Bayesian inference has had great success in modeling cognition on a computational level, but this success raises two questions: what is happening on the algorithmic level in the brain to produce these computational-level results, and how can we explain systematic differences between actual performance and what Bayesian models predict? I will study these questions in the context of the Bayesian number game, which examines concept learning. I will model empirical data with stochastic algorithms based on general-purpose inference. My results may lead us closer to understanding the algorithmic processes used in the human brain and designing better artificial intelligence systems.

I have worked with Dr. Fedorenko and Professor Kanwisher at MIT studying the neural basis of language with fMRI. I have worked with Professor Winston at MIT CSAIL helping to build a computer vision system incorporating feedback.