Hierarchical prediction errors in midbrain and septum during social learning.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Social Cognition

Background:

• Social learning is crucial for human interaction but its underlying computational and physiological mechanisms remain unclear.
• The specific roles of neuromodulatory transmitters in social learning are a key area of investigation.

Purpose of the Study:

• To investigate the neuromodulatory mechanisms of social learning by examining belief updates about advice validity and adviser trustworthiness.
• To differentiate the neural encoding of low-level and high-level prediction errors in social contexts.

Main Methods:

• Functional magnetic resonance imaging (fMRI) combined with computational modeling and genetic analysis.
• Participants engaged in a task requiring inference about an adviser's changing intentions.
• Analysis of prediction errors related to advice accuracy and adviser fidelity.

Main Results:

• Low-level prediction errors (PEs) about advice accuracy activated the dopaminergic midbrain and theory of mind regions.
• High-level PEs, termed 'expected uncertainty', about adviser fidelity engaged the cholinergic septum.
• Dopaminergic midbrain responses were modulated by the Catechol-O-Methyltransferase (COMT) gene polymorphism.

Conclusions:

• Social learning involves hierarchically organized PEs encoded by distinct midbrain and septal systems, analogous to other learning processes under volatility.
• Dopaminergic and cholinergic projections may broadcast these hierarchical PEs to cortical areas, facilitating plasticity in belief representation about others.