Proceedings of the Annual Meeting of the Cognitive Science Society

Beyond an ability to represent other people’s mental states,people can also represent different types of minds, like those ofnewborn babies, pets, and even wildlife that we rarely interactwith. While past research has shown that people have a nu-anced understanding of how minds vary, little is known abouthow we infer what kind of mind different agents have. Here wepresent a computational model of mind attribution as Bayesianinference over a space of generative models. We tested ourmodel in a simple experiment where participants watched shortvideos in the style of Heider & Simmel, 1944, and had to in-fer the representations in the agent’s mind. We find that, fromjust a few seconds, people can make accurate inferences aboutagents’ mental capacities, suggesting that people can quicklyinfer an agent’s type of mind, based on how they interact withthe world and with others.