Cooperation in the Iterated Prisoner’s Dilemma is Learned by Operant Conditioning Mechanisms

D. GUTNISKY, B.S. ZANUTTO

ARTIFICIAL LIFE

The Mit Press

Lugar: Estados Unidos; Año: 2004 vol. 10 p. 433 - 461

1064-5462

Abstract The prisoner’s dilemma (PD) is the leadingmetaphor for the evolution of cooperative behavior inpopulations of selfish agents. Although cooperation in theiterated prisoner’s dilemma (IPD) has been studied for overtwenty years, most of this research has been focused onstrategies that involve nonlearned behavior. Anotherapproach is to suppose that players’ selection of the preferredreply might be enforced in the same way as feeding animalstrack the best way to feed in changing nonstationaryenvironments. Learning mechanisms such as operantconditioning enable animals to acquire relevant characteristicsof their environment in order to get reinforcements and toavoid punishments. In this study, the role of operantconditioning in the learning of cooperation was evaluated inthe PD. We found that operant mechanisms allow the learningof IPD play against other strategies. When random moves areallowed in the game, the operant learning model showed lowsensitivity. On the basis of this evidence, it is suggested thatoperant learning might be involved in reciprocal altruism.