ConSearcher: Supporting Conversational Information Seeking in Online Communities with Member Personas DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Human-centered computing Interactive systems and tools CCS: Information systems Search interfaces

Shiwei Wu, Xinyue Chen, Yuheng Liu, Xingbo Wang, Qingyu Guo, Longfei Chen, Chuhan Shi, Zhenhui Peng · 2026

Read ConSearcher: Supporting Conversational Information Seeking in Online Communities with Member Personas DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Human-centered computing Interactive systems and tools CCS: Information systems Search interfaces by Shiwei Wu, Xinyue Chen, Yuheng Liu, Xingbo Wang, Qingyu Guo, Longfei Chen, Chuhan Shi, Zhenhui Peng (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-19747.txt.

Introduction

Wu and co-authors study information seeking inside online communities. Users often read forums to borrow plans and views from others. A first study with ten users showed a classic chat search tool helped. It also left clear room to do better. Their ConSearcher tool builds member personas on the fly from each query. Users can preview what a like-minded member might ask. In a study with 27 users, it beat two baselines on outcomes and engagement. It also raised worries about over-personalization.

Catalog record

CREATORShiwei Wu, Xinyue Chen, Yuheng Liu, Xingbo Wang, Qingyu Guo, Longfei Chen, Chuhan Shi, Zhenhui Peng
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73427
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.19747

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

ConSearcher: Supporting Conversational Information Seeking in Online Communities with Member Personas DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Human-centered computing Interactive systems and tools CCS: Information systems Search interfaces Shiwei Wu, Xinyue Chen, Yuheng Liu, Xingbo Wang, Qingyu Guo, Longfei Chen, Chuhan Shi, Zhenhui Peng Abstract . Many people browse online communities to learn from others’ experiences and opinions, e.g., for constructing travel plans. Conversational search powered by large language models (LLMs) could ease this information-seeking task, but it remains under-investigated within the online community. In this paper, we first conducted an exploratory study (N=10) that indicated the helpfulness of a classic conversational search tool and identified room for improvement. Then, we proposed ConSearcher , an LLM-powered tool with dynamically generated member personas based on user queries to facilitate conversational search in the community. In ConSearcher , users can clarify their interests by checking what a simulated member similar to them may ask and get responses from diverse members’ perspectives. A within-subjects study (N=27) showed that compared to two conversational search baselines, ConSearcher led to significantly higher information-seeking outcome and user engagement but raised concerns about over-personalization. We discuss implications for supporting conversational information seeking in online communities. 1. Introduction Many people come to online communities to seek information from other members who have diverse backgrounds and experiences but have similar interests ( 17 ; 13 ; 24 ; 1 ) . For example, people can learn from the member-contributed posts and comments to construct a travel or bodybuilding plan ( 15 ) and acquire visual design knowledge ( 28 ) . Recently, the integration of LLM-powered conversational search tools into online communities has become increasingly common ( e.g., Reddit Answer, RedNote’s conversational search agent). While these tools facilitate natural language queries and efficiently summarize relevant information, they tend to flatten the distinct standpoints and rich details embedded in user-generated content which are essential for sensemaking given users’ diverse priorities. For example, when planning a trip, a budget backpacker might prioritize a station’s transport links, whereas a family traveler might find the same station’s crowds overwhelming, preferring a quiet apartment in a residential district instead. Previous HCI work has studied the usage of conversational search but has largely focused on its comparisons to traditional web search, e.g., regarding biased information querying ( 34 ) and user-preferred search tasks ( 43 ) . It is under-explored how a classic conversational search tool is used and can be improved for information seeking within an online community, which values the personal experiences, opinions, and suggestions contributed by human members ( 42 ; 15 ) . For personalized decision-making tasks ( e.g., career exploration ( 20 ) ), sensemaking goes beyond fact retrieval. It involves identity construction ( 41 ) ( i.e., identify who they are and clarify their latent needs) and dialectical engagement with diverse perspectives ( 45 ) to understand trade-offs. To fill this gap, we first implemented a classic conversational search tool (noted as BaseAgent ) using community data, which retrieves posts and comments related to the user query, generates an answer grounded on retrieved content, and recommends follow-up questions based on dialog history. We then conducted an exploratory study with ten participants to understand their practices and challenges when using BaseAgent for information seeking within online communities. We found that participants had a back-and-forth interaction between BaseAgent and original community content, i.e., they conversed with BaseAgent to get compact answers and suggested follow-up queries, and they read posts and comments to situate the information and develop interests in other members’ contexts. However, participants reported challenges in clarifying personal interests in the queries and getting responses from different members’ perspectives, which limit the helpfulness of BaseAgent . To address these challenges, we introduced the concept of member personas, i.e., the compact and authentic representations of community members who seek information (noted as seeker ) of their interests and provide information (noted as provider ) from different perspectives. The seeker personas can help users identify their information needs by looking at members with similar interests in the community, while the provider personas can enable the conversational search tool to offer human-like responses from different members’ standpoints. HCI researchers have explored the usage of persona to act as target users for eliciting user needs ( 21 ; 8 ; 33 ; 49 ) and facilitating design decisions ( 27 ; 6 ; 35 ) , and engage users in thoughtful interaction with social media content ( 45 ; 37 ) . However , it remains unclear how member personas can be derived and incorporated to support conversational search within an online community. To instantiate member personas, we developed ConSearcher , which dynamically generates member personas that are grounded on posts and comments related to user queries. With ConSearcher , users can view and customize the generated personas of information seekers, who have clear interests in the factors ( e.g., travel arrangements, accommodation, and sightseeing) about the general search intent ( e.g., Japan travel plan). Users can select their interested queries that these simulated seekers are likely to ask, and get responses and follow-up queries from the conversational agents role-playing different information providers ( e.g., anime filmmaker, cultural explorer). To enable these system features, we developed an LLM-powered pipeline that 1) decomposes the initial search keywords into related factors, 2) groups posts relevant to each factor for generating personas of information seekers and the queries they are likely to ask, 3) groups comments relevant to these queries for generating personas of information providers, and 4) adopts these personas to guide each agent’s response to the user query. To evaluate the impact of ConSearcher on information seeking in online communities, we conducted a within-subjects study with 27 participants, asking them to seek information on given topics ( i.e., make a Japan travel plan, integrate benefits of digital and traditional education, reflect on whether pursuing a PhD or not) in relevant communities. The results revealed that compared to BaseAgent and another baseline (noted as BaseSearcher ) with personas of information seekers but not providers, ConSearcher led to significantly more fruitful information-seeking outcome and engaging experience. ConSearcher ’s suggested queries and generated responses were also rated significantly more satisfying, helping users clarify their needs and get contextual answers from diverse perspectives. However, the member personas sometimes complicated the information-seeking process and raised concerns on over-personalization. Our contributions are as follows: • We conducted an exploratory study to identify the challenges and opportunities for enhancing conversational information seeking within online community contexts. • We develop ConSearcher , an LLM-powered system that supports users to clarify information needs and get contextual answers from diverse perspectives by interacting with data-driven member personas. • Our within-subjects user study demonstrates the strengths of ConSearcher for improving the outcome and experience of conversational information seeking in online communities. • We provide implications for LLM-powered, persona-driven information-seeking support in online communities. 2. Related Work We first review literature about information seeking support in online communities, which focuses on human members’ shared experiences, opinions, and suggestions. Then, as we propose conversational search and persona methods in online communities,

Read the complete text