Theory of Mind as Intrinsic Motivation for Multi-Agent Reinforcement Learning
This addresses the challenge of social intelligence for artificial agents in multi-agent settings, though it appears incremental as it builds on existing theory of mind concepts.
The paper tackles the problem of improving social dynamics in multi-agent reinforcement learning by using theory of mind to model other agents' beliefs as an intrinsic reward signal, with preliminary results showing enhanced performance in a mixed cooperative-competitive environment.
The ability to model the mental states of others is crucial to human social intelligence, and can offer similar benefits to artificial agents with respect to the social dynamics induced in multi-agent settings. We present a method of grounding semantically meaningful, human-interpretable beliefs within policies modeled by deep networks. We then consider the task of 2nd-order belief prediction. We propose that ability of each agent to predict the beliefs of the other agents can be used as an intrinsic reward signal for multi-agent reinforcement learning. Finally, we present preliminary empirical results in a mixed cooperative-competitive environment.