How social networks influence human behavior: An integrated latent space approach for differential social influence

Abstract

How social networks influence human behavior has been an interesting topic in applied research. Existing methods often utilized scale-level behavioral data to estimate the influence of a social network on human behavior. This study proposes a novel approach to studying social influence that utilizes item-level behavioral measures. Under the latent space modeling framework, we integrate the two interaction maps for respondents' social network data and item-level behavior measures. The interaction map visualizes the association between the latent homophily of the respondents and their behaviors measured at the item level in a low-dimensional latent space, revealing the potential, differential social influence effects across specific behaviors measured at the item level. We also measure overall social influence as the impact of the interaction map configuration contributed by the social network data on the behavior data. The performance and properties of the proposed approach are evaluated via simulation studies. We apply the proposed model to an empirical dataset to demonstrate how the students' friendship network influences their participation in school activities.