Inferring the internal structure of groups through the integration of statistical learning and causal reasoning

Area of Science:

• Cognitive Science
• Social Psychology
• Computational Neuroscience

Background:

• Human social interactions form complex networks (friendships, hierarchies).
• Observable interactions are often sparse and noisy, hindering structural inference.
• Understanding social networks is crucial for social cognition and behavior.

Purpose of the Study:

• To investigate how humans infer latent social structures from limited interaction data.
• To test a computational model integrating statistical learning and social models.
• To determine the mechanisms underlying social network inference and prediction.

Main Methods:

• Three behavioral experiments using abstract videos of social interactions.
• Participants inferred social structures, predicted behavior, and reasoned about influence.
• A computational model based on statistical learning and causal reasoning was developed and tested.

Main Results:

• Participants successfully inferred underlying social structures.
• Judgments aligned with predictions from the computational model.
• Performance could not be explained by simpler cue-based models.

Conclusions:

• Humans integrate domain-general statistical learning with domain-specific social models.
• This integration forms causal representations for social understanding.
• Statistical learning and causal reasoning work together for flexible social cognition.