Reinforcement learning produces dominant strategies for the Iterated Prisoner’s Dilemma

Marc Harper; Vincent Knight; Martin Jones; Georgios Koutsovoulos; Nikoleta E. Glynatsi; Owen Campbell

doi:10.1371/journal.pone.0188046

Loading metrics

Open Access

Peer-reviewed

Research Article

Reinforcement learning produces dominant strategies for the Iterated Prisoner’s Dilemma

Marc Harper ,

Contributed equally to this work with: Marc Harper, Vincent Knight

Roles Conceptualization, Methodology, Software, Writing – original draft, Writing – review & editing

Affiliation Google Inc., Mountain View, CA, United States of America
⨯
Vincent Knight ,

Contributed equally to this work with: Marc Harper, Vincent Knight

Roles Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing

* E-mail: knightva@cardiff.ac.uk

Affiliation Cardiff University, School of Mathematics, Cardiff, United Kingdom

http://orcid.org/0000-0002-4245-0638

⨯
Martin Jones ,

Roles Conceptualization, Software

¶‡These authors also contributed equally to this work.

Affiliation Independent Researcher, Edinburgh, Scotland
⨯
Georgios Koutsovoulos ,

Roles Conceptualization, Software

¶‡These authors also contributed equally to this work.

Affiliation INRA, Université Côte d’Azur, CNRS, ISA, Nice, France
⨯
Nikoleta E. Glynatsi ,

Roles Visualization, Writing – original draft

¶‡These authors also contributed equally to this work.

Affiliation Cardiff University, School of Mathematics, Cardiff, United Kingdom
⨯
Owen Campbell

Roles Software, Writing – review & editing

¶‡These authors also contributed equally to this work.

Affiliation Independent Researcher, Chester, United Kingdom
⨯

Reinforcement learning produces dominant strategies for the Iterated Prisoner’s Dilemma

Marc Harper,
Vincent Knight,
Martin Jones,
Georgios Koutsovoulos,
Nikoleta E. Glynatsi,
Owen Campbell

Published: December 11, 2017
https://doi.org/10.1371/journal.pone.0188046

Reader Comments

Post a new comment on this article

not impressed

Posted by pknuth on 26 Mar 2018 at 21:29 GMT

This is an empirical study of just a bunch of 2-player IPD games. Not clear what---if any---insights are gained from this paper. Far, far more interesting results (and consequently, more likely to gain insight into why humans cooperate with unrelated individuals) is with IPD games where there are

1) N>2 players
2) alternative strategies are allowed (e.g., altruistic punishments or loners)

Re-running tournaments where people are playing against TFT etc. is hardly groundbreaking.

No competing interests declared.

Subject Areas
?

For more information about PLOS Subject Areas, click here.
We want your feedback. Do these Subject Areas make sense for this article? Click the target next to the incorrect Subject Area and let us know. Thanks for your help!

Artificial neural networks
Is the Subject Area "Artificial neural networks" applicable to this article?

Thanks for your feedback.
Prisoner's dilemma
Is the Subject Area "Prisoner's dilemma" applicable to this article?

Thanks for your feedback.
Hidden Markov models
Is the Subject Area "Hidden Markov models" applicable to this article?

Thanks for your feedback.
Machine learning algorithms
Is the Subject Area "Machine learning algorithms" applicable to this article?

Thanks for your feedback.
Evolutionary algorithms
Is the Subject Area "Evolutionary algorithms" applicable to this article?

Thanks for your feedback.
Machine learning
Is the Subject Area "Machine learning" applicable to this article?

Thanks for your feedback.
Neural networks
Is the Subject Area "Neural networks" applicable to this article?

Thanks for your feedback.
Game theory
Is the Subject Area "Game theory" applicable to this article?

Thanks for your feedback.

Reinforcement learning produces dominant strategies for the Iterated Prisoner’s Dilemma

Reinforcement learning produces dominant strategies for the Iterated Prisoner’s Dilemma

Reader Comments

Post Your Discussion Comment

Why should this posting be reviewed?

Thank You!

not impressed

Posted by pknuth on 26 Mar 2018 at 21:29 GMT