Marcelo Mattar

My lab studies the neural computations that generate intelligent, goal-directed behavior. Using mathematical models, behavioral experiments, and neural recordings, we try to uncover how our memory systems build internal models of the world, and how we use these representations to simulate the future when making a decision. We draw heavily from the fields of Artificial Intelligence and Computational Neuroscience, particularly the area of Model-Based Reinforcement Learning.

Research in the lab consists of developing mathematical models of learning and decision making -- typically expressed in the language of Reinforcement Learning and Bayesian statistics -- and evaluating theoretical predictions against behavioral and neural data from humans. We also collaborate closely with other experimentalists, including those specialized in animal electrophysiology and psychiatric disorders in humans. We hope our work will help us understand the computations underlying healthy and pathological cognition, and inform the development of brain-inspired AI algorithms that approach human performance.