No winning strategy in the iterated prisoner's dilemma: Game theory and simulated evolution

Julian García, Matthijs van Veelen

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

Abstract

The iterated prisoner's dilemma (IPD) is a test bed for adaptation and cooperation. Computational experiments are regularly used for studying the competition of IPD strategies in multi-agent settings. This experimental work rarely links their results to game theoretical results with the potential to enlighten the analysis and the questions being asked. Here we focus on simulated evolution and results from Evolutionary Game Theory (EGT) and the IPD. The theory implies that all Nash equilibria can be upset by a sequence of mutants. If strategies are not restricted, populations of agents should move between Nash equilibria with different levels of cooperation. We argue this instability is inescapable, regardless of how strategies are represented. We present algorithms that show that simulated evolution perfectly aligns with EGT predictions. This implies that cognition itself may only have limited impact on the cycling dynamics of cooperation and defection. We argue that the role of mutations or exploration is more important in determining levels of cooperation.

Original languageEnglish
Title of host publicationALA 2019 - Adaptive and Learning Agents Workshop at AAMAS 2019
EditorsPatrick MacAlipine, Patrick Mannion, Bei Peng, Roxana Radulescu
Place of PublicationNew York NY USA
PublisherAssociation for Computing Machinery (ACM)
Number of pages2
Publication statusPublished - 2019
EventAdaptive and Learning Agents Workshop at AAMAS 2019 - Montreal, Canada
Duration: 13 May 201914 May 2019
https://ala2019.vub.ac.be/ (Website)
https://ala2019.vub.ac.be/#program (Proceedings)

Conference

ConferenceAdaptive and Learning Agents Workshop at AAMAS 2019
Abbreviated titleALA 2019
Country/TerritoryCanada
CityMontreal
Period13/05/1914/05/19
Internet address

Cite this