Towards a Multi-avatar Macroeconomic System

In this paper we provide a prototype Multi-Agent macroeconomic system where artificial agents' behavioral rules are calibrated using human behaviors. The artificial representation of humans, i.e. ``molded'' avatars, derives from a two steps process. The first step consists in performing laboratory experiments in order to gather microeconomic data. The second relies on the use of data mining heuristic techniques on the experimental data which allow for the identification of the behavioral models that better approximate humans' behaviors. The implementation of these techniques, which we call \textit{microeconomic calibration}, leads to the identification of heterogeneous agents. We maintain that the Multi-Avatar systems which arise from the heterogeneous microeconomic calibrated agents give the possibility to improve the existing Agent-based models. The proposed approach will generate consequently more reliable policies evaluation tools. The model discussed in the paper is motivated by the investigation on the role of entrepreneurs' financial decisions over the business cycle. This topic was addressed especially by Hyman Minsky since the Fifties and recently gained a new momentum in the scientific community after the onset of the financial crisis.