Exploring Community-Powered Conversational Agent for Health Knowledge Acquisition: A Case Study in Colorectal Cancer

TL;DR

This work tackles how people learn health knowledge from online communities by building CanAnswer, a community-powered conversational agent that fuses doctor-led content with peer-generated posts and a professional base dataset. The authors develop a computational workflow that grounds responses in professional sources while leveraging real-world cases for empathy and engagement, evaluated through a lab study and expert interviews in the colorectal cancer domain. Results show CanAnswer improves recall and learning engagement with credible grounding, though experts caution about the variability of peer data and emphasize safety/credibility considerations. The study offers design guidelines and demonstrates a generalizable approach for powering health-learning CAs with mixed-source community data, with potential application to other diseases and domains.

Abstract

Online communities have become key platforms where young adults, actively seek and share information, including health knowledge. However, these users often face challenges when browsing these communities, such as fragmented content, varying information quality and unfamiliar terminology. Based on a survey with 56 participants and follow-up interviews, we identify common challenges and expected features for learning health knowledge. In this paper, we develop a computational workflow that integrates community content into a conversational agent named CanAnswer to facilitate health knowledge acquisition. Using colorectal cancer as a case study, we evaluate CanAnswer through a lab study with 24 participants and interviews with six medical experts. Results show that CanAnswer improves the recalled gained knowledge and reduces the task workload of the learning session. Our expert interviews (N=6) further confirm the reliability and usefulness of CanAnswer. We discuss the generality of CanAnswer and provide design considerations for enhancing the usefulness and credibility of community-powered learning tools.

Figures (5)

• Figure 1: CanAnswer's interface and interaction.
• Figure 2: Computational workflow that integrates doctor-led and peer-led community data with a professional base dataset to support health knowledge acquisition through reliable responses, suggested follow-up questions, input suggestions, filtered community posts and topic switching.
• Figure 3: Baseline's interface. The webpages, chat interfaces and documents included in Baseline. Participants can freely choose the information sources they want to browse or interact with.
• Figure 4: Experimental design and procedure.
• Figure 5: (a) Provide suggested follow-up questions; (b) Switch example questions about other topics; (c) Test learning feedback with single-choice questions; (d) Offer input suggestions when the user enters the keyword; (e) Assume a scenario to ask for an explanation; (f) Provide real-world cases for reference.