Foolproof Cooperative Learning
This work addresses the challenge of fostering cooperation in multi-agent systems, particularly in game theory, but it is incremental as it extends existing concepts from matrix games to stochastic games.
The paper tackles the problem of achieving cooperative strategies in stochastic games without being exploited by selfish players, introducing Foolproof Cooperative Learning (FCL) which converges to Tit-for-Tat behavior and is proven to be a learning equilibrium in repeated symmetric games.
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.