The document discusses applying theory of mind, or the ability to infer the intentions of other agents, to multi-agent reinforcement learning. It introduces three papers that use Bayesian reasoning to model other agents in the Hanabi game. Specifically, the papers develop methods for Bayesian action decoding and simplified action decoding to enable agents to reason about each other's intentions during both learning and testing. The document notes challenges in multi-agent reinforcement learning like non-stationarity from other agents learning and incomplete information about other agents. Applying theory of mind techniques may help address these challenges by allowing agents to infer what other agents know or intend to do.