Analogical Reasoning for Intent Recognition and Action Prediction in Multi-Agent Systems

The ability to infer internal states of others is a hallmark of human social reasoning. It is central to both interpersonal interactions and team dynamics. This includes reasoning about relationships between others, such as cooperative and competitive relationships. Here, we present a system that enables virtual agents to exhibit the same social reasoning abilities. In particular, we show that analogical reasoning is sufficient for modeling relationships between virtual agents in stag-hunt, a simple stochastic prisoner’s dilemma-style game, without the need for an explicit underlying model of cooperation. Furthermore, we show that the analogical model can predict agents’ future actions based on their prior behavior.