Complex Systems

Simulating the Evolution of Behavior: the Iterated Prisoners' Dilemma Problem Download PDF

David M. Chess
Computing Systems Department, IBM T. J. Watson Research Center,
Post Office Box 218, Yorktown Heights, NY, USA


A system is described in which a number of artificial agents (represented by simple mathematical expressions) compete for the right to "reproduce" (that is, to cause new agents with similar properties to be generated). By simulating some of the essential features of biological evolution, the system makes possible some novel insights into the behavior of communities of agents over time. The results of Fujiki and Dickinson on the Iterated Prisoners' Dilemma problem (IPD) are essentially confirmed. The typical course of evolution of a community of IPD players is described, and possibilities for further work are outlined. This study is also relevant to machine learning, and adaptive systems in general.