Conversational Linguistic Features Inform Social-Relational Inference

Abstract

Whether it is the first day of school or a new job, individuals often find themselves in situations where they must quickly and accurately learn novel social networks. Prior work has shown that success in learning social networks predicts more positive individual and social outcomes, but the mechanisms through which social network learning occurs are unclear. We posit that individuals use linguistic features of observed conversations to identify the valence and strength of social relationships. To investigate this, 57 adults completed a naturalistic behavioral task wherein they watched an episode of the reality television show, Survivor, and indicated which contestants they thought were friends or rivals throughout the episode. In Study 1, we take a person-focused approach using 34,735 participant observations to understand how well individuals learned the social network structure. In Study 2, we take a novel stimulus-focused approach that employs natural language processing to quantify 486 sentences of episode dialogue and examine how distinct linguistic features (similarity, emotional tone, self-assurance) predict network learning. In Study 3, we analyze the time course of the entire preceding season of Survivor using 296 dyadic similarity assessments to explore the association between semantic similarity and relationship formation. Across three studies, we found that participants successfully learned the social network structure, linguistic features predicted network learning, and greater semantic similarity was associated with friendship formation. These findings suggest that naturalistic conversational content is both a potential mechanism of social network learning and a promising avenue for research on social relational inference.

People are represented and remembered through their relationships with others

“…Our findings lend additional support to this view and suggest how we remember people is similarly shaped by the process of determining how and through whom people are connected rather than how similar they are along abstract social dimensions (Tavares et al, 2015) or what social group features they have in common (Son et al, 2021). Our findings also complement recent naturalistic work demonstrating that participants use linguistic features from observed conversations to make accurate inferences about changes in social networks (Schmidt et al, 2022), underscoring the importance of social interactions in social learning and prediction.…”