Prototype Structure and Conceptual Clustering

Agent-based modeling has proven to be particularly effective for fine-grain social modeling. However, so far, the modeling of social agents has not involved dynamics based on interpretive interaction. Meaning or interpretation plays a significant role in human interactions. These micro effects might create interesting effects on aggregate dynamics. The goal of the Interpretive Agent (IA) project is to develop a framework in which agent-based modeling can incorporate interpretive mechanisms. We believe that introducing interpretive mechanisms in agents will improve the depth and quality of the simulation models [Sallach, 2003; Sallach & Mellarkod 2004]. The innovation of interpretive behavior in agent models involves at least the following three mechanisms: prototype inference, orientation accounting and situational definition [Sallach 2003]. These mechanisms follow closely the mode by which human interpretation works. Introducing these mechanisms is not straight-forward, as the computational complexity generated by these mechanisms needs to be carefully aligned and integrated. We are working on each of these mechanisms separately and putting them together in the broad framework. The prototype structures are viewed as clusters formed from objects/events/actions of agents’ responses. Thus, the work summarized in this paper captures the computational aspects of prototype structure and inference. Contact: David L. Sallach, Associate Director Center for Complex Adaptive Agent Systems Simulation (CAS) Decision Information Sciences, Bldg. 900 Argonne National Laboratory Argonne, IL 60439