Multiple Levels of Cognitive Modeling within Agent-Based Modeling

We are conducting basic research on the factors affecting the rise of conflict among peoples in an environment with multiple societies, cultures, and biomes. The decisions that some agents make are more likely to have a significant impact on the trajectory of our simulation than those of other agents. This implies that the fidelity of the decision-making process needs to be high in some cases in order to achieve cognitive plausibility and model accuracy. However, to use cognitively-plausible architectures such as Soar (Newell, 1990) or ACT-R (Anderson et al., 2004) to provide such high fidelity for all agents becomes computationally intractable. To meet our needs, we use a hybrid approach. Our approach is to use agent-based modeling (Epstein, 2002; Luke, Cioffi-Revilla, Panait, Sullivan, & Balan, 2005) and to vary the sophistication of the agents’ decision making process with the agent’s level within the simulation’s hierarchy of agents. This extended abstract describes our approach, results, and future work using an existing simulation.