ConSearcher: Supporting Conversational Information Seeking in Online Communities with Member Personas

Why Read This from an HCI Perspective

This article frames information seeking in online communities as an HCI challenge—not just simple search, but a process in which users articulate their needs, compare different viewpoints, and construct meaning. In particular, it shows how combining ‘member personas’ with LLM-based conversational search can be designed to balance contextuality, engagement, and personalization. For practitioners, it offers interface design cues; for researchers, it prompts consideration of the risks of over-personalization and bias.

CIT's Commentary

From a CIT perspective, the core contribution of this study is that it reinterprets ‘conversational search’ not as an answer-generation tool, but as a meaning-construction tool that structures users’ interests and surfaces the community’s diverse voices. In particular, the design that separates seeker/provider personas is compelling from an HCI standpoint because it supports both users’ self-understanding and their adoption of others’ perspectives during information seeking. However, it is important to examine how faithfully dynamically generated personas reflect the real diversity of the community, and at what moments over-personalization may narrow the scope of exploration. Going forward, the system should be designed with transparency, editability, and confidence indicators so users can distinguish between ‘helpful personalization’ and ‘trapping personalization.’

Questions to Consider While Reading

• Q.How can we verify whether dynamically generated member personas sufficiently represent the diversity of perspectives among actual community members?
• Q.What interaction mechanisms would be most effective in mitigating the problem of over-personalization narrowing the breadth of exploration?
• Q.Can the design that separates seeker and provider personas in the context of information seeking be generalized to other domains?