Wealth exchange models and machine learning: Finding optimal risk strategies in multiagent economic systems

Abstract

The Yard-Sale Model, a well known wealth exchange model whose observed macroscopic behavior hides many underlying aspects of particular complexity, was studied at the microscopic level. The performance of the agents during the successive transactions allows for the definition of successful or disadvantageous strategies according to the profit they achieve at the end of the process. Optimal strategies were found that maximize the individual wealth of each agent by performing their training through a genetic algorithm. The addition of different levels of rationality given by the amount of available information from their environment showed promising results, at both the microscopic and macroscopic levels. Remarkably, after the training process, the rational agents were able to determine when it would be convenient to interact with their opponents. Additionally, a region of parameters was found for which the distribution of wealth is a power law throughout the whole wealth range. As a general result, the incorporation of rational agents in this type of systems leads to greater inequality at the collective level.