Understanding Human-AI Collaboration in Music Therapy Through Co-Design with Therapists

Title of the Paper

Understanding Human-AI Collaboration in Music Therapy Through Co-Design with Therapists

Paper Information

• Research Areas: Human-AI Collaboration, Music Therapy, Design Research
• Keywords: Music Therapy, AI Collaboration, Human-Computer Interaction, Co-Design, Emotional Therapy, Multimodal Generation, Personalized Treatment, Music Generation Technology

Research Background and Issues

• Identified Problems or Challenges:

  • Music therapy has been proven to positively impact mental health, but due to its complexity, the therapy often relies heavily on the professional expertise and technical support of music therapists.
  • Although music-related AI technologies (e.g., music generation, style transfer) have advanced rapidly, research on their application in music therapy remains limited.
  • Integrating AI technologies may require therapists to acquire new skills, potentially disrupting existing workflows.

• Significance:

  • Innovations in music AI technologies, from algorithms to application layers, can significantly enhance the efficiency, diversity, and personalization of music therapy.
  • Understanding therapists' needs and concerns regarding music AI technologies will help design more effective human-AI collaborative systems to support complex therapeutic processes.

• Research Motivation and Related Work:

  • Current research on music AI primarily focuses on algorithmic development, with little consideration of real-world application contexts and therapists' needs.
  • This study aims to address the following questions:
    1. What are therapists' perceptions of current music AI technologies?
    2. How do therapists envision music AI assisting in the music therapy process?
    3. What are therapists' main concerns about the application of music AI?

Solution

• Proposed Solution:

  • Design and implement a mixed-methods approach, including semi-structured interviews and co-design workshops, to explore how music AI technologies can be integrated into music therapy.
  • Develop emotion-focused therapeutic solutions by involving therapists to create an emotion-centered therapy framework combined with music AI.

• Innovative Aspects of the Solution:

  • Propose a typical workflow for music therapy, integrating potential application scenarios for music AI technologies to form a comprehensive therapeutic solution.
  • Co-create design recommendations based on real clinical scenarios with therapists, laying the foundation for music therapy applications in human-AI collaboration.

• Implementation Steps and Key Technologies:

  • Implementation Methods:

    1. Semi-structured interviews: Understand therapists' workflows, common challenges, and views on music AI.
    2. Co-design workshops: Provide representative AI music generation technology materials (e.g., emotion-driven music generation models) for therapists to participate in designing therapeutic solutions.
    3. Data analysis: Use thematic coding to analyze interview and workshop results, summarizing key insights.

  • Music AI Technologies Used:

    • General Melody Generation (Ge-Gen): Random melody generation.
    • Emotion-based Music Generation (Emo-Gen): Generating music aligned with emotions.
    • Melody Harmonization (Melo-Har): Harmony generation.
    • Music Genre Transfer (Genre-Tran): Music style transfer.
    • Tone Transfer (Tone-Tran): Transforming tones into specific instrumental effects.

Research Outcomes

• Specific Outcomes:

  1. Proposed a typical workflow commonly used by music therapists (information gathering, emotional issue resolution, therapy consolidation).
  2. Summarized the application of music AI technologies during the emotional issue resolution phase, including emotion perception, emotion acceptance, emotion coping, and emotion transformation.
  3. Explored the potential advantages of music AI in supporting therapists, such as improving therapeutic effectiveness, enriching therapy content, personalizing treatment, and expanding client reach.

• Advantages Compared to Existing Solutions:

  • This study allocates music AI technologies to specific therapy stages, providing a theoretical framework for technology design and application.
  • It not only focuses on the technical capabilities of AI in therapy but also delves into therapists' practical needs and system usability.

• Experimental or Evaluation Results:

  • Therapists believe music AI tools can significantly enhance therapy efficiency.
  • However, current AI-generated music still struggles to match subtle emotional expressions, and technical performance needs further improvement.
  • Participants emphasized high requirements for the safety and user-friendliness of AI tools.

• Limitations and Future Directions:

  • Limitations:
    1. Small sample size, with participants concentrated in China and North America.
    2. Gender imbalance (majority of participants were female therapists).
    3. Results may be influenced by cultural and regional contexts, lacking broad representativeness.
  • Future Directions:
    1. Develop prototypes of music AI interfaces based on design recommendations and conduct user testing.
    2. Enhance the emotional recognition capabilities of music AI technologies, improving the quality and adaptability of generated music.
    3. Expand research to include more diverse cultural and professional backgrounds.