Foolproof Cooperative Learning

TL;DR

This paper introduces Foolproof Cooperative Learning (FCL), an algorithm that achieves cooperative behavior in stochastic and symmetric games, ensuring convergence to a stable equilibrium and robustness against selfish players.

Contribution

The paper extends learning equilibrium concepts to stochastic games and proposes FCL, a novel algorithm that promotes cooperation and resists exploitation in repeated symmetric games.

Findings

FCL converges to Tit-for-Tat behavior in symmetric games.

FCL is a learning equilibrium in repeated symmetric games.

FCL demonstrates robustness to selfish learners.

Abstract

This paper extends the notion of learning equilibrium in game theory from matrix games to stochastic games. We introduce Foolproof Cooperative Learning (FCL), an algorithm that converges to a Tit-for-Tat behavior. It allows cooperative strategies when played against itself while being not exploitable by selfish players. We prove that in repeated symmetric games, this algorithm is a learning equilibrium. We illustrate the behavior of FCL on symmetric matrix and grid games, and its robustness to selfish learners.