Social Cognition as Reinforcement Learning: Feedback Modulates Emotion Inference.

Area of Science:

• Neuroscience
• Social Cognition
• Reinforcement Learning

Background:

• Traditional social cognition research often uses single-instance judgments of strangers.
• Real-world social inference involves repeated interactions and feedback, crucial for learning about known individuals.
• Understanding the interplay between social cognition and reinforcement learning is key to real-world social understanding.

Purpose of the Study:

• To investigate how individuals learn about others' emotional states through repeated interactions and feedback.
• To explore the intersection of social cognition and reinforcement learning mechanisms in forming social impressions.
• To identify neural correlates associated with learning social information and updating impressions over time.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to scan participants.
• Participants repeatedly inferred the emotions of three social targets presenting conflicting cues (visual and verbal).
• Reinforcement learning models estimated learning signals to correlate with brain activity.

Main Results:

• Participants successfully updated their social judgments based on feedback over repeated encounters.
• Distinct cue types (visual vs. verbal) predicted target emotions differently, aiding learning.
• Brain activity in the ventral striatum, ventromedial prefrontal cortex (vmPFC), and temporoparietal junction (TPJ) correlated with learning signals and social impression updating.

Conclusions:

• Learning about others' emotions relies on domain-general reinforcement learning mechanisms.
• Social information processing, particularly updating impressions, involves both domain-general and domain-specific neural systems.
• This study bridges social cognition and reinforcement learning, highlighting feedback-driven learning in social contexts.