The explicit representation of context in agent-based models of complex adaptive spatial systems

Abstract

The dynamic behavior of complex adaptive spatial systems is driven by interactions that occur among system components. The proper representation of individual-individual and individual-environment interaction in agent-based simulations of such systems is, therefore, of paramount importance to model design. In this article we develop and implement a context-driven, agent-based modeling approach that supports the explicit representation of situation-dependent information for decision making within dynamic spatial environments. This context-driven approach is characterized by the identification of primitive contextual elements, the organization of such elements into context patterns, and contextualized learning. We synthesize two existing frameworks for the representation of spatiotemporal context and extend them to include higher level constructs and feedback mechanisms needed for learning and adaptation. Contextualized information is organized, analyzed, and used by agents to enhance their problem-solving capabilities. Dynamically changing information relevant to the decision-making processes of agents is captured in this approach and used to drive spatiotemporal learning in agents. We examine the utility of this context-driven approach via an agent-based model of elk movement. Elk in this model are represented as contextually aware intelligent agents that learn optimal movement patterns and adapt to heterogeneous landscape dynamics. The internal and external stimuli received by elk during migration are incorporated into the representation of agent context. The context-driven approach, as experimental results indicate, provides solid support for the representation of individual-centric interactions within complex adaptive spatial systems. © 2010 by Association of American Geographers.