Redesigning Online Discussions for Informed Discourse at Scale
Abstract
Online communities, such as Reddit and Twitter, are the de facto platforms where people with different backgrounds and knowledge converse on diverse issues. Ideal discussions should encompass diverse viewpoints supported with information, and such discussions contribute to the formation of well-informed public opinions on complex issues. However, what we observe from the online communities isan exchange of emotional and subjective arguments built on top of overt simplifications of important issues, exacerbated by the current designs of communities favoring short, provocative content.We frame the problem based on the need for discussion participants# informedness on the subject matter. The lack of understanding of the background knowledge, controversies, and ambiguities surrounding the issue, as well as a failure to appreciate the diversity of perspectives and value systems, can debilitate civil and productivediscussion.To facilitate informed discussion, we propose a novel online discussion platform supporting two design goals: (a) enhancing users# understanding of complex issues through the multifaceted exploration of discussion and (b) supporting the formation and iteration of argumentation by effectively integrating information into argumentation. To support the first goal, we propose a discussion design to support exploring diverse information as well as viewpoints related to the issue, built on top of a human-AI collaborative pipeline for clustering diverse viewpoints from existing discussion threads and identifying potential viewpoints uncovered in the discussion. To support the second goal, we propose a set of novel user interactions to support creating an initial argument, fortifying the argument, and positioning the argument within the discussion thread, by leveraging the power of large language models.We expect that our project would enable more informed discussion on complex social issues and the interactions and the technical innovations from our project could guide future researchers to study online discussions in depth and to build advanced discussion support tools.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2024
- Source ID
- N000142412290
Entities
People
- Juho Kim
Organizations
- KAIST
- Office of Naval Research
- United States Navy