Striatal dopamine-wave mechanisms for agentic reward learning
Abstract:
In this talk, I will explore computational motivations and empirical evidence for regionally specialized dopamine (DA) signals for learning and performance in fronto-striatal networks. I will focus on the cognitive dorsomedial striatum as a case study to illustrate that regional DA signals, arriving as spatiotemporal wave-like patterns, are specialized to local functions, in stark contrast to the standing view of globally broadcast DA error signals. These findings connect novel spatiotemporal circuit specializations employed by DA to implement computations described by formal reinforcement learning models in the service inferring agentic control.