--- title: "Conversational User Interfaces in Healthcare: A Scoping Review of ACM CUI Conference Contributions" authors: - Nidhi Kothari - Andreia Valente - Mark Billinghurst year: 2026 venue: "ACM Conversational User Interfaces 2026 (CUI '26)" doi: "10.1145/3816046.3816289" url: "https://andreia-valente.com/publications/kothari2026conversational.html" pdf: "https://andreia-valente.com/pdfs/kothari2026conversational.pdf" topics: - Conversational User Interfaces - Healthcare - Scoping Review - Agency - Evaluation --- # Conversational User Interfaces in Healthcare: A Scoping Review of ACM CUI Conference Contributions ## Citation Metadata - Authors: Nidhi Kothari, Andreia Valente, Mark Billinghurst - Venue: ACM Conversational User Interfaces 2026 (CUI '26) - Year: 2026 - DOI: https://doi.org/10.1145/3816046.3816289 - HTML: https://andreia-valente.com/publications/kothari2026conversational.html - PDF: https://andreia-valente.com/pdfs/kothari2026conversational.pdf ## Plain-Language Summary This paper maps how healthcare has been studied within the ACM Conference on Conversational User Interfaces. It reviews 22 healthcare-related full papers from ACM CUI between 2019 and 2025 and analyzes them through four linked dimensions: healthcare context, conversational design, agency and authority, and evaluation practice. The review shows that healthcare CUI research in this venue is concentrated in prevention, wellbeing, behavior change, and long-term self-management, while more clinically embedded or high-authority uses remain comparatively rare. The central finding is that healthcare CUIs in this corpus mostly advise, guide, and support users rather than making decisions. No reviewed system exercised directive authority. This matters because healthcare CUI design is not only a question of modality or language model capability; it is also a question of who holds responsibility, how much authority is delegated to the system, and what evidence is required before such systems can be trusted in practice. ## Research Questions - What healthcare subdomains and stages of care are represented in ACM CUI research? - How are healthcare CUIs designed in terms of modality, conversation strategy, and target users? - How are roles, agency, and authority distributed between users and conversational systems? - How are healthcare CUIs evaluated, and how is acceptance reported? ## Method The paper conducts a venue-bounded scoping review following established scoping review guidance and PRISMA-ScR reporting principles. The search was conducted in March 2026 through the ACM Digital Library and manual enumeration of ACM CUI proceedings. From 377 records, the authors retained full papers, screened 133 full papers, and included 22 papers that addressed healthcare, care, or wellbeing through a conversational interface. The analysis used a four-pillar charting framework. The healthcare context pillar captured subdomain and stage of care. The design pillar captured modality, target user, conversation strategy, and design approach. The agency pillar captured the role and authority of the CUI. The evaluation pillar captured study type, evidence, and acceptance measures. A thematic analysis of reported limitations and challenges complemented the taxonomy. ## Key Findings - Healthcare CUI research in ACM CUI clusters around prevention, behavior change, mental health, wellbeing, and self-management. - Systems are usually positioned as assistants, guides, coaches, or supportive companions rather than decision-makers. - No system in the reviewed corpus exercises directive authority, revealing a ceiling in how much responsibility these systems currently take. - Conversation strategies remain largely structured or hybrid, even as generative AI becomes more visible. - Evaluations are mostly short-term and usability-focused, with limited clinical, longitudinal, or ecologically valid evidence. - Recurring challenges include privacy, trust, response naturalness, limited personalization, modality constraints, system failures, weak context awareness, and difficulty supporting multi-user care situations. ## Detailed Review Results The review found that healthcare CUI work at ACM CUI became more visible after 2023. The included corpus contained no healthcare full papers from 2019 or 2020, two from 2021, none from 2022, five from 2023, seven from 2024, and eight from 2025. Most papers addressed prevention, wellbeing, behavior change, or long-term management rather than diagnosis, acute care, treatment planning, or clinical decision-making. Behavior intervention and self-management were the largest application area, appearing in 7 of the 22 included papers. Prevention was the most common stage of care, represented in 12 of 22 papers, while long-term management appeared in 6 of 22 papers. The reviewed systems usually targeted patients, consumers, or the general population rather than clinicians. Older adults appeared in 4 of the 22 papers, and most systems for older adults used voice interaction. The design analysis showed that structured dialogue remained common: 9 of 22 systems used structured conversations, and 5 used hybrid designs. Among papers that reported a design process, most used user-centered or participatory methods. Evaluation practice was dominated by short-term studies: 13 of 22 papers used qualitative evaluations, 7 used experimental designs, and 2 used surveys. Technical robustness, long-term use, ecological validity, and clinical outcome evidence were rarely central evaluation targets. The agency analysis distinguishes informational assistants, guiding coaches, supportive companions, and systems with stronger authority. The reviewed systems stayed mostly advisory. This preserves user control and reduces immediate clinical risk, but it also limits integration into higher-stakes healthcare workflows unless future systems define clear boundaries, escalation procedures, provenance, accountability, and evidence standards. ## Limitations and Research Gaps The paper explicitly frames the findings as venue-bounded rather than a complete map of all healthcare CUI research. The review is limited to full papers from ACM CUI, used a single reviewer workflow, did not use a registered protocol, and depended on author-reported system descriptions. The corpus also reflects early-stage and Global North research patterns. The main research gaps are longitudinal clinical validation, stronger technical evaluation, more work with healthcare professionals and affected communities, better support for multi-user care settings, and clearer treatment of safety, privacy, escalation, and responsibility when conversational agents move beyond low-authority support. ## Why This Paper Matters This review is useful for researchers designing conversational agents for healthcare because it shows where the field has been cautious and where evidence is missing. It argues that stronger healthcare CUI roles require stronger evaluation standards, clearer authority boundaries, explicit escalation pathways, and greater involvement from healthcare professionals and affected users. For AI and HCI researchers, the paper is especially relevant to discussions of agency, accountability, clinical decision support, and responsible deployment of conversational systems in health contexts. ## Recommended Citation Kothari, N., Valente, A., & Billinghurst, M. (2026). Conversational user interfaces in healthcare: A scoping review of ACM CUI conference contributions. In ACM Conversational User Interfaces 2026 (CUI '26). Association for Computing Machinery. https://doi.org/10.1145/3816046.3816289 ```bibtex @inproceedings{kothari2026conversational, author = {Kothari, Nidhi and Valente, Andreia and Billinghurst, Mark}, title = {Conversational User Interfaces in Healthcare: A Scoping Review of {ACM} {CUI} Conference Contributions}, year = {2026}, booktitle = {ACM Conversational User Interfaces 2026 (CUI '26)}, publisher = {Association for Computing Machinery}, doi = {10.1145/3816046.3816289}, url = {https://doi.org/10.1145/3816046.3816289} } ```