Seminar of the Mind, Brain and Reasoning PhD Seminar 2021-2022. On line on Zoom. H. 10:00 AM. Abstract
There is increasing consensus around the idea that the brain is a predictive machine, which uses an internal generative model to continuously generate predictions, in the service of online action-perception and future-oriented forms of cognition. In this talk, I will present a theoretical and computational perspective on the functioning of the predictive brain, by appealing to the notions of ‘predictive coding’ and ‘active inference’. In particular, I will focus on how the predictive brain may support prospective and future-oriented functions, such as planning and imagination, by temporarily detaching from the action-perception loop. I will highlight two distinct but complementary usages of planning in computational modelling that might be relevant to understand predictive brain functions: planning ‘at decision time’, to support goal-directed behaviour; and ‘in the background’, to learn behavioural policies and to optimize internal models. I will exemplify this distinction by presenting a series of simulations and by reviewing empirical evidence on internally generated sequences of neuronal activity in the hippocampus during online navigation and offline periods. The perspective offered in this talk suggests that the brain's generative model supports predictive and prospective functions even when it is temporarily detached from the action-perception loop; and this might have implications for our understanding of spontaneous brain activity at rest.
06 maggio 2022