Humans flexibly integrate social information despite interindividual differences in reward.

Area of Science:

• Cognitive Science
• Behavioral Economics
• Computational Neuroscience

Background:

• Human social learning research has advanced, incorporating social information into reinforcement learning.
• Previous models often assume identical payoffs, limiting realism in diverse social interactions.

Purpose of the Study:

• To develop a more realistic model of social learning that accounts for payoff differences between individuals.
• To investigate the flexibility of human social learning in integrating diverse social information.

Main Methods:

• Introduced a socially correlated bandit task to simulate realistic payoff variations.
• Developed and tested the Social Generalization (SG) model using evolutionary simulations and online experiments.

Main Results:

• The SG model outperformed existing models in predicting social learning behavior.
• Found that social information is treated as noisier than individual observations in the SG model.
• Evidence suggests a resource-rational trade-off where social learning can substitute for individual exploration.

Conclusions:

• Human social learning is highly flexible, adapting to varying social contexts and individual differences.
• The SG model provides a more accurate framework for understanding how humans integrate social information from diverse sources.
• This research opens new avenues for studying adaptive social learning strategies in complex environments.