Center for Behavioral Decision Research

CBDR : Seminar Series : Seminar by Julie Fiez

Dorsomedial Striatum, Reinforcement Learning, and Declarative Knowledge Acquisition

	presented by Julie Fiez (University of Pittsburgh)


	Thursday, March 5	link to paper
	12pm
	Porter Hall 223D	link to Speaker's Site

	Abstract:

	Doya (2000) highlighted three neural systems for learning: a statistical learning system involving the cerebral cortex that adapts to consistently co-occurring inputs, a reinforcement learning system involving the striatum (or basal ganglia) that uses positive and negative outcomes to optimize the accrual of future rewards, and an error-based learning system involving the cerebellum that detects and corrects performance errors. These systems would seem to be prime targets for research in the learning sciences, but they are often thought to represent relatively slow, inflexible, and consciously inaccessible forms of procedural memory that have little do with the formation of declarative knowledge, the acquisition of deep expertise, and the use of past experience to guide complex decision- making. We are investigating a contrastive view: that humans can strategically use reinforcement learning signals mediated by a particular sector of the striatum -- the dorsomedial striatum -- to achieve rapid and robust learning that is accessible to explicit awareness. In this view, investigations of the dorsomedial striatum should be central to efforts to understand when and how humans use feedback to modify their understanding of the world, and then use what they have learned to guide their future behavior.

	Host at CMU: Morewedge