Fostering Collective Discourse: A Distributed Role-Based Approach to Online News Commenting

TL;DR

This study tackles fragmentation and polarization in online news discussions by introducing a distributed, role-based commenting system that organizes conversations through clustering, summarizing, and threading. Implemented as a browser extension and evaluated with a within-subject design ($N=38$) across CNN articles, the system uses three user levels to collaboratively structure discourse, aided by AI prompts. Results show increased engagement with shorter, more balanced contributions, reduced emotional expression, and improved comprehension during reading and writing phases, but with trade-offs in depth and potential delays in updating the discourse. The work advances design space for collective discourse by outlining how distributed roles, structured workflows, and AI-assisted supports can foster more constructive participation while highlighting key design tensions, such as workload, echo-chamber risk, and preserving novelty. The findings inform practical considerations for deploying collaborative discourse tools in real-world news ecosystems and point to future work on motivated participation, dynamic role assignment, and scalable governance.

Abstract

Current news commenting systems are designed based on implicitly individualistic assumptions, where discussion is the result of a series of disconnected opinions. This often results in fragmented and polarized conversations that fail to represent the spectrum of public discourse. In this work, we develop a news commenting system where users take on distributed roles to collaboratively structure the comments to encourage a connected, balanced discussion space. Through a within-subject, mixed-methods evaluation (N=38), we find that the system supported three stages of participation: understanding issues, collaboratively structuring comments, and building a discussion. With our system, users' comments displayed more balanced perspectives and a more emotionally neutral argumentation. Simultaneously, we observed reduced argument strength compared to a traditional commenting system, indicating a trade-off between inclusivity and depth. We conclude with design considerations and trade-offs for introducing distributed roles in news commenting system design.

Figures (6)

• Figure 1: Assigned roles for each user level: Lower levels contribute to smaller units of the discussion space, ordered as clusters, summarizations, and threads. All users can comment, while clustering, summarizing, and threading are collaboratively created and reviewed across different user levels. AI assists by suggesting summarizations and thread topics.
• Figure 2: The overview of our system: (a) guiding discussion topics, three of which are initially generated by AI, (b) discussion thread showing the overview of the corresponding comment section for each discussion topic, (c) summaries of clusters displayed in each discussion thread, along with timestamps for the summaries created, and (d) guiding question to prompt the discussion in each discussion thread, generated when the topic is created
• Figure 3: Workflow of Clustering: LV0 users propose clusters by dragging and dropping comments into the desired locations (left). LV1 users then review these clusters by comparing the changes before and after the clustering activity. The clusters become visible to all users once they are approved by the required number of LV1 users (right).
• Figure 4: Workflow of Summarizing: (1)LV1 users summarize accepted clusters by clicking the 'Summarize' button within the cluster. (2) A modal will display an AI-generated summary suggestion, which users can revise or replace with their own (left). (3) Once finalized, the summary becomes visible to all users and is displayed at the top of the cluster (right).
• Figure 5: Workflow of Threading: (1) LV2 users propose new thread topics by selecting from AI-suggested topics or by suggesting their own. (2) Users review these topics by approving or declining each proposed thread. (3) A new discussion thread is created and becomes visible to all users once a topic is approved by the required number of LV2 users.