Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Loss-calibrated attention in a neural sampling model of perceptual decision-making

Khayla Black¹, Ralf M Haefner¹; ¹University of Rochester

Presenter: Khayla Black

Attention is a central component of perception and cognition, yet its computational and neural basis remains debated. Attention is commonly conceptualized as the allocation of limited resources, but the nature of this resource remains unclear. Prior models link it to metabolic constraints, while other accounts emphasize selection mechanisms such as biased competition or perceptual load. Here, we develop the idea that the resource is computational: if the brain performs approximate inference by generating a finite number of samples from a posterior distribution, then these samples constitute the resource. We develop this idea in a 2AFC discrimination task, building on a neural sampling model that predicts task-specific neural and behavioral signatures and extending it to two locations to model spatial attention. The model predicts improved performance at attended locations and reproduces canonical neural signatures of attention, including increased gain, reduced response variability, and systematic changes in noise correlations.

Topic Area: Decision-Making, Cognitive Control & Event Cognition