J. M. Vidal and Ed Durfee
We build an economic society of agents in which buyers and sellers compete with each other, and try to increase their total values and total profits, respectively. We give a precise method for the construction of these agents and for the incremental incorporation of modeling capabilities, following the intuitions behind the Recursive Modeling Method (RMM). These agents were built and run in a simulated economic society. Early test results show that deeper (i.e. l-level) models are more effective in heterogeneous societies than in homogeneous ones, and that price volatility is a good predictor of the relative benefits of deeper models.