ELSC Seminar Series

Learning through rewards is a fundamental aspect of human behavior, and it is mediated by the dopamine system in the brain. Traditionally, the dopamine system was thought to facilitate learning by transmitting a single error signal to the brain’s learning circuits. Computationally, this is analogous to the implementation of a simple algorithm (temporal difference learning) previously discovered by the machine learning community. However, recent research has challenged this notion by demonstrating that the dopamine system exhibits significant diversity in its activity, which is inconsistent with the single error signal hypothesis. What does this mean for our understanding of how the brain learns? In the talk, I will explore these recent findings and describe the research we are conducting in our lab to elucidate how diversity in dopamine activity can underlie learning. Our work is based on the hypothesis that this diversity is especially important when learning something complex.