MétaCan
Menu
Back to cohort
Record W4414606802 · doi:10.2196/72553

Generative AI Chatbot for Diabetes Management: Formative 2-Part Qualitative Study Using DTalksBot Involving Patients and Clinicians

2025· article· en· W4414606802 on OpenAlexvenueno aff
Soyun Jeon, Seolhee Lee, Esther Hehsun Kim, Jinsu Eun, Kwangwon Lee, Hajin Lim, Joonhwan Lee

Bibliographic record

VenueJMIR Formative Research · 2025
Typearticle
Languageen
FieldPsychology
TopicDigital Mental Health Interventions
Canadian institutionsnot available
Fundersnot available
KeywordsFormative assessmentChatbotGenerative grammarWorkflowQualitative researchComplement (music)Empirical researchHealth care

Abstract

fetched live from OpenAlex

BACKGROUND: Diabetes mellitus requires continuous self-management to prevent complications. Patients frequently rely on online resources and mobile apps for diabetes-related information; however, these often lead to information overload, limited personalization, and difficulty in navigation. Generative artificial intelligence (AI) chatbots may address these challenges by providing accessible, personalized, and responsive guidance. OBJECTIVE: This study aimed to explore the potential role of generative AI chatbots in diabetes management through a 2-part qualitative evaluation. Part 1 examined patients' information needs, user experiences, and expectations. Part 2 investigated specialists' perspectives on the practical utility of generative AI chatbots in supporting diabetes self-management. By incorporating perspectives from both patients and specialists, the study aimed to identify appropriate boundaries for the involvement of generative AI chatbots, reflecting the needs and expectations of both stakeholder groups. METHODS: This study was conducted using DTalksBot, a generative AI chatbot powered by GPT-4 (OpenAI) and enhanced with retrieval-augmented generation. In Part 1, we aimed to understand the experiences, needs, and expectations of patients with diabetes. To achieve this, 24 participants engaged in structured chatbot sessions, completed postinteraction surveys, and participated in in-depth interviews. Data were analyzed using thematic and content analysis to identify patterns in user queries and experiences. In part 2, we invited 4 family medicine specialists to assess the accuracy of DTalksBot's responses by reviewing conversation logs and to share expert insights on the future role of generative AI chatbots in diabetes management. RESULTS: In part 1, a total of 24 patients submitted a total of 643 questions, which were categorized into 4 primary themes: personalized health advice and guidance (n=281, 44.6%), complications and comorbidities (n=174, 27.1%), medication and treatment exploration (n=111, 17.3%), and mental health management and support (n=30, 4.7%). Patients emphasized the advantages of generative AI chatbots over traditional information sources, including faster access to reliable content, reduced cognitive burden, and the ability to comfortably discuss sensitive topics. In part 2, specialists recognized the generative AI chatbots' value in answering routine inquiries, but noted limitations in contextual accuracy, real-time data integration, and response personalization. CONCLUSIONS: Generative AI chatbots showed promise as complementary tools for diabetes self-management by offering accessible, reliable, and tailored support. This formative evaluation provides empirical evidence on how generative AI chatbots can address patient information needs and complement existing health care resources. To maximize utility, future generative AI chatbots need to integrate real-time health data, enhance contextual relevance, and align with clinical workflows to ensure safety, trust, and broader applicability.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.018
metaresearch head score (Gemma)0.021
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Qualitative · Consensus signal: Qualitative
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.018
Threshold uncertainty score0.096

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0180.021
Meta-epidemiology (narrow)0.0010.001
Meta-epidemiology (broad)0.0010.000
Bibliometrics0.0010.001
Science and technology studies0.0080.006
Scholarly communication0.0030.003
Open science0.0020.007
Research integrity0.0020.003
Insufficient payload (model declined to judge)0.0040.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.185
GPT teacher head0.581
Teacher spread0.396 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designQualitative
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations9
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueJMIR Formative ResearchSame topicDigital Mental Health InterventionsFrench-language works237,207