Perspective on Agent-Based Modeling of Complex Spatial Systems

Perspective on Agent-Based Modeling of Complex Spatial Systems

Piotr Jankowski

Department of Geography

San Diego State University

My perspective on the complexity of spatial systems and agent-based modeling (ABM) paradigm comes from my view of participatory spatial decision making, which fundamentally is a complex socio-technological system. This view is corroborated by a definition of complex system given by Casti (1999), according to whom it (complex system) involves a number of agents interacting based on limited, partial knowledge. There are obvious parallels between complex spatial systems and participatory spatial decision making. Participatory spatial decision making involves process-driven interactions of agents (e.g. decision makers, experts, members of the public) who make choices based on the partial knowledge of problem domain. Just like in complex spatial systems, in which the system properties result from collective interactions of its constituting parts and where the emergence of properties cannot be traced by separately analyzing the attributes of internal components, so in participatory spatial decision processes, outcomes resulting from interactions of actors cannot be easily traced to actor characteristics.

A key attribute of system complexity is emergence. The term emergence captures the spirit of complexity by mediating the notion of unpredictability, non-linearity, and synergy of individual actions and behaviors that cannot be easily deduced from system’s characteristics. Emergence results from a finite set of atomic rules applied at a local (individual) level among a large number of heterogeneous entities. The dynamic interactions between those individuals are capable of generating complex phenomena on higher systemic levels, which may be perceived as either novel, surprising or counterintuitive, or simply as patterns that cannot be obtained otherwise (Ligmann-Zielinska, 2005). Consequently, complex systems are difficult to frame. Similarly, participatory spatial decision making is difficult to frame. Nyerges and Jankowski (1997) proposed an assessment framework for participatory spatial decision making called enhanced adaptive structuration theory (EAST). The framework outlines a detailed set of concepts and relationships linking the content, process, and outcome of collaborative spatial decision making. The content constructs of EAST examine the socio-institutional, group participant, and technology influences. The process constructs examine the social

interactions between humans and technology. The outcome constructs address

societal impacts of the decisions. The intent of EAST has been to help identify

collaborative spatial decision making structures leading to the emergence of meaningful outcomes. In this sense, the function of EAST has been to help organize participatory decision making processes in order to reduce the element of surprise in emerging outcomes. However, because of complexity of participatory decision making processes, such element of surprise is unavoidable, which makes large-scale participatory decision making a challenge. Another problem to reckon with in large-scale participatory decision making processes is the inherent difficulty in enlisting a representative participation of different groups of the public. This is why the idea of agent-based modeling is particularly attractive.

The idea of simulating participatory decision making processes such as collaborative planning with agent-based modeling environments has been around for quite some time (e.g. Innes and Booher, 1999), but not much progress in terms of working simulation environments has been made. There are many challenges to overcome before virtual agent decision making processes can be used to augment/supplement human decision making processes. One of the main challenges is the representation of human values and behavior in agent-based modeling environments. Various methods of representing individual behavior in agent-based models have ranged from simple rules of thumb, through first-order logic rules to random utility and game theory-based equations. Little attention, however, has been paid to behavioral decision theories explaining human choice making, such as the Prospect Theory of Tversky and Kahneman (1981). Prospect Theory was tested in various situations involving individual choice making but not, to my knowledge, in spatially-explicit decision contexts involving location and spatial pattern as determinants of choice. It may be profitable to employ agent-based simulation to test behavioral decision-making theories such as Prospect Theory in spatially-explicit virtual decision situations. Verifying results of such experiments against empirical results can help us learn and eventually build better representations of spatial decision making behavior.

References

Casti J.L., 1999, The Computer as a Laboratory, Complexity, 4(5), pp.12-14

Innes J.E., and D.E. Booher, 1999, Consensus Building and Complex Adaptive Systems A Framework for Evaluating Collaborative Planning, APA Journal, 65(4), pp. 412-423

Ligmann-Zielinska A., 2005, What is so special in spatial Complex Adaptive Systems?

The role of spatialized agent-based approach in social science, unpublished paper, Department of Geography, San Diego State University

Nyerges, T., and P. Jankowski, 1997, . Enhanced Adoptive Structuration Theory: A theory of GIS-supported Collaborative Decision Making, Geographical Systems, 4:3, pp. 225-257.

Tversky, A., and D. Kahneman, 1981, The Framing of Decisions in the Psychology of Choice, Science, New Series, 211(4481), pp.453-458