Modeling Variation in Empathic Sensitivity Using Go/No-Go Social Reinforcement Learning.

Area of Science:

• Computational behavioral modeling
• Social learning
• Reinforcement learning

Background:

• Social learning, particularly concerning aggression, is understudied.
• Individual variation in social behavior requires rigorous quantification.

Purpose of the Study:

• To quantify individual differences in social learning using a novel reinforcement learning task.
• To investigate the relationship between social learning, empathy, and antisociality.

Main Methods:

• Adapted a go/no-go reinforcement learning task with social and non-social contexts.
• Monetary outcomes affected the subject, a partner, or no one.
• Quantified sensitivity to others' rewards/losses and Pavlovian influences.

Main Results:

• Participants learned from partner's rewards and punishments similarly to self-relevant outcomes.
• Sensitivity to others' punishments inversely correlated with trait antisociality.
• Sensitivity to others' losses related to motivation but not self-reported empathy.

Conclusions:

• Social learning may engage empathic processes, as individuals respond to partners' outcomes akin to their own.
• Computational modeling of social reinforcement learning offers a new method to assess empathic sensitivity.
• Findings link social learning mechanisms to individual differences in antisocial behavior.