Affiliation: New York University

Abstract: Attention and learning are intertwined: attention prioritizes information for learning and past learning determines how attention is deployed. In this talk, I will first cover past work on modeling how visual feature-based attention enhances performance, and then motivate the need to extend this to the question of how attention is guided during learning. I will present modeling work based off of an existing experimental dataset of monkey behavioral choices and neural recordings collected during a color value learning task. Through modeling, we are able to explore the relationship between internal value estimates and top-down attention in this dataset. Our results support the idea that attention is directed toward the color with highest estimated value, but also reverses after trials with high reward prediction error.