LAUSR

Growing evidence shows that social media facilitate diffusion of both pro-recovery and anti-recovery information among people affected by mental health problems, while little is known about the associations of people’s activities in sharing different types of information. Our work explores this question by analyzing a large set of Twitter conversations among users who self-identified as eating disordered. We use clustering algorithms to identify topics shared in online conversations and represent interpersonal interactions by a multilayer network in which each layer represents user-to-user communication on a different topic. By measuring structural properties of the multilayer network, we find that (i) the same set of users form social networks with different structures in communicating different types of information and (ii) exposure to content on body image can reinforce individual engagement in anti-recovery communication and weaken engagement in pro-recovery communication. By measuring structural changes in a sequence of temporal, multilayer networks built based on users’ conversations over time, we further find that (i) actors previously engaged in pro-recovery communication are likely to engage in anti-recovery communication in the future and (ii) actors in anti-recovery communication have frequent entries into and exits from such communication system. Our results shed light on the organization and evolution of communication in online eating disorder communities.