Brain meeting: Dr Rani Moran

Brain meeting

An extensive body of psychological and neuroscientific literature on Reinforcement Learning indicates that behaviour is governed by two distinct systems— a rigid, retrospective model-free (MF) system and a flexible, prospective model-based (MB) system. Unlike the MF system, which tends to repeat actions that were successful in the past, the MB system uses representations of transition structures and reward-functions to plan efficient courses of action. In my talk, I will present novel theoretical ideas and behavioural findings that broaden our understanding of the scope of MB functions and system interactions. I will first show that MB planning is a sophisticated self-reflective process that takes into account one’s own habits. Thus, planners prioritize plans that cohere rather than conflict with the momentary state of their own MF propensities. Second, I will show that the MB system engages not only in planning but also in retrospective-inference, a process that resolves state-uncertainty and thus guides credit assignment.

There will be coffee, tea and cake in the conservatory directly after the talk.