Effects of configuration of agents with different strategy representations on the evolution of cooperative behavior in a spatial IPD game

Abstract

It is well-known in the literature that the choice of a strategy representation scheme has a large effect on the evolution of cooperative behavior in an IPD (Iterated Prisoner's Dilemma) game. Different representation schemes often lead to totally different results. The evolution of cooperative behavior is easy under some representation schemes while it is difficult under other schemes. An interesting research issue is to examine the evolution of cooperative behavior in an inhomogeneous population consisting of heterogeneous agents with different representation schemes. In our former study, we used an inhomogeneous population where one of two representation schemes was randomly assigned with the same probability to each cell in a two-dimensional grid-world. In this setting, every agent is likely to have both types of representation schemes in its neighborhood. That is, an IPD game is likely to be played not only between homogeneous agents with the same representation scheme but also between heterogeneous agents with different representation schemes everywhere in the grid-world. In this paper, we examine the effect of spatial configurations of two types of agents on the evolution of cooperative behavior in a spatial IPD game. In computational experiments, agents with one representation scheme are clustered in a specific region of a two-dimensional grid-world (e.g., 7×7 cells at the center of an 11×11 grid-world). Agents with the other representation scheme are placed in the remaining cells. In this setting, the interaction through the IPD game between heterogeneous agents is limited only around the boundary between the two types of agents in the grid-world. Experimental results are compared between this setting and the random assignment in our former study.