Simulation approach to learning problem in hypergame situation by genetic algorithm

This paper presents a simulation approach to adaptation process of two interacting parties (or groups), each of which adopts learning behavior in hypergame situation. That is, we try to clarify which learning behavior facilitates the adaptation process to convergence on equilibria of the traditional game situation (TGS), and facilitates each agent to learn the equilibria correctly. First, we define the hypergame situation, in which each agent is assumed to have only internal model of the situation. Then, we develop adaptation process model of the groups, and a simulation of the process. In the model, genetic algorithm (GA) has role to improve population of perceptions according to the past experiences. Finally, we point out that by examining the simulation results, action choice and perception evaluation based on subjective Nash equilibria are critical to the performance of the adaptation process, in the situations with one or more TGS Nash equilibria.