Characterizing Anxiety Disorders with Online Social and Interactional Networks

Anxiety disorders are closely associated with an individual’s interactions, manifested in the way an individual expresses themselves and interacts with others in their social environment. However, little is explored empirically about the association of social network structure and the interactions of an individual with aspects of mental health functioning, such as anxiety. In recent years, individuals have begun to appropriate social media to self-disclose about their mental illnesses, seek support, and derive therapeutic benefits. The study examines the online social network and interaction characteristics of Twitter users who self-disclose about their anxiety disorders. We analyze a sample of 200 Twitter users and their over 200,000 posts shared on the platform, who were expert-validated to have self-disclosed about suffering from an anxiety disorder. On their data, a variety of attributes of the users’ online social networks, interactions, and social behaviors using natural language and network analysis approaches were modeled using state-of-the-art network science measures. A number of state-of-the-art supervised learning classification frameworks are built using these attributes, to identify whether an individual’s anxiety disorder status could be automatically inferred. Results show that these social network, behavior, and interaction attributes, when incorporated in a support vector machine classifier, signal an individual’s self-reported anxiety disorder status, in contrast to a demographically and activity matched control group, with 79% accuracy and 84% area under the receiver-operating characteristic curve. The work provides the first insights into the role that the social interactions and social network structure on online platforms play in characterizing an individual’s mental health experience, such as anxiety. We discuss the implications of our work in instrumenting online social platforms in ways that yield positive affordances and outcomes for individuals vulnerable to mental illnesses.