers may have widely different and inconsistent explanations of the problems at hand, or opinions on the course to adopt. Bayesian expert systems can, for example, be applied to help structuring the debate on various global change issues such as the economic, environmental and social impacts of carbon capturing and storage in oceans. Stakeholders within the ECF hold strong views on the desirability of this option to tackle the greenhouse gas problem. Even the framing of the research question caused serious conflicts between involved companies and environmental NGOs. For some, carbon capturing and sequestration is a potential low-cost solution, for others an unacceptable technical fix associated with risks that cannot currently be quantified properly. Thus a structuring process can greatly benefit from the use of Bayesian belief networks. Cain’s (2001) illuminating guidelines provide concrete steps to capture and represent the world as described by different stakeholders in simple conceptual models. Stakeholder interviews or group discussions are conducted to elicit expert information and different subjective probabilities. Stakeholder Bayesian networks (BNs) are created. A BN is basically a set of nodes representing system variables and a set of links representing causal relationships between these nodes (see Fig. 1). Stakeholder Bayesian networks can at a later stage be simplified and merged to master BNs. In a next step Conditional probability tables (CPT) are created: a set of probabilities, one for each node, specifying the belief that a node will be in a particular state given the states of those nodes that affect it directly (its parents). In other words, CPTs express how relationships between nodes operate (see Table 3). Each row in a CPT implies a question. Using the belief network in Fig. 1 as an example, we can ask the following question: ‘If the status of seafloor habitats (bottom feeders) is poor and the intensity of industrial fishing is high, what is the chance that sustainability of fish stocks is acceptable?’ If it appears to be difficult to frame these questions then it is likely that the master BN is illogical. The structure or the states of the nodes have to be subsequently altered. As mentioned above, Stakeholder Bayesian networks can be simplified and merged to master BNs. When the master BN is completed it can be turned into a fully functioning BN, which can be used to help decisions and to carry out further dialogues with stakeholders. This is done by filling in the CPTs using the best and most appropriate data or expert judgement available and by playing around with the created BN (by changing probabilities). By building an expert belief system and reviewing it together with stakeholders, a better picture of the problems at hand can be obtained. The whole exercise provides the involved scientists and stakeholders an opportunity to reflect on their basic assumptions, revise their views and learn as individuals and as a team. Such a procedure will reveal gaps in our present knowledge and thus point at new research questions. Expert belief systems can be used to develop empirical explanations (a causing b with a certain probability) but also normative argumentation. Thus both factual uncertainty and conflict about values can be addressed. This helps to identify areas where agreement can be found and where disagreement over issues or values prevails. The ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 Public support Yes 50% No 50% Carbon sequestration in oceans Yes 50% No 50% 177 Marine ecosystem resilience High 50% Low 50% Status of seafloor habitats (bottom feeders) Good 50% Poor 50% Intensity of industrial fishing High 50% Low 50% Sustainability of fish stocks Good 33.3% Acceptable 33.3% Poor 33.3% Income from fisheries High 50% Low 50% Fig. 1. A simple Bayesian belief network. possible fields of application encompass a broad range of decision-making situations ranging from natural resources to business management decisions. 4.3.3. How do stakeholders learn?—constructing a model of learning As mentioned above an important aspect of Bayesian learning is that the update of beliefs when new evidence occurs is possible. This takes place formally by experts changing the probabilities of a statement being true (cf. Fig. 1). An application of Bayesian learning could be to study ‘how and on what basis stakeholders update their beliefs when confronted with new, albeit uncertain insights?’ It becomes possible to develop formal models of how stakeholders or ‘agents’ learn. Such models, even though they may remain anecdotal, explicitly aim at simulating more realistic present and future behaviour such as consumer behaviour, investment decisions or positions in negotiations.7 Research in this area, although crucial to improve current global change research and modeling, is in its infancy. Agent-based modelling is one approach which is actively developed and experimented within climate and global change research (Moss et al., 2001). The Bayesian approach presented in the present paper seems promising for exploring stakeholders’ mental models of the world and in turning qualitative descriptions into simple quantitative models. An encouraging feature of Bayesian networks is that several time steps can be built 7 The strongest motivation for this kind of research seems to lie in the potential to make predictions in the finance and insurance sector. Just (2001) argues that his approach can be useful for predictions in the agricultural sector, such as crop insurance and production contract negotiations. Table 3 Conditional probability table (CPT) of an imaginary stakeholder Sustainability of fish stocks Intensity of industrial fishing Low Low High High Status of seafloor habitats Good Poor Good Poor Good Acceptable Poor 0.60 0.00 0.40 0.00 0.40 0.10 0.60 0.00 0.00 0.90 0.00 1.00 into the system. Thus interventions in a management system can be explored in an iterative way. 4.4. Organisational learning A formal approach to stakeholder assessments, such as outlined above, is an essential, though insufficient part, of a conceptual framework of stakeholder dialogues. A stakeholder dialogue is a social learning process, where communication and interaction in small groups play a fundamental role. Management science and organisational learning have greatly influenced business practices in the last decade (Senge, 1990), but have in practice been largely neglected in global change research. Bohm’s distinction between discussion and dialogue is helpful in understanding the special character of dialogues. In discussions, individual views are presented and defended. Discussions can be seen as a ping–pong game: the subject of common interest is analysed from many points of view; the purpose of the game is normally to win (Bohm, 1996). Winning a discussion means that one’s view is ARTICLE IN PRESS 178 M. Welp et al. / Global Environmental Change 16 (2006) 170–181 accepted by the group and prevails. In a dialogue, in contrast, participants are neither negotiating positions nor trying to reach a consensus or to win. Dialogues are based on mutual respect and on the notion that everyone has a valid viewpoint. In a dialogue a free flow of meaning between participants takes place and individuals gain insights, which could not be achieved individually. For Bohm the greatest impact is realised through a synergy between the processes of dialogue and discussion. In practice it is often difficult to move from the mode of structured discussion to one of dialogue, and if and when this happens is to some extent unpredictable. Of paramount importance is that a certain degree of trust is necessary to enter dialogues in Bohm’s sense. The first necessary condition for a dialogue is that participants treat one another as colleagues. Second, they should ‘suspend’ their assumption, which means to hold them ‘as if it were ‘hanging in front of you’, constantly accessible to questioning and observation’ (Bohm, 1996). It does not mean throwing the assumptions away or suppressing them. By holding their assumptions up for examination, scientists involved in such dialogues can learn about the mental models of stakeholders and their own. Third, the process is structured by a skilled facilitator. All three conditions are highly relevant in the potential success of science-based stakeholder dialogues, as well as a number of other factors.8 Regular interaction with stakeholders over a longer period of time is a necessary requirement to create a common language and build mutual trust. Regular participation to a dialogue enables a richer understanding of the uniqueness of each person’s point of view. However, there exist no general recipes for organizing a dialogue, which is beneficial for all participants. Experimenting with different formats and ways of communication as well as learning by doing play a fundamental role in bringing together the traditionally separated domains of academic and ‘outside’ worlds. The development of the concept of the learning organisation has been closely linked to various key people in the business world. Wack, de Geus, Collyns and Carroll all worked for Shell but also contributed significantly to the development of the theory and practice of learning organisations. Many other business persons have been mentioned in the ‘history’ of the learning organisation (cf. www.fieldbook.com). Scenario practice as a learning activity has been used and further developed especially within Royal Dutch Shell. The same methodology has been later expanded and implemented in a radically different environment, that of national and international peace negotiations under the name ‘civic scenarios’ (Kahane, 2002, 2004). The close link between scenario practice and learning organisations further has been pointed out for example by Wack (1985). 8 These are equally relevant for other dialogues, e.g. in the field of management or policy making. On the scientific side, the work of the theoretical physicist Bohm has been acknowledged widely. Isaacs, an associate of Bohm, introduced the concept of discussion and dialogue to Senge in the late 1980s. This influenced strongly Senge’s work on the five disciplines (Personal Mastery, Mental Models, Shared Vision, Team Learning, and Systems Thinking), which have been applied widely in organisational learning (Senge, 1990). The ‘Fifth Discipline’—Systems Thinking—implies the necessity of seeing inter-relationships rather than linear cause-effect chains, and seeing processes of change rather than snapshots. Systems thinking and analysis being familiar to global change scientists can thus form a common ground where stakeholders and scientists can meet. 5. Discussion Global change and climate change are becoming issues of great public interest. In particular at the interface between science, policy and society new ways of inquiry and dialogue have to be developed. Traditionally, science has had great authority in defining what is a socially relevant problem and what approaches are appropriate in investigating them. Also in global change research and climate research political institutions for support and legitimacy play a significant role (Miller, 2004). Ravetz (2005) argues that science needs to open up for new ways of framing problems. In such a process stakeholder dialogues play a vital role. In the following, we will discuss the three theoretical frameworks outlined in the previous section and relate these to some areas of application and tools for stakeholder dialogues. We make a distinction between two kinds of tools: tools for facilitating communication (communication tools) and tools for formalizing actors’ mental models and assessments (analytical tools). Theories of organisational learning have great relevance for stakeholder dialogues. In dialogue exercises scientists and stakeholders meet either once or (preferably) several times. Typical settings for dialogues are conferences and workshops. Outside plenary stakeholders and scientists may team up in smaller break-out groups to debate a specific question. Key issues to consider are thus under what conditions such groups become learning teams, rather than debating opponents or negotiating parties, and how the process of learning can be extended beyond such meetings. Applying and further developing Senge’s disciplines, in particular systems thinking and team learning can contribute to the emergence of long-term dialogues and socially relevant research results. A stakeholder theory developed along the lines of learning organisations thus docks on the experiences and language of the business world and management. There are various communication tools that can enhance team learning (cf. Preskill and Torres, 1999). In a focus group session a small number of people meet to discuss, while a World Café session provides opportunities for a larger group to share and collect the ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 179 Table 4 Relevance of selected theories for the practice of stakeholder dialogues Approaches Areas of application Tools Organisational Learning Necessary conditions for dialogues in contrast to discussions Working and learning as small teams consisting of stakeholders and researchers Identify and agree on problem formulation Communication tools Focus groups Conceptualizing agents’ behaviour Framing problems in a stakeholder dialogue setting Finding differences and inconsistencies in stakeholders’ assessments How do stakeholders learn (update their beliefs)? Analytical tools Computer models Bayesian belief networks Formal approaches: Rational Actor Paradigm, Bayesian Learning variety of views. Professional moderation is important to keep discussions on track while climate games (board and computer games) can be used as a ‘kick-off’ of a stakeholder dialogue. For an overview of different participatory methods, see van Asselt and Rijkens-Klomp (2002) in this journal (Table 4). While organisational learning has relevance for stakeholder dialogues on a procedural level, formal approaches such as Rational Actor Paradigm and the Bayesian Approach are interesting on an analytical level. They provide tools to structure dialogues and analyse different perspectives that stakeholder may have. Rational Actor Paradigm is a starting point for discussing the preferences and expectations of different actors. Although it has been subject to substantial critique, among others Bounded rationality, it still plays a significant role in social science research. From these discussions we learn that stakeholders are not only utility maximising agents, but rather actors that have limited information-processing capacity, have to make decisions under uncertainty and have different degrees of risk aversion. Bayesian networks are a promising technique for formalizing stakeholders’ assessments. The Bayesian approach enables the mathematical representation of stakeholder assessments, which is a similarity with the Rational Actor Paradigm. Conceptually, Bayesian learning can also be related to Bounded rationality, since both acknowledge the fact that people have limited information storage capacity and tend to make decisions incrementally. Bayesian learning and in particular Bayesian belief networks is a promising technique for framing problems, finding differences and inconsistencies in stakeholders’ thinking and studying how stakeholders update their beliefs. There are computer-based software tools which are helpful in creating Bayesian belief networks. The Bayesian approach can also be linked with other types of analytical tools such as modelling. Since the Bayesian approach is not yet widely used in management and policy making, its use may face institutional obstacles which have to be considered. Rayner et al. World Café Professional mediation Climate games (2005) have analysed what obstacles lie in the use of shortterm climate forecasting for improved water management in the case study areas in the US. While managers identify ‘poor reliability’ of the forecasts as the main reason, Rayner et al. identify several institutional reasons such as organisational conservatism and complexity, political disincentives of to innovation and regulatory constraints. Thus if the Bayesian approach is to help the use of scientific knowledge in decision-making of insurance companies, farmers, and natural resources managers, it needs to be embedded in organisational routines and become part of the mental models of managers problem solving. Similarly as the Bayesian approach is not common knowledge among resource managers, computer modelling is not something everybody is acquainted with. We should keep in mind that the practice of stakeholder dialogues needs not only theory and related analytical and communication tools, but many skills that can be learned and trained, such as active listening, moderating, observing group dynamics, etc. (Stoll-Kleemann et al., 2001). Indeed Kasemir et al. (2003a) see a danger in putting a too heavy theoretical overload on stakeholder dialogues, since this puts the participants in unequal positions. The process should be foremost reflexive, implying that not only the content of the dialogue, but also the rules of the dialogue themselves are openly discussed (Webler, 1995). Also, the way analytical and communication tools are used should be discussed together with stakeholders. For example Patt et al. (2005) recommend that vulnerability assessment to climate change should be embedded in a social learning process. 6. Conclusions In science dialogues relevant stakeholders usually include, for example business and sectoral representatives, policy makers at different scales, as well as citizens. Each of these groups possesses distinctive knowledge, which can range from records on past flood and storm damages owned by insurance companies and sector management ARTICLE IN PRESS 180 M. Welp et al. / Global Environmental Change 16 (2006) 170–181 expertise to the everyday life experience of lay-persons. Their role in the dialogue ranges from witnessing the scientific process and commenting research results to being intensively involved in the process of generating new knowledge and meaning. Each theoretical approach discussed in the two previous sections has practical relevance for science-based dialogues. The controversial discussion about the Rational Actor Paradigm plays a key role in the way we see actors in global change research. They have different degrees of risk aversion, have to make decisions under uncertainty, and thus do not have complete knowledge to base their decisions on. The Bayesian approach is relevant for framing problems, visualizing stakeholders mental models and observing how stakeholders learn. The theory of organisational learning as outlined by Senge and others is vital as it points to necessary conditions for dialogues. It gives practical guidance to how people with different educational and professional backgrounds can team up to small groups, which provide opportunities for learning and joint problem solving. A research process usually includes several iterations and a dialogue may take place over a long period of time. Cycles of stakeholder dialogues may start with identifying relevant research questions and move on to phases of consultation, reviewing drafts and modifying concepts and models. Thus different tools and approaches are needed throughout the research process. A promising path is to integrate both the use of communication and analytical tools in dialogue practice. Reflecting on different mental models and arguments is key to the creation of common meaning and development of formal models. Existing formal models, even rudimentary and incomplete, can be a starting point for new discussions and dialogues on stakeholders’ perceptions and beliefs, which in turn can serve as iterations in model improvements, testing and use. Research on global environmental change increasingly takes place in interdisciplinary teams, in which scientists in different fields have regular discussions and dialogues, and think together in an interdisciplinary way. Science-based stakeholder dialogues are an extension of this practice and an effort to link research with knowledge domains outside the academic world. Stakeholder dialogues are not a substitute for scientific thinking but rather they foster the art and practice of thinking together. Acknowledgements We are grateful for the productive dialogues that have been conducted within the European Climate Forum (ECF). We would like to acknowledge the Potsdam Institute for Climate Impact Research for enabling this research on stakeholder dialogues. Anne de la Vega-Leinert benefited from financial and otherwise support within the ATEAM Project (Advanced Terrestrial Ecosystems Analysis and Modelling) funded within the 5th Framework Programme of the European Commission ‘Energy, Envir- onment and Sustainable Development’ (Project Number: EVK2-2000-00075). References Bärlund, I., Carter, T.R., 2002. Integrated global change scenarios: surveying user needs in Finland. Global Environmental Change 12, 219–229. Bohm, D., 1996. On Dialogue. Routledge, London. Bostrom, A., Fischhoff, B., Morgan, M.G., 1992. Characterizing mental models of hazardous process: a methodology and an application to radon. Journal of Social Issues 48, 85–100. Boudon, R., 1996. The ‘Cognitivist Model’: a generalised ‘Rational-choice Model’. Rationality and Society 8, 123–150. Breen, R., 1999. Beliefs, rational choice and Bayesian learning. Rationality and Society 11, 463–479. Cain, J., 2001. Planning Improvements in Natural Resources Management. Guidelines for Using Bayesian Networks to Support the Planning and Management of Development Programmes in the Water Sectors and Beyond. Centre for Ecology and Hydrology, Wallingford. Christensen, J.H., Christensen, O.B., 2003. Climate modelling: severe summertime flooding in Europe. Nature 421, 805–806. de la Vega-Leinert, A.C., Schröter, D., Leemans, R., Fritsch, U., Pluimers, J., in preparation. A stakeholder dialogue on European vulnerability. Regional Environmental Change (ATEAM special issues). Dunkerley, D., Glasner, P., 1998. Empowering the public? Citizens juries and the new genetic technologies. Critical Public Health 8, 181–192. Economist, 2003. Hot potato revisited, November 6th 2003. European Climate Forum, 2004. What is dangerous climate change? Initial results of a Symposium on Key Vulnerable Regions Climate Change and Article 2 of the UNFCCC Buenos Aires, 14 December 2004. Internet: http://www.european-climate-forum.net/pdf/ECF_beijing_results.pdf Freeman, R.E., 1984. Strategic Management: A Stakeholder Approach. Pitman, Boston. Gardner, G.T., Stern, P.C., 1996. Environmental Problems and Human Behavior. Allyn and Bacon, Boston. Garrison, R.M., Greer-Wootten, B., 2000. Negotiating congruencies among multiple interpretive frameworks: elite representations of global climate change. In: Climate Change Communication. University of Waterloo. Harrison, S., Qureshi, M.E., 2000. Choice of stakeholder groups and members in multicriteria decision models. Natural Resources Forum 24, 11–19. Hasselmann, K., Latif, M., Hooss, G., Azar, C., Edenhofer, O., Jaeger, C.C., Johannessen, O.M., Kemfert, C., Welp, M., Wokaun, A., 2003. The challenge of long-term climate change. Science 302, 1923–1925. Hellström, E., 2001. Conflict cultures—qualitative comparative analysis of environmental conflicts in forestry. Silva Fennica Monographs 2. Hemmati, M., 2002. Multi-Stakeholder Processes for Governance and Sustainability—Beyond Deadlock and Conflict. Earthscan, London. Innes, J.E., Booher, D.E., 2003. Collaborative policymaking: governance through dialogues. In: Hajer, M.A., Wagenaar, H. (Eds.), Deliberative Policy Analysis: Understanding Governance in the Network Society. Cambridge University Press, Cambridge. Jaeger, C.C., 1998. Risk management and integrated assessment. Environmental Modeling and Assessment 3, 211–225. Jaeger, C.C., 2003. A note on domains of discourse. Logical know-how or integrated environmental modelling. PIK Report No. 86. http:// www.pik-potsdam.de/publications/pik_reports. Jaeger, C.C., Renn, O., Rosa, E.A., Webler, T., 1998. Decision analysis and rational action. In: Rayner, S., Malone, E.L. (Eds.), Human Choice and Climate Change, vol. 3. Battelle Press, Columbus/OH. Jaeger, C.C., Renn, O., Rosa, E.A., Webler, T., 2001. Risk, Uncertainty, and Rational Action. Earthscan, London. Jesper, G., 1998. Corporate legitimacy in risk society: the case of the Brent Spar. Business Strategy and the Environment 7, 213–222. ARTICLE IN PRESS M. Welp et al. / Global Environmental Change 16 (2006) 170–181 Just, D.R., 2001. Electronic Source: http://are.berkeley.edu/courses/ envres_seminar/just.pdf Kahane, A., 2002. Civic scenarios as a tool for societal change. Strategy and Leadership 30 (1), 32–37. Kahane, A., 2004. Solving tough problems. An Open Way of Talking, Listening, and Creating New Realities. Berret-Koehler Publishers, Inc., San Francisco. Kasemir, B., Dahinden, U., Swartling, A.G., Schibli, D., Schüle, R., Tabara, D., Jaeger, C.C., 2003a. Collage processes and citizen’s visions for future. In: Kasemir, B., Jäger, J., Jaeger, C.C., Gardner, M.T. (Eds.), Public Participation in Sustainability Science. Cambridge University Press, Cambridge, pp. 81–104. Kasemir, B., Jäger, J., Jaeger, C.C., Gardner, M.T. (Eds.), 2003b. Public Participation in Sustainability Science. Cambridge University Press, Cambridge. Kates, R.W., Clark, W.C., Corell, R., Hall, J.M., Jaeger, C.C., Lowe, I., McCarthy, J.J., Schellnhuber, H.J., Bolin, B., Dickson, N.M., Faucheux, S., Gallopin, G.C., Gruebler, A., Huntley, B., Jager, J., Jodha, N.S., Kasperson, R.E., Mabogunje, A., Matson, P., Mooney, H., Moore III, B., O’Riordan, T., Svevin, U., 2001. Sustainability science. Science 292, 641–642. Lemos, M.C., Morehouse, B.J., 2005. The co-production of science and policy in integrated climate assessments. Global Environmental Change 15, 57–68. Maturana, H.R., Varela, F.J., 1998. The Tree of Knowledge: The Biological Roots of Human Understanding. Shambhala Publications. Miller, C.A., 2004. Climate science and the making of a global political order. In: Jasanoff, S. (Ed.), States of Knowledge. The Co-Production of Science and Social Order. Routledge, London, pp. 46–66. Moss, S., Pahl-Wostl, C., Downing, T., 2001. Agent-based integrated assessment modelling: the example of climate change. Integrated Assessment 2, 17–30. Patt, A.G., Klein, R.J.T., de la Vega-Leinert, A.C., 2005. Taking the uncertainty in climate change vulnerability assessment seriously. Comptes Rendus Geoscience 337, 411–424. Pihlström, S., 2001. Putnam and Rorty on their pragmatist heritage: rereading James and Dewey. Paper presented at AIER/BRC Symposium: Dewey: Modernism, Postmodernism and Beyond Great Barrington, July 20–22, 2001, MA, USA. http://www.helsinki.fi/science/ commens/papers/pragmatistheritage.htm Preskill, H., Torres, R., 1999. The role of evaluative enquiry in creating learning organisations. In: Easterby-Smith, M., Araujo, L., Burgoyne, J. (Eds.), Organisational Learning and the Learning Organisation. Sage, London. Putnam, H., 2002. The Collapse of the Fact/Value Dichotomy and Other Essays. Harvard University Press, Cambridge, MA. Rahmstorf, S., Zickfeld, K., 2005. Thermohaline circulation changes: a question of risk assessment. Climatic Change 68, 241–247. Ravetz, J., 2005. The post-normal science of safety. In: Leach, M., Scoones, I., Wynne, B. (Eds.), Science and Citizens. Zed Books, pp. 43–53. Rayner, S., Lach, D., Ingram, H., 2005. Weather forecasts are for wimps: why water resource managers do not use climate forecasts. Climatic Change 69, 197–227. Reason, P., 2002. The practice of co-operative inquiry. Systemic practice and Action Research 15, 169–176. 181 Rorty, R., 1991. Solidarity or objectivity? In: Rorty, R. (Ed.), Objectivity, Relativism and Truth. Cambridge University Press, Cambridge, pp. 21–34. Senge, P., 1990. The Fifth Discipline. The Art and Practice of the Learning Organisation. Doubleday, New York. Senge, P., 2003. Creating desired futures in a global society. Reflections 5, 1–12. Simon, H.A., 1955. A behavioral model of rational choice. Quarterly Journal of Economics 69, 174–183. Stoll-Kleemann, S., O’Riordan, T., Jaeger, C.C., 2001. The psychology of denial concerning climate mitigation measures. Global Environmental Change 11, 107–117. SustainAbility (1996a) Engaging Stakeholders—Volume 1—The Benchmark Survey. Series, SustainAbility Ltd. London, UK. http:// www.sustainability.com/home.asp SustainAbility (1996b) Engaging Stakeholders—Volume 2—The Case Studies. SustainAbility Ltd. London, UK. http://www.sustainability.com/home.asp Vallejo, N., Hauselman, P., 2004. Governance and Multistakeholder Processes. International Institute for Sustainable Development. http:// www.iisd.org/pdf/2004/sci_governance.pdf van Asselt, M.B.A., Rijkens-Klomp, N., 2002. A look in the mirror: reflection on participation in Integrated Assessment from a methodological perspective. Global Environmental Change 12, 167–184. van Daalen, C.E., Thissen, W.A.H., Berk, M.M., 1996. The Delft process: experiences with a dialogue between policy makers and global modellers. Global Environmental Change 6, 267–285. van den Hove, S., Le Menestrel, M., de Bettignies, H.-C., 2002. The oil industry and climate change: strategies and ethical dilemmas. Climate Policy 2 (1), 3–18. von Neumann, J., Morgenstern, O., 1947. Theory of Games and Economic Behavior. Wiley, New York. Wack, P., 1985. Scenarios: unchartered waters ahead. Harward Business Review September–October, 73–89. Webler, T., 1995. ‘‘Right’’ discourse in citizen participation: an evaluative yardstick. In: Renn, O., Webler, T., Wiedemann, P. (Eds.), Fairness and Competence in Citizen Participation: Models for Environmental Discourse. Kluwer Academic Publishers, London, pp. 35–86. Welp, M. (Ed.), 2001. Stakeholder successes in global environmental management. Report of Workshop, Potsdam, 8 December 2000. PIK Report No. 70. 49pp. Welp, M., Kasemir, B., Jaeger, C.C. Citizens’ voices in environmental policy: the contribution of integrated assessment focus groups to accountable decision making. In: Coenen, F.H.J.M., Paterson, R., (Eds.), The Promise and Limits of Participatory Processes for the Quality of Environmentally Related Decision-Making, Kluwer Academic Publishers, Dordrecht (in preparation). Wynne, B., 1996. May the sheep safely graze? A reflexive view of the expert-lay knowledge divide. In: Lash, S., Szerszynski, B., Wynne, B. (Eds.), Risk, Environment & Modernity. Towards a New Ecology. Sage, London, pp. 44–83. Wynne, B., 2005. Risk as globalizing ‘democratic’ discourse? Framing subjects and citizens. In: Leach, M., Scoones, I., Wynne, B. (Eds.), Science and Citizens. Globalization and the challenge of engagement. Zed Books, London.