Title: Flexible decision-making through long-term integration of evidence
Abstract: Many decisions in natural environments depend on combining information over long timescales. The mechanisms that underlie such decisions have been rarely quantified. However, based on limited time constants of neural network models developed to explain decision-making over shorter timescales, one may predict acute limitations for evidence integration over long timescales. I will explore two classes of decision-making processes: decisions that involve integration of discrete sensory information separated by long gaps, and decisions to revise strategy for future choices based on the outcome of past choices. For both types of decisions, humans and monkeys approximate a normative integration process that extends over tens of seconds without accruing significant memory leak or noise, contrary to existing neural network models. My lab is exploring neural mechanisms that could support these properties using fMRI and large-scale extra-cellular recordings from a network of frontoparietal cortices. I suggest that supplementary and pre-supplementary motor cortices as well as hippocampus may be key for implementing evidence integration over long timescales.
Hosted by Doris Tsao