Hierarchical Bayesian inference underpins human social learning

Abstract: Bayesian theories of brain function view perception as an active process with the brain constantly generating predictions about the causes of sensory inputs to anticipate future events. Since our knowledge about states in the world is incomplete, the extent to which surprising signals (or prediction errors) update predictions depends on their uncertainty. The role of uncertainty and (its inverse, precision) is particularly critical for social contexts where predictions about others’ hidden intentions are based on incomplete or ambiguous inputs. Consistent results from EEG and functional MRI studies suggest that the brain utilizes Bayesian inference machinery, in particular hierarchical precision-weighted prediction errors to generate a model of another person and his/her intentions. This computational approach is now extended to psychiatry, in the context of disorders where theory of mind deficits predominate.

Brief Bio: Dr Andreea Diaconescu is an Independent Scientist at the Krembil Centre for Neuroinformatics (CAMH) and Assistant Professor in the Department of Psychiatry at the University of Toronto.

Andreea completed her PhD in Cognitive Neuroscience with Prof. Randy McIntosh at the Rotman Research Institute (University of Toronto). Following her PhD training, she held a postdoctoral position at the Translational Neuromodeling Unit (University of Zurich and ETH Zurich). Under the supervision of Prof. Klaas Enno Stephan, she developed, validated, and applied hierarchical Bayesian models of social learning and decision-making to examine persecutory ideation in early psychosis. Andreea has recently been supported by the Swiss National Foundation at the University in Basel in the Department of Psychiatry to lead a project on early detection of psychosis using computational models of persecutory delusions fit to behaviour and neuroimaging (EEG and fMRI) data.

At the Krembil Centre for Neuroinformatics, Andreea focuses on neurocomputational models of suicidal ideation in psychoaffective disorders.

See below for link to join via zoom:

Topic: KCN Event: Andreea Diaconescu

Time: Dec 10, 2020 10:00 Eastern Time (US and Canada)

https://utoronto.zoom.us/j/89681767784

Meeting ID: 896 8176 7784

Passcode: 593617

Recording available on kcnhub youtube channel:

https://www.youtube.com/watch?v=QC2E0wstV4s&t=294s