Research Interests
What are the representations, computations, and approximations that allow people to navigate a complex and social world with only limited cognitive resources?
The study of human cognition and artificial intelligence are tightly intertwined. Research on human learning has directly inspired algorithms for training artificial agents, while advances in AI have provided novel insights into the nature of human intelligence. Although some machine learning algorithms have reached human-level expertise, there remain important gaps between the efficiency of human and machine learning. How do people learn so rapidly and from so few examples? And how do we reason about and learn from other people?
To address these questions, we study human behavior through online experiments in the form of an interactive game, in the lab using computers or virtual reality equipment, or using naturally occurring data from online games. We then use computational models inspired by reinforcement learning or machine learning methods, to understand key features of human cognition, such as how we rapidly learn in novel environments or how we use Theory of Mind to infer the hidden beliefs of other people. These models are then validated by predicting human behavior, neural activity, or through evolutionary simulations.
Ultimately, the goal of the lab is to understand the specific shortcuts and cognitive algorithms that separates human intelligence from our best artificial algorithms.
