Cognitive maps of social characteristics allow flexible inference in social networks



How do people learn about the vast and complex web of social relationships that surround them? We test how people use information on social characteristics (such as being in the same club or sharing hobbies) to fill in gaps in their knowledge of friendships and make inferences about unobserved friendships in the social network. We find that the ability to infer friendships depends on a simple but inflexible heuristic that induces friendship when two people share the same characteristics, and on a more complex but flexible cognitive map that encodes the relationships between characteristics rather than ‘between people. Our results reveal that cognitive maps play a powerful role in the way people represent and think about relationships in a social network.


In order to navigate a complex web of relationships, an individual must learn and represent the connections between people in a social network. However, the size and complexity of the social world makes it impossible to gain direct knowledge of all the relationships within a network, suggesting that people need to make inferences about unobserved relationships to fill in the gaps. . Through three studies (m = 328), we show that people can encode information about social characteristics (e.g., hobbies, clubs) and then deploy this knowledge to infer the existence of unobserved friendships in the network. Using computer models, we test various feature-based mechanisms that might support such inferences. We find that people’s ability to generalize successfully depends on two representational strategies: a simple but inflexible similarity heuristic that takes advantage of homophilia, and a complex but flexible cognitive map that encodes the statistical relationships between social characteristics and friendships. Together, our studies reveal that people can create cognitive maps encoding arbitrary patterns of latent relationships in many abstract characteristic spaces, allowing social networks to be represented in a flexible format. Additionally, these findings highlight open questions across disciplines about how people learn and represent social media, and may have implications for generating more human link prediction in machine learning algorithms.


  • Author contributions: research designed by J.-YS, AB and OFH; J.-YS carried out research; J.-YS, AB and OFH analyzed the data; and J.-YS, AB and OFH wrote the article.

  • The authors declare no competing interests.

  • This article is a direct PNAS submission.

  • This article contains additional information online at

Data availability

The data and code supporting the conclusions of this manuscript are available online on the Open Science Framework at the following URL:

Leave A Reply

Your email address will not be published.