Developing and Testing the Efficacy of AI-Chatbot-Based

For the first and the second years, Phase I using retrieval-augmented generation (RAG), developing a rule-based "CalmRecoveryJourney" AI-Chatbot step to step guided BMS-MC, feedback system, 24-hour emotional support with human-loop escalation. Phase II Pilot evaluation using a mixed-methods design including the quantitative (a randomized waitlist-controlled trial, RCT) and qualitative methods (in-depth interview) in the 68 breast cancer survivors who completed active cancer treatments and will be recruited from breast cancer organizations, and social media. For the third year, a large-scale single-arm intervention with three repeated measurements (baseline, post-intervention, and follow-up) will be conducted across four geographic regions in Taiwan (North, Central, South, and East). A total of 200 breast cancer survivors from four regions will be recruited. The outcome measures include self-report questionnaires (WHO Well-Being Index and Patient Health Questionnaire-9, PHQ-9, Emotion Regulation Questionnaire, Five Facet Mindfulness Questionnaire, self-compassion scale, EORTC QLQ-BR45, wearable EEG online data and survivors' learning process data from the AI-Chabot system including adherence and engagement with AI-Chatbot. Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3). With intentional-to-treat GEE analysis and thematic qualitative analysis will be conducted.

• AIAI-Chatbot CalmMindRecoveryJourney delivering BMS-MC therapy (Behavioral): CalmRecoveryJourney is a rule-based AI-Chatbot implemented through Line platform which guides survivors step by step complete 6-week Modules: BMS empowerment strategies, Structured Mindfulness skills and Compassion-based skills. The modules for each session contents will be developed based on our previous intervention studies on BMS-MC for breast cancer survivors, lung cancer survivors and their caregivers (Hsiao et al., 2012; Hsiao et al., 2016; Hsiao et al., 2022; Hsiao et al., 2025a; Hsiao et al.,2025b), and update literature review. BMS-MC consists of BMS empowerment strategies and mindfulness/compassion to enhance breast cancer survivors' holistic well-being and emotional regulation skills (mindfulness, self-compassion).

Primary outcomes

• Patient Health Questionnaire-9, PHQ-9 [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3)]: Unabbreviated Scale Title: Patient Health Questionnaire-9 Minimum and Maximum Values: 0 to 27 Outcome Direction: Higher scores indicate a worse outcome (greater severity of depressive symptoms).
• WHO Well-Being Index [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3).]: Unabbreviated Scale Title: World Health Organization-Five Well-Being Index Minimum and Maximum Values: 0 to 25 (often converted to a percentage score from 0 to 100 by multiplying the raw score by 4) Outcome Direction: Higher scores indicate a better outcome (higher level of well-being).

Secondary outcomes

• Emotion Regulation Questionnaire [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3)]: Unabbreviated Scale Title: Emotion Regulation Questionnaire Minimum and Maximum Values: Cognitive Reappraisal Subscale: 6 to 42 Expressive Suppression Subscale: 4 to 28 Outcome Direction: Higher scores indicate greater endorsement/use of that specific emotion regulation strategy. Whether higher scores mean a better or worse outcome depends on context, though Cognitive Reappraisal is generally considered adaptive (better outcome) and Expressive Suppression is often associated with maladaptive outcomes (worse outcome).
• Five Facet Mindfulness Questionnaire [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3).]: Unabbreviated Scale Title: Five Facet Mindfulness Questionnaire (15-item short form) Minimum and Maximum Values: 15 to 75 total score (or 3 to 15 for each of the 5 individual subscales) Outcome Direction: Higher scores indicate a better outcome (higher overall level of mindfulness).
• self-compassion scale [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3).]: Unabbreviated Scale Title: Self-Compassion Scale (or Self-Compassion Scale - Short Form) Minimum and Maximum Values: 1 to 5 mean score (calculated by averaging item scores) Outcome Direction: Higher scores indicate a better outcome (higher level of self-compassion).
• EORTC QLQ-BR45 [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3)]: Unabbreviated Scale Title: European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire - Breast Cancer Module (45 items) Minimum and Maximum Values: Raw scores are transformed linearly into a standardized score ranging from 0 to 100 for all subscales. Outcome Direction: Functional Scales / Single Items: Higher scores indicate a better outcome (a higher/better level of functioning). Symptom Scales / Single Items: Higher scores indicate a worse outcome (a higher level of symptoms or problems).
• wearable EEG power in alpha band [Measurement time-points are at baseline (T1), the 6 week (the end of AI-Chatbot guide) (T2), the 12 week (self-practices period) (T3)]: wearable EEG power in alpha band

Sex: All, ages: 18 Years to any

Inclusion Criteria:

Diagnosis of breast cancer completing the active cancer treatments (surgery, chemotherapy, radiotherapy) Aged at least 20 Owned a smartphone (Android or iOS) with Line access Ability to communicate with the AI chatbot Willingness to participate in the research and interact with the AI-chatbot during the intervention.

Exclusion Criteria:

Diagnosis of another type of cancer Current receipt of psychotherapy Another online psychosocial or mental health intervention The presence of cognitive impairment or severe mental health conditions

• Abd-Alrazaq, A. A., Rababeh, A., Alajlani, M., Bewick, B. M., & Househ, M. (2020). Effectiveness and safety of using chatbots to improve mental health: Systematic review and meta-analysis. Journal of Medical Internet Research, 22(7), e16021. https://doi.org/10.2196/16021 Boie, S. D., Glastetter, E., Lux, M. P., Balzer, F., von Kalle, C., Lenz, C., & Müller, U. (2025). Evaluating a chatbot as a companion for patients with breast cancer: Collaborative pilot study. JMIR Cancer, 11, e68426. Brown, T., Cruickshank, S., & Noblet, M. (2021). Specialist breast care nurses for support of women with breast cancer. The Cochrane Database of Systematic Reviews, 2(2), CD005634. https://doi.org/10.1002/14651858.CD005634.pub3 Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav. Dec 1983;24(4):385. Bornemann B, Singer T. Taking time to feel our body: steady increases in heartbeat perception accuracy and decreases in alexithymia over 9 months of contemplative mental training. Psychophysiology. Mar 2017;54(3):469-482. Chakraborty, C., Pal, S., Bhattacharya, M., Dash, S., & Lee, S.-S. (2023). Overview of chatbots with special emphasis on artificial intelligence-enabled ChatGPT in medical science. Frontiers in Artificial Intelligence, 6. Conţ, M., Ciupe, A., Orza, B., Cohuţ, I., & Niţu, G. (2022). Career counseling chatbot using Microsoft bot frameworks. In I. Kallel, H. Kammoun, & L. Hsairi (Eds.), 2022 IEEE Global Engineering Education Conference (EDUCON) (pp. 1387-1392). IEEE. Duan, W., Klibert, J., Schotanus-Dijkstra, M., Llorens, S., van den Heuvel, M., Mayer, C. H., Tomasulo, D., Liao, Y., & van Zyl, L. E. (2022). Editorial: Positive psychological interventions: How, when and why they work: Beyond WEIRD contexts. Frontiers in Psychology, 13, 1021539. https://doi.org/10.3389/fpsyg

• Study Protocol (PDF, 2026-06-29)

• Fei-Hsiu FH Hsiao, PhD, Principal investigator, National Taiwan University