When a child spends hours building a Lego rocket ship or running a fictional restaurant, are they developing their extraordinary learning abilities? Conventional wisdom suggests they are. But despite many empirical investigations into this claim, it has proved difficult to substantiate. 

In collaboration with a world-class team of developmental and computational cognitive scientists, primary investigators Tomer Ullman and Alison Gopnik will explore how play can help people and machines learn. They will use a novel computational model of play to understand two central questions: Why do humans play? And how do they develop new goals and set new objectives?

Typically, computational systems have a well-defined set of goals or problems that the system must solve. Humans, however, have the ability to pose brand new problems and goals and develop new ways to solve them. Play—which, by definition, is designed without a specific purpose—may provide an important underpinning of this ability. 

This project proposes that much of children’s imaginative and exploratory play arises from “competence-directed curiosity”—a fundamental motivation to figure out how they can achieve goals. By adopting a novel computational framework, it will explore the full richness and potential of play.