Answers are grounded in this paper and cite their sources.
In many Multi-Agent Systems (MAS), self-interested agents need to cooperate in order to maximize their own utilities in time. The goal of this work is to improve cooperation among agents that use best-response Reinforcement Learning (RL) algorithms (Q-Learning), by the introduction of two new principles (Change or Learn Fast and Change and Keep) that foster the reaching of Pareto efficient stable outcomes.
No takes yet. Share an insight, caveat, or question.
Cote et al. (2006) studied this question.