A NEW UNDERSTANDING OF HUMAN AGENCY
Integrated assessment may be defined as the scientific discipline that integrates knowledge and makes it available for decision processes. Integrated assessment has made considerable progress over the recent years. First approaches relied more or less on models as means for integration. The decision process was perceived as utility maximizing choice of a single decision maker. Measures taken into consideration were mainly of the centralized kind, like taxes. However, given the fact that most global change phenomena result form the added effect of numerous activities at regional scales, IA faces now major challenges that can be be summarized here by adopting a polycentric approach to integrated assessment:
è polycentric regarding the integration of scales across space and time - geographical dimensions and kinds of choice processes.
è polycentric regarding the integration of different levels of societal organization and different types of societal groups and measures. This implies to address a broad range of institutional settings.
The notion of scale, the integration of different scales of analysis is central to this approach. An improved understanding of the nature of decision making processes is of vital importance for both modelling and designing integrated assessment processes.
Current practices to represent the human dimension in Integrated Assessment models have severe limitations. Economics approaches are not sufficient to account for the complex dynamics we encounter in real world systems in general. New approaches that seem to be more appropriate emerge in particular in the field of agent based social simulation. It is intended to build up a more permanent research network among partner institutions to establish links between the Integrated Assessment and the Social Simulation community. In particular the following topics will be addressed:
- Improved understanding of human agency, both individual and social.
- The mutual dependence between individual and collective action.
- The importance of scale for the representation of agency, the level of aggregation, the nature of choice processes.
- Representation of institutions and institutional change.
- The implications of a new understanding of human agency for the predictability of socio-economic systems and decision making under uncertainty.
- Link analytical modelling and participatory approaches.
Examples from the problem domain
To tackle environmental problems proves in most cases to be limited by our lack of understanding the nature of change in human systems, in general and in human-environment systems, in particular. The sustainable management of environmental resources will require major societal transitions and changes in today's socio-economic systems. This invokes in many cases the surmounting of lock-in situations. An example is given by water resource management. Today's system of water resource management grew slowly over the last century. Requirements for changes to account for further environmental needs were always met by extending the current system and by building on the large body of knowledge in the engineering community. Rules of practitioners, technical skills, consumer behaviour co-evolved with the technical infrastructure. Now major changes are impeded since current infrastructure, the shared knowledge of practitioners, expectations of stakeholders stabilize the status quo. Both the inertia in the “hardware” (longevity of infrastructure and high fixed costs) and the “software” (shared rules and habits) prevent change. One may talk of lock-in situations where transitions towards systems that may be better adapted to today's needs are prevented. Today's needs are for example characterized by huge uncertainties in water demand, by increased environmental awareness, by pressure towards cost-efficient solutions, and by fast changing socio-economic boundary conditions. New institutional arangements are required to improve the efficient allocation of scarce water resources. The introduction of water markets – either formal or informal - institutional barriers to implementing them, constraints raising transaction costs are important issues that require new approaches to deal with change and transformation processes in social systems.
Major transitions towards sustainability cannot be brought about by conventional policy measures alone. Conventional policy refers here to centralized political measures such as taxes. We advocate a new approach of a polycentric understanding of policy making that invokes instances of social learning at different levels of societal organization. Social learning is defined here as the mutual shaping of expectations of the involved social groups. The shaping of expectations depends on institutions. In the light of the new approaches in institutional economics, an institution may be defined very broadly as shared rules of human conduct. For example, if one is driving on a road one expects other drivers to respect the red light and stop. Without such shared rules of conduct live in a society would be impossible. Some institutions (laws) are enforced by legislation (e.g. traffic regulations). Others (customs) are shared by the members of a society and evolve and change in a social setting (e.g. shake hands for welcome). Rules enable individuals to form expectations concerning the actions of others. This insight is important for change to overcome lock-in situations where mutually dependent expectations stabilize each other. Institutions also influence the competitiveness of socio-economic systems since they determine transaction costs and performance characteristics. An important research question is the development of concepts and models of institutional and technological change. One need to focus on the mutual relationship between processes of social and individual learning, between individuality and the embeddedness in social networks.
We advocate a new approach to represent human actors in models by so-called software agents. There does not exist an overall accepted and shared definition of an agent. In its most abstract form an agent may be defined as an object with a sensory input and an output to the environment. However, it should be emphasized that we use the notion of an agent for the representation of real human beings – hence neither in the sense of computer science and software engineering nor in the sense of object oriented programming where one may talk of the "agentification" of the models of physical systems.
We identify three dimensions along which the representation of human agency needs to be improved (Fig 1):
§ internal agent complexity referring to the processes inside of an agent
§ agent interaction referring to the embedding of an agent in its social environment
§ agent diversity referring to the characteristic differences among groups of agents.
Fig. 1 Important dimensions to improve the understanding of human agency.
Fig 1 emphasizes the processes inside of an agent that need to be improved for understanding human agency. The thick arrows represent the simple reasoning scheme of the RAP. Information is used to derive subjective representations of the world. According to the concept of bounded rationality in micro-economig theory, two actors with the same state of knowledge should per definition have the same subjective probabilities (e.g. their subjective assessment of the market potential of a new product). In a more advanced perspective one has to acknowledge that the processing of information is inherently motivational. Personal values, previous experience, the embedding in a social network define what one may call a cognitive filter for the aquisition and processing of information. A rather reductionist approach seems to be warranted to improve our understanding for the cognitive processes related to the subjective aquisition and processing of information, the dynamics of preferences and the processes of learning. The ^space – time dependence of knowledge aquisition and learning are important for the scaling behaviour.
Fig. 2 Processes of information processing and decision making within an agent to improve the understanding of internal agent complexity.
Another model to human agency used in the social simulation community is the belief desire intention (BDI) model. It has its roots in the philosophical tradition of understanding practical reasoning of humans. The two main activities that need to be distinguished are the processes of deliberation, deciding what state of affairs to achieve, and means-end reasoning, deciding how to achieve these states of affairs. The BDI models offers a framework to organize processes of human agency. The exact nature of these processes needs to be defined for specific applications.
A simple attempt to replace the RAP by another model seems not to be desirable at the current stage, given the multiplicity of theoretical approaches to deal with cognition. One may as well extend the RAP by accepting that rational agency is defined as action in one's own interests given ones beliefs about the world. This leaves ample room for defining how to determine what are ones own interests and how subjective beliefs can be derived. It is time for an exploratory phase where different approaches will florish. What may be required are coherent frameworks for description that allow to derive a "taxonomy" of human behaviours: e.g. the goal directed planning engineer, the profit maximizing investor, the need satifying and habit driven consumer etc.
Another important question refers to the level of aggregation at which an agent should be represented. How should we represent a firm, an association? Economic theory has devoted much effort to derive procedures for aggregation to come to the so-called representative agent device. This implies for example that in most economic models a "representative consumer" is used to represent consumer behaviour for a whole nation. One may also state that the RAP implies the nature of human agency to be independent of scale – it can be applied to the individual as much as to the choice processes of the whole of humankind. In the extreme, in some models dealing with the implication of climate change, a single decision maker optimizes the unique welfare function of the whole of humankind. Given the presence of agent heterogeneity and local interaction patterns among agents, a simple scheme of aggregation cannot be derived (economic theory is based on centralized interactions via the price mechanism only). The representation of human agency at different scales of aggregation is a research topic of paramount importance. It is closely related to our understanding of institutions. To understand the emergence of norms and the dynamics of institutions has to take into account the embeddedness of indviduals in social networks and the internal representation of institutions (e.g. shared norms, rules) in an individual's mind. There are only a few attempts that may offer guidance regarding cultural change at the level of a society as a whole.
A research agenda around the questions addressed above will provide the base for an improved understanding of evolutionary change in socio-economic systems in general and in human environment-systems in particular. It will be vital to improve our understanding for the limitations to predictability and the indeterminacy inherent in the dynamics of socio-economic systems. It should link theoretical and applied research, approaches focusing on agents (representation of individual human actors with cognitive function of varying complexity) and approaches focusing on system behavior (interaction of agents). Figure 3 sketches the main areas of research:
Fig. 3 Important dimensions of research that have to be bridged for deriving an improved understanding of human environment systems.
Fig. 3 shows the different dimensions of research: Starting from the focus of complex individual agents and acknowledging cognition as a source for complexity is rather new. The system's perspective has a longer tradition. Here agents are typically very simple and the complexity arises from agents' interactions in social networks. There is a certain tradeoff between making individual agents very complex and investigating the dynamics that arise from agents' interactions. However, a new approach to human agency has to take both levels into account.
An improved understanding for the complexity and indeterminacy of social systems leads also to an improved understanding for the nature of decision processes and the application of models for decision making. The development and application of models should be closely linked to participatory processes as outlined in Fig.4.
Fig. 4 Suggested relationships between different activities of research and different types of models to derive a new research agenda for improving the understanding of human-environment systems and for approaches to joint problem solving in participatory settings. Agent based models are particularly suited for applications in participatory settings to foster processes of social learning and to provide assistance in specific decision making processes.
Applied models are taylored for applications in specific problem domains. Conceptual models are abstract models that include only the mere essentials for explaining stylized facts derived from empirical data. It will be important to assure that from the applications in specific case studies generic approaches regarding for example the representation of agent behaviour and agent types will emerge.
The suggested tight link between modelling and participatory approaches is based on the insight that a new understanding of human agency, the nature of decision processes and the limitations to prediction call also for an improved understanding for the use of models. Analytical and modelling approaches should be tightly linked to participatory settings. How can one build models in a participatory fashion and how to apply them in participatory settings. Applied agent based models are based on factual knowledge derived from domain experts and empirical data. The validation in the participatory setting is based on their potential how they structure the debate about a problem by integrating factual and local knowledge.
The steps to take
A first workshop to intitiate the network is planned for next year. We invite responses to the suggested research and network. In a first stage the network will run a series of workshops, lecture notes and foster the exchange of researches among the participating institutions.
Furubotin, E.G., and Richter, R. 2000. Institutions and economic theory. The University of Michigan Press, Ann Arbor.
Gibson, C., Ostrom, E., and Ahn, T.-K. 1998. Scaling Issues in the Social Sciences. IHDP Working Paper No 1. IHPD, Bonn.
Kottonau, J., Burse, J., and Pahl-Wostl, C. submitted. A Consumer Memory-based Model of New Product Diffusion within a Social Network.
Minsch, J., Feindt, P.-H., Meister, H.-P., Schneidewind, U., Schulz, T. 1998. Institutionelle Reformen für eine Politik der Nachhaltigkeit. Springer, Berlin.
Morgan, G.M., and H. Dowlatabadi 1996. Learning from integrated assessment of climate change. Climatic Change 34: 337-368.
Moss, S., Pahl-Wostl, C., and Downing, T. in press, Integrated Assessment. Agent based integrated assessment modelling: The example of climate change.
Pahl-Wostl, C., Jaeger, C.C., Rayner, S., Schär, C., Van Asselt, M., Imboden, D.M., Vckovski, A. Integrated Assessment of Climate Change and the Problem of Indeterminacy, in: Views from the Alps: Regional Perspectives on Climate Change eds P. Cebon, U. Dahinden, H.C. Davies, D.M. Imboden and C.C. Jaeger, The MIT Press, Cambridge MA, 1998, pp. 435-498.
Rotmans, J. and Dowlatabadi, H. Integrated Assessment of Climate Change: Evaluation of methods and strategies, in: Human Choice and Climate Change eds S. Rayner and L. Malone, Battelle Press, Columbus Ohio, 1998.
Simon, H. 1982. Models of Bounded Rationality, The MIT Press, Cambridge, MA.
Tillman, D., Larsen, T., Pahl-Wostl, C. and Gujer, W. 1999. Modelling the actors in water supply systems. Water Science and Technology, 39: 203-211.
Wooldridge, M. 2000. Reasoning about rational agents. The MIT Press, Cambridge, MA.
Young, O.A., Agrawal, A., King, L.A., Sand, P.H., Underdal, A., and Wasson, M. 1999. Institutional Dimensions of Global Environmental Change. IHDP Report No. 9. IHDP, Bonn, Germany.