Beyond the Dialogue: Multi-chatbot Group Motivational Interviewing for Premenstrual Syndrome (PMS) Management

Research Background and Issues

• What problems or challenges did the authors identify?

  • Premenstrual Syndrome (PMS) causes women to experience a variety of physical, emotional, and behavioral symptoms before menstruation. However, due to societal factors such as stigmatization, many women lack peer support and tend to handle these issues alone.
  • Existing technological interventions (e.g., period tracking apps) focus more on symptom logging but rarely address how to provide social support or self-help resources.
  • Most existing mental health chatbot tools are designed for one-on-one interactions, and the potential of multi-chatbot systems in mental health interventions—particularly in creating social support scenarios—remains underexplored.

• Why is this issue important?

  • Women lacking peer support may experience feelings of loneliness, which could further exacerbate their symptoms and disrupt their daily lives.
  • Research on group therapy has shown that supportive social environments can foster a sense of belonging and improve mental health. Therefore, shifting to a group-based support approach in PMS management could yield significant benefits.

• Research Motivation and Related Work

  • The authors aim to leverage the advantages of group therapy (e.g., cognitive behavioral therapy and motivational interviewing) by introducing a multi-chatbot system to provide an integrated intervention approach for PMS management that includes both symptom tracking and social support.

Solution

• What methods or solutions did the authors propose?

  • They developed a multi-chatbot system that simulates a group counseling environment, featuring a "guide bot" and two "peer bots," implemented using a large language model (e.g., GPT-4).
  • This system employs motivational interviewing questions to help users explore their symptoms, understand coping strategies, and simulate a sense of social support.

• What is innovative about this solution?

  • This is the first application of multi-chatbot features in mental health interventions, transitioning from traditional one-on-one models to more complex group dynamic support systems.
  • The system incorporates mechanisms for social learning and linguistic style adaptation (e.g., linguistic style matching and semantic resonance) into user-machine interactions, a method not previously applied in technological solutions for PMS management.

• What are the implementation steps and key technologies used?

  1. System Design:
     • The "guide bot" leads users through predefined questions (e.g., symptom descriptions, coping strategies) using a motivational interviewing approach.
     • The "peer bots" play the roles of a "role model" (extroverted, optimistic) and an "empathizer" (introverted, hesitant), jointly providing coping strategies and emotional support.
  2. The chatbots' language generation is implemented using the GPT-4 model, with emotionally styled statements designed to interact with participants via the Slack platform.
  3. A two-month user study was conducted in Japan, comparing three experimental conditions: "no intervention," "one-on-one chatbot," and "multi-chatbot group interaction," to analyze user experiences and outcomes.
  4. Data collection included questionnaires, chat log analysis (e.g., character count, linguistic style matching, semantic analysis), and interview feedback.

Research Outcomes

• What specific outcomes were achieved?

  1. Participants in the multi-chatbot group exhibited higher engagement, including longer response lengths and more complex language content.
  2. Participants demonstrated significant convergence in linguistic style and cognitive expressions with the chatbots (especially the peer bots), indicating dynamic linguistic and cognitive adjustments.
  3. Participants learned new coping strategies by observing the peer bots and experienced a sense of belonging and emotional support.

• What advantages does it have compared to existing solutions?

  • Compared to single one-on-one chatbot models, the multi-chatbot system simulates a group social support environment, providing a richer cognitive and emotional experience.
  • The system integrates both symptom tracking and social support functions, addressing the limitations of existing PMS management tools that focus on a single functionality.

• What were the experimental or evaluation results?

  • In the multi-chatbot group, linguistic style matching (LSM) scores were significantly higher than in the one-on-one group, indicating greater interactional synchrony.
  • Interview results revealed that some participants felt a sense of belonging from the group and became more willing to share PMS-related experiences with friends in real life.
  • On the downside, some participants reported feelings of jealousy and helplessness due to social comparison with the chatbots or felt overwhelmed by the volume of information provided.

• Limitations and Future Directions

  • Limitations:
    • The study was limited to the cultural context of Japan, making it difficult to generalize findings to cultures with different social norms.
    • The study duration was relatively short (two months), preventing observation of long-term effects.
    • The motivational interviewing questions were predefined, limiting the system's flexibility to respond to specific user needs.
    • Personality dynamics modeling and cultural factors were not thoroughly explored.
  • Future Directions:
    • Extend the research to diverse cultural contexts and conduct longer-term experiments to verify the durability of the effects.
    • Explore high-flexibility and personalized interaction models, with deeper analysis of the dynamic role adaptation of chatbots.
    • Investigate ways to mitigate the negative psychological effects of social comparison while enhancing the perceived empathy of the chatbots.
    • Further develop personalized mental health intervention strategies in multi-chatbot environments based on linguistic and cognitive style modeling.

Through this study, the authors demonstrated the potential of multi-chatbot systems in the field of mental health support, emphasizing the need to balance user experience and ethical responsibility when designing multi-bot interaction systems.