Exploring the Design of Generative AI in Supporting Music-based Reminiscence for Older Adults

Title of the Paper

Exploring Generative AI Designs to Support Music-Based Reminiscence Activities for Older Adults

Paper Information

• Research Domain: Human-Computer Interaction, Generative AI for Supporting Older Adults' Mental Health
• Keywords: Human-AI Interaction, Generative AI, Reminiscence Therapy, Music-Based Reminiscence, Older Adults

Research Background and Issues

• Problems and Challenges:

  1. Older adults' mental health may benefit from reminiscence therapy, but memory decline and language limitations can hinder their ability to recall.
  2. Existing studies on reminiscence therapy often focus on traditional media (e.g., photos or music) and rarely explore how generative AI can enhance the reminiscence experience for older adults.
  3. Older adults' acceptance of technology is complex, involving privacy concerns, usability challenges, and emotional factors.

• Significance: Music is a deeply emotional medium that effectively triggers memories. Leveraging generative AI to produce images, questions, and other content can help older adults recall past details more fully, enhancing emotional well-being.

• Motivation and Objectives: This study aims to address the following questions:

  • RQ1: What are older adults' perceptions of the application of generative AI in music-based reminiscence?
  • RQ2: What factors should be considered when designing generative AI to support music-based reminiscence?

Solution

• Proposed Solution: The study adopts a user-centered design approach, including interviews with social workers, iterative prototyping (video prototype and digital prototype), and two rounds of design workshops to explore how generative AI can guide and support music-based reminiscence.

• Innovations:

  1. Conducting the first dedicated study on the potential application of generative AI in music-based reminiscence therapy.
  2. Extracting user perspectives and design considerations for supporting generative AI systems.

• Implementation Steps and Key Technologies:

  1. Collaborating with social workers to identify older adults' habits, technological preferences, and usage scenarios in music activities.
  2. Creating a video prototype (AI-DJ) to demonstrate how generative AI generates questions and images to support individual or group reminiscence.
  3. Iteratively designing a digital prototype (MusicJourney) using ChatGPT and Bing Chat to generate dialogues and images, allowing older adults to experience AI-supported reminiscence.
  4. Analyzing interview data using open coding and reflective thematic analysis to extract design insights.

Research Outcomes

• Key Findings:

  1. Older adults found that generative AI-generated dialogues and images helped them recall details, but they preferred personalized interactions over group interactions.
  2. Compared to existing reminiscence tools, AI-guided methods increased the enjoyment of reminiscence but required AI-generated content to be precise, personalized, and emotionally resonant.
  3. Older adults expressed strong concerns about privacy and data control, especially when sharing personal memories with AI systems.

• Advantages:

  1. AI's interactive features can provide older adults with personalized music reminiscence experiences, avoiding the potentially limited triggers of traditional methods.
  2. AI-generated dialogues and visual content can help older adults explore deeper memories and evoke positive emotions.

• Experimental or Evaluation Results:

  • Results from the video prototype and workshops indicated that participants preferred using AI for individualized music reminiscence activities.
  • Testing of the digital prototype (MusicJourney) demonstrated the preliminary feasibility of generative AI for question and image generation, effectively supporting deeper reminiscence triggered by music.

• Limitations and Future Directions:

  1. The study was geographically limited to Hong Kong; future research should explore the impact of cultural contexts on older adults' technology acceptance and experiences.
  2. Some older adult participants had relatively high digital literacy, which may affect the generalizability of the findings.
  3. The current study focused on individual reminiscence; future research should examine the application of generative AI in group reminiscence scenarios.
  4. Privacy design needs further refinement to balance enhanced privacy protection with personalized experiences.

This study provides significant theoretical and practical support for the future design of generative AI systems to assist older adults in reminiscence activities.