Conversational Dynamics in Online Support Groups

Funded by: The National Science Foundation
PIs: Robert Kraut , Carolyn Rosé, and John Levine

Online health support groups are popular, being used by almost 2/3 of American adults. If these groups are effective, it is because of the conversations that are their “active ingredient.” The overall goals of the proposed research are to clarify how the conversational dynamics of online cancer support groups influence partici-pants’ health quality of life and to improve the operations of these groups.

Our first goal is to understand how conversations in online cancer support groups produce social support at the level of the conversational episode. Our research will examine the dynamics of peer-to-peer in-teraction that lead these groups to make available different amounts and types of social support. To achieve this goal, we will conduct longitudinal studies using tens of thousands of archived conversational exchanges in online cancer support groups. Our second goal is to understand the impact of online social support on (a) members’ commitment to the group, as indicated by their willingness to remain in it and contribute to it both as information recipients and information providers, and (b) their health quality of life, as indicated by their self-reported distress, depression, and experienced pain. Our studies of commitment will involve quantitative life history analyses of group members who do and do not become active consumers and/or sources of social support. Our studies of health quality of life will involve panel regression analyses predicting changes in health quality based on the social support that participants exchange. Our third goal is to develop a suite of automated content analysis tools (the Analyst’s Helper), using text mining and language processing technol-ogy, to facilitate fast and accurate coding of conversations in online support groups. Such tools are needed be-cause analyzing conversational exchanges in online groups is not practical with hand-coding of messages or dictionary-based automated content-analysis tools. The Analyst’s Helper will enable health researchers to ana-lyze large corpora of conversational interactions in novel ways (e.g., by identifying the types of support that people seek and receive) and will be the basis of interventions to improve support groups. Finally, our fourth goal is to use the Analyst’s Helper to identify helpful and harmful conversations in online support groups and then to develop interventions for improving the training of support-group facilitators and reducing problematic conversations between support-group members.

Health support groups can yield substantial benefits for their members, but the social processes responsible for these benefits need to be specified. The proposed research will utilize theoretically-driven studies to clarify the conversational dynamics underlying the effectiveness of online cancer support groups, will create state-of-the-art tools for analyzing interactions in such groups, and will build interventions for improving group effectiveness. Identifying how communication in online cancer support groups influences participants’ commitment and health quality of life will have both theoretical and applied payoffs. Moreover, creating tools for analyzing large corpora of conversational data will facilitate the work of researchers interested in conversational behavior in other kinds of online groups. Finally, the proposed research will clarify the determinants of fundamental group processes (e.g., member commitment) that are critical to the success of offline as well as online groups.

The proposed work is innovative in several important ways. First, it tests theoretically-interesting questions in the context of online groups that matter deeply to large numbers of people, both cancer victims and their loved ones. Second, it focuses on a critical, but neglected, mechanism underlying the effectiveness of online cancer support groups, namely conversational dynamics between group members. Third, it provides a sophisticated set of methodological tools (the Analyst’s Helper) for analyzing conversations in online groups and enhancing the effectiveness of these groups. And, fourth, it benefits from the expertise of experienced investigators from a range of disciplinary backgrounds, including social, clinical, and health psychology and computational linguistics.

Selected Recent Publications

  1. Mayfield, E. & Rosé, C. P. (in press). LightSIDE: Open Source Machine Learning for Text Accessible to Non-Experts, Invited chapter in The Handbook of Automated Essay Grading, Routledge Academic Press.
  2. Mayfield, E., Adamson, D., & Rosé, C. P. (2013). Recognizing Rare Social Phenomena in Conversation: Empowerment Detection in Support Group Chatrooms, Proceedings of the Association for Computational Linguistics
  3. Wang, Y.-C., Kraut, R. E., & Levine, J. M. (2012). To Stay or Leave? The Relationship of Emotional and Informational Support to Commitment in Online Health Support Groups. In Proceedings of the ACM Conference on Computer-Supported Cooperative Work (CSCW'2012). Best Paper Award New York: ACM Press.
  4. Mayfield, E., Adamson, D. & Rosé, C. P. (2012). Hierarchical Conversation Structure Prediction in Multi-Party Chat. Proceedings of the SIGDIAL 2012 Conference, The 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue, 5-6 July 2012, Seoul National University, Seoul, South Korea. The Association for Computer Linguistics 2012, pp 60-69.
  5. Wen, M. & Rosé, C. P. (2012). Understanding Participant Behavior Trajectories in Online Health Support Groups Using Automatic Extraction Methods, GROUP ’12 Procedings of the 17th ACM International Conference on Supporting Group Work, ACM, New York, pp 179-188.
  6. Mayfield, E., Wen, M., Rosé, C. P., Golant, M. (2012). Discovering habits of Effective Online Support Groups, GROUP ’12 Procedings of the 17th ACM International Conference on Supporting Group Work, ACM, New York, pp 263-272.
  7. Nguyen, D. & Rosé, C. P. (2011). Language use as a reflection of socialization in online communities, in Proceedings of the Workshop on Language in Social Media (LSM 2011), pp 76–85, Portland, Oregon, 23 June 2011.
  8. Nguyen, D., Smith, N. A., & Rosé, C. P. (2011). Author age prediction from text using linear regression, in Proceedings of ACL-HLT 2011 Workshop on Language Technologies for Cultural Heritage, Social Sciences, and Humanities (LaTeCH 2011), pp 115-123.