Over the last several decades, there has been a wealth of illuminating work on processes implicated in social cognition. Much less has been done in articulating how we learn the contours of particular concepts deployed in social cognition, like the concept MENTALISTIC AGENT. Recent developments in learning theory afford new tools for approaching these questions. In this article, I describe some rudimentary ways in which learning theoretic considerations can illuminate philosophically important aspects of the MENTALISTIC AGENT concept. I maintain that MENTALISTIC AGENT is an essentialized concept (cf. Gelman, in The essential child, 2003; Keil, in Concepts, kinds, and cognitive development, 1992) and that learning-theoretic considerations help explain why the concept is not tied to particular traits.