Iterated Prisoner’s Dilemma: An extended analysis

Abstract

Abstract-The motivation of this work is to investigate the classic problem of the Iterated Prisoner's Dilemma in an environment with many players with different behaviors. We review previous analysis of the tournaments proposed by Robert Axelrod, focusing primarily on the strategies adopted by their players. These agents have been recreated in this new tournament, based on the information available, and several others were included, from periodically non-rational agents to agents based on modern paradigms such as neural networks, genetic algorithms, reinforcement learning, trackers, etc. This gives heterogeneous and rich alternatives to the basic tournament. As a result, we verified the classification of these players for various game modes (variation of maximum number of iterations), and analyze their performance based on this criterion.

Dominance, sharing, and assessment in an iterated Hawk–Dove game

“…The large variety of different models often makes it difficult for biologists without detailed mathematical knowledge to understand the differences and similarities in the implications of different models. In general, models in which resources are ultimately divided unequally between 'winners' and 'losers' (Eshel and Sansone, 2001;Eshel, 2005;Fawcett and Johnstone, 2010;Hammerstein, 1981;Houston and McNamara, 1991;Kura et al, 2015Kura et al, , 2016Mesterton-Gibbons et al, 2014 are often based on the Hawk-Dove framework described in Maynard Smith (1979), whereas models that con-centrate on the evolution of sharing and other apparently paradoxical acts of cooperation (Baek et al, 2017;Carvalho et al, 2016;Doebeli and Hauert, 2005;Nowak and Sigmund, 1993a,b;Nowak, 2012;Trivers, 2006) are often based on the Iterated Prisoner's Dilemma (IPD) framework described in Axelrod (1984). This distinction poses the question of whether one needs to consider rather different models for outcomes with cooperation versus overt aggression.…”