[2O6-GS-5] Agents: game thoery

[2O6-GS-5-03] Dynamics of Cooperation in Iterated Prisoner's Dilemma with Implementation Errors

〇Shinnichiro Murai¹, Ryohei Igarashi¹, Atsushi Iwasaki¹ (1. The University of Electro-Communications)

[[Online]]

Keywords:Evolutionary Games, Repeated Games, Implementation Errors

We analyze how players behave to maintain (or not maintain) cooperation in an environment which they may cause implementation errors from an evolutionary game theoretic point of view. In the previous studies, the Grim-Trigger (GRIM) strategy, which is a representative strategy, was not correctly represented when describing the behavior after an implementation error under memory-one strategy representation. We thus use replicator-mutator dynamics to examine the consequences in a strategy space represented by finite state automata withing two states, that accurately illustrates the behavior of GRIM. As a result, we observe that a variant of GRIM survives by returning to cooperation when they cause implementation errors in an attempt to defect each other. We also show that Tit-For-Tat (TFT) survives by increasing its continuation payoffs by maintaining cooperation when they cause implementation errors.