Origins of Social Intelligence in Human Infants.
Abstract: We cannot help but see ourselves and others as intentional agents, who have goal states of the world in mind, and plan costly actions in order to realize them. My research program investigates the origins of these abilities in human infants, and aims to describe these origins in engineering-specific terms. I ask: What are the representations and computations that support infants’ analysis of other people’s goal-directed actions? In this talk, I will present evidence that when infants observe other agents act, they appreciate that agents seek to minimize the cost of their actions (Liu & Spelke, 2017), trade off the cost of acting against the rewards actions bring (Liu, Ullman, Tenenbaum, & Spelke, 2017), and act so as to cause changes in the world (Liu, Brooks, & Spelke, 2019). I will end by briefly presenting newer work on infants' understanding of dangerous situations (e.g. cliffs and trenches) (Liu, Ullman, Tenenbaum, & Spelke, under review), their co-developing intuitive physics and psychology, and my future plans to develop rich measures of infants' learning and reasoning. Collectively, this work shows that human infants rapidly develop a skeletal understanding of other people as intentional causal agents acting in a physical world, which could serve to help them learn about other minds and from other minds.