Neurocomputational basis of learning when choices simultaneously affect both oneself and others

Area of Science:

• Neuroscience
• Cognitive Science
• Social Psychology

Background:

• Prosocial and antisocial behaviors impact self and others, necessitating social learning from joint outcomes.
• Neurocomputational mechanisms underlying social learning from multi-outcome scenarios are not well understood.

Purpose of the Study:

• To investigate the neurocomputational processes supporting social learning from choices with joint self- and other-outcomes.
• To determine whether social learning relies on simulating others' valuations or integrating self- and other-relevant information.

Main Methods:

• Three pre-registered studies involving participants learning choice outcomes for themselves and others.
• Computational modeling to test simulation versus integrated value frameworks for social learning.
• Analysis of brain regions associated with value updating and social decision-making.

Main Results:

• Social learning is characterized by an integrated value framework, not a simulation approach.
• Individuals update choice values using distinct prediction errors for self and other, and for positive and negative outcomes.
• Asymmetric value updating is neurally represented in regions including the ventral striatum, anterior cingulate cortex, insula, and amygdala.

Conclusions:

• Distinct encoding of self- and other-relevant information guides future social behaviors.
• The integrated value framework explains learning across diverse social scenarios, including beneficial, costly, altruistic, and harmful ones.
• Neurocomputational processes involve asymmetric value updates based on prediction errors for self/other and outcome valence.