Evolutionary instability of selfish learning in repeated games

13 May 2021  ·  Alex McAvoy, Julian Kates-Harbeck, Krishnendu Chatterjee, Christian Hilbe ·

Across many domains of interaction, both natural and artificial, individuals use their past experiences to shape their future behaviors. The result of such learning processes depends on what individuals wish to maximize. A natural objective is one's own success, and indeed this kind of "selfish" learning has become a standard paradigm to model adaptation processes. However, when two selfish learners interact with each other, the outcome can be detrimental to both, especially when there are conflicts of interest. Here, we explore the dynamics that arises when learning rules themselves are subject to evolutionary pressure. By combining extensive simulations and analytical techniques, we demonstrate that selfish learning is unstable in most classical two-player games. Instead, if evolution operates on the level of long-run payoffs, selection favors learning rules that incorporate other-regarding preferences. To further corroborate these results, we analyze data from a repeated prisoner's dilemma experiment. We find that selfish learning is insufficient to explain human behavior when there is a trade-off between payoff maximization and fairness. Our results have important implications on evolutionary game theory, behavioral economics, and multi-agent learning.

PDF Abstract


  Add Datasets introduced or used in this paper

Results from the Paper

  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.


No methods listed for this paper. Add relevant methods here