Abstract

Evolutionary programming experiments are conducted to examine the relationship between the durations of encounters and the evolution of cooperative behavior in the iterated prisoner's dilemma. A population of behavioral strategies represented by finite-state machines is evolved over successive generations, with selection made on the basis of individual fitness. Each finite-state machine is given an additional evolvable parameter corresponding to the maximum number of moves it will execute in any encounter. A series of Monte Carlo trials indicates distinct relationships between encounter length and cooperation; however, no causal relationship can be positively identified.

This content is only available as a PDF.