A Computational Analysis on the Role of Social Relationships in Online Communication and Information Diffusion

Abstract

Social relationships play a crucial role in shaping daily conversations and information sharing within social networks, both in person and on online platforms like Twitter and Facebook. These platforms have become immensely popular for accessing a wide range of information. While previous studies have contributed to understanding the properties of social ties, less attention has been devoted to directly identifying the characteristics of individual social relationships and their influence on dyadic interactions in online social networks.

In this dissertation, I present three computational studies to identify and analyze the key characteristics of social relationships within large online social networks. These studies seek to shed light on how social relationships impact interactions and information diffusion. The first study approaches relationships through the lens of social dimensions, such as conflict or trust, wherein a dyad exhibits varying levels of dimension strength. The findings indicate that the strength of inferred dimensions accurately represents the nature of social relationships in Twitter ties. Additionally, these inferred dimensions can reflect community-level outcomes, such as the stability of organizations or well-being indices. The second study proposes a novel method for identifying different types of interpersonal relationships using a combination of text- and network-based features. Linguistic and diurnal communication patterns are found to differ significantly among various types of relationships, and it is possible to build accurate classifier models for inferring categories of social relationships based on communication on Twitter. Moreover, incorporating information about these relationships enhances the performance of retweet prediction models. Building upon the relationship classification model developed in Study 2, the third and final study investigates the responses of dyads of users facing unexpected life-shock events. Interestingly, the research uncovers relationship-specific reactions to different types of shocks, providing valuable insights into how social ties are influenced during challenging times. The findings from the three computational studies provide a comprehensive understanding of the dynamics of social relationships in the digital age.