Modeling one human decision maker with a multi-agent system

Abstract

In this paper, we propose an agent-based model of human decision making: CODAGE (Cognitive Decision AGEnt). Classical Decision Theories have been widely used in multi-agent systems, but imply a too rational behavior when faced with real-world human data. Moreover, classical model usually exceeds human capabilities. Therefore, we derived our decision model from several cognitive psychological theories (e.g. Simon's decision theory, Montgomery's search of dominance structure, etc.) to take human bounded rationality into account. While most of existing cognitive agents use the BDI framework, we propose a new kind of architecture. In the CODAGE model, the decision maker is modeled by an entire multi-agent system, where each agent is in charge a particular sub-process of the whole decision. The architecture is intended to be as generic as possible. It could be viewed as an agent-based decision framework, in which different decision heuristics and biases could be implemented. We illustrate this approach with a simulation of a small experimental financial market, for which our model was able to replicate some human decision behaviors.