The goal of this randomized controlled trial is to evaluate the effectiveness of a mindfulness-centered stress management AI chatbot in reducing perceived stress and enhancing mental health among Chinese men who have sex with men (MSM). The main questions it aims to answer are: 1. Does the AI chatbot show a greater reduction in perceived stress compared to standardized psychoeducational materials? 2. Does the AI chatbot improve mental health outcomes, including depressive and anxiety symptoms, subjective well-being, and emotion regulation? Researchers will compare the AI chatbot to a control group receiving psychoeducational materials to see if the chatbot works to reduce stress and improve mental health. Participants will: 1. Be randomized to use the AI chatbot (recommended at least 2 sessions per week, 10-15 minutes per session) or read psychoeducational materials (2 times per week) for 4 weeks. 2. Complete online self-report questionnaires at baseline, post-intervention (Week 4), and 4-week follow-up.
Chinese MSM often experience unique social stress factors related to stigma, discrimination, and the pressures of concealing their sexual identity. These ongoing challenges are closely linked to elevated levels of anxiety, depressive symptoms, and general psychological distress. While traditional face-to-face therapy can be effective, common barriers such as waiting lists, costs, and privacy concerns frequently prevent individuals from seeking care. To help bridge this gap, the trial evaluates an accessible, private, and real-time AI tool that delivers safe, low-intensity mental health preventive support directly to a user's smartphone. The investigational tool is a web-based chatbot built on a open-source Chinese large language model and tailored using Retrieval-Augmented Generation technology. This design ensures that the chatbot's responses are empathetic, culturally sensitive, and tailored to MSM-related stress and mental health topics. During conversations, the chatbot guides users through brief, evidence-based mindfulness practices-such as grounding, cognitive defusion, self-compassion, and sleep support-designed for 10-to-20-minute sessions. The tool focuses strictly on stress management and coping skills; it is not intended to provide clinical diagnosis, formal psychotherapy, or emergency crisis intervention. The research follows a randomized controlled trial design spanning four weeks of intervention followed by a post-study assessment four weeks later. Adult participants with mild-to-moderate stress are recruited via major social media platforms, including Xiaohongshu, Weibo, and WeChat. After completing online consent and baseline surveys, participants are randomly assigned to either the intervention group or a control group. Those in the intervention group attend a brief online orientation and gain access to the AI chatbot, which they are recommended to use at least twice a week. Participants in the control group receive non-interactive educational readings on general stress management twice weekly. Researchers measure changes in stress levels, anxiety, depression, emotion regulation, rumination, and coping strategies across three key evaluation points.
Study Type
INTERVENTIONAL
Allocation
RANDOMIZED
Purpose
PREVENTION
Masking
NONE
Enrollment
76
An investigational, web-based conversational AI chatbot providing preventive, low-intensity psychological support for Chinese MSM experiencing mild-to-moderate stress. It has the following functions: (1) Built on a Chinese open-source LLM optimized through RAG to deliver culturally adapted, empathetic responses addressing minority stress, stigma, and identity concealment. (2) Delivers interactive, 10-20 min guided mindfulness-based stress reduction exercises (grounding, self-compassion, sleep-support, etc).
Participants receive standardized, text-based psychoeducational readings covering general stress, coping concepts, and mental health awareness. The delivery contains no interactive features, conversational elements, or guided mindfulness coaching.
The University of Hong Kong
Hong Kong, Hong Kong
Score of Perceived Stress
Chinese version of Perceived Stress Scale (PSS-14). Score range: 0-56. The higher the score, the greater the perceived stress. PSS-14 is highly sensitive to both acute and subtle changes in stress levels, making it effective for capturing variations in stress even within short periods of intervention. It can detect moderate fluctuations in perceived stress, even in individuals with low to moderate baseline stress levels.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Difficulties in Emotion Regulation
Chinese version of 16 item Difficulties in Emotion Regulation Scale Short Form (DERS-SF). Score range: 16-80. The higher the score, the lower the individual's level of emotional regulation. DERS-SF is highly sensitive to short-term fluctuations in emotion regulation.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Stress Coping
Chinese version of Brief Coping Orientation to Problems Experienced Inventory (Brief-COPE). Score range: 28-112. The higher score of the subscale demonstrates more likely the coping strategies were used by the respondents. Brief-COPE is highly sensitive to changes in coping strategies and can capture variations in stress responses, especially following interventions aimed at improving coping mechanisms.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Rumination
Chinese version of Ruminative Responses Scale (RRS). Score range:22-88. A higher score indicates a more severe tendency toward ruminative thinking. RRS is sensitive to both moderate and subtle changes in rumination. It can detect shifts in rumination levels, particularly in individuals with depression or stress-related disorders.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Distress Tolerance
Chinese version of distress tolerance scale (DTS). Score range: 15-75. A higher score indicates that an individual has a higher tolerance for emotional distress.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Depressive symptom
Chinese version of Patient Health Questionnaire-9 (PHQ-9). Score range:0-27. A higher score indicates more severe symptoms of depression. PHQ-9 is sensitive to both moderate and mild changes in depressive symptoms. It has been shown to detect reductions in depressive symptoms following short-term interventions, even those targeting stress or emotion regulation rather than directly treating depression.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Anxiety Symptom
Chinese version of Generalized Anxiety Disorder-7 (GAD-7). Score range: 0-21. A higher score indicates more severe symptoms of anxiety. GAD-7 is particularly sensitive to changes in anxiety, capturing both the intensity and frequency of anxiety symptoms. Even short-term interventions can result in significant reductions in anxiety.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Subjective Well-Being
Chinese version of WHO Well-Being Index (WHO-5). Score range:0-30. A higher score indicates a higher level of wellbeing. WHO-5 is sensitive to short-term fluctuations in subjective well-being, making it ideal for measuring the effects of interventions like mindfulness.
Time frame: From enrollment to the end of intervention at week 1, week 4, and week 8.
Score of Acceptability of Intervention
Chinese version of Acceptability of Intervention (AIM). Score range: 1-5. Higher scores indicate greater acceptability. 4-item Acceptability of Intervention Measure (AIM) will be used to assess acceptability. It is a 5-point Likert scale, with good readability, reliability, and validity. Higher scores indicate greater acceptability.
Time frame: From enrollment to the end of intervention at week 4.
Score of Intervention Appropriateness
Chinese version of Intervention Appropriateness Measure (IAM). Score range: 1-5. Higher scores indicate greater appropriateness. 4-item Intervention Appropriateness Measure (IAM) will be used to assess the appropriateness of the intervention. It is a 5-point Likert scale, with good readability, reliability, and validity. Higher scores indicate greater appropriateness.
Time frame: From enrollment to the end of intervention at week 4.
Score of AI Psychological Dialogue Quality
AI Psychological Dialogue Quality Assessment. Score range: 1-5. Higher scores indicate higher quality. AI Psychological Dialogue Quality Assessment is a 5-point Likert scale for evaluating AI psychological dialogue quality including normativity, professionalism, emotional understanding and empathy, consistency and coherence and potentially harmful information).
Time frame: From enrollment to the end of intervention at week 4.
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.