Content-based social network analysis methods for data-driven health promotion



 DESCRIPTION (provided by applicant): The effects of social relationships on health behavior change and management of chronic conditions has been well documented. Investigators from a variety of disciplines have attempted to understand and harness the role of these social ties in health promotion. Online communities, which digitize peer-to-peer communication, provide a unique opportunity to researchers to understand the mechanisms underlying human behavior change. The onset of digital communication platforms has led investigators to question the use of existing behavior change theories that were formulated in the context of face-to-face communication. Manifestations of socio-behavioral constructs in online communities are often embedded in the content of the messages exchanged by users. Most studies of online and offline social networks have not considered communication content of peer-to-peer interactions when attempting to discern the theoretical underpinnings of network influence. In the proposed research, we will develop a framework for the incorporation of message content into network-based models of social influence. While generally applicable, the methods will be developed based on communications among individuals attempting to quit smoking in an online community, QuitNet, the first online social network for smoking cessation. The methods will be extended to analyze communication among members managing chronic conditions on TreatmentDiaries, an online community where patients share their daily experiences of disease management. Combining qualitative methods with automated text analysis, our work will facilitate high-throughput empirical analysis of user communication in online social networks allowing the identification of emergent user needs and community culture. Temporal models that capture the evolution of cognitive factors pertinent to health behavior change and chronic disease management will be developed. Content-specific communication characteristics of network influence and dependencies will be studied by adapting affiliation exposure models and exponential random graph modeling techniques. This research proposal will result in 1) novel methods to incorporate communication content into network models of social influence to enhance our understanding of the theoretical roots of behavior change in digital communication platforms; 2) scalable techniques to model temporal changes and network dependencies of user interactions in terms of communication content and structure; and 3) new proposals for the development of digital interventions and technology features that harness social ties to support individuals engaging in health behavior change and chronic disease management.