Shared Stories, Shared Bonds: People with Dementia Exploring Generative AI Together

Paper Title

Shared Stories, Shared Bonds: People with Dementia Exploring Generative AI Together

Publication Info

• Topic area: Generative AI applications in dementia care for fostering social connectedness.
• Keywords: Dementia, social connectedness, Generative AI, participatory design, agency, personhood, shared realities, multimodal interaction.

Background and Problem

• Problem / challenge: People with dementia often experience social isolation due to cognitive and communication challenges, limiting their ability to engage in meaningful social interactions. Existing GenAI applications for dementia care focus primarily on reminiscence but lack exploration of active engagement and co-creation by people with dementia.
• Significance: Addressing social isolation in dementia care is critical for improving mental health, well-being, and quality of life. Exploring GenAI as a tool for fostering social connectedness could offer new opportunities for meaningful interactions.
• Motivation and related work: Prior research has shown that technology-based interventions, including GenAI, can enhance social connectedness through reminiscence and storytelling. However, little is known about how people with dementia actively experience and interact with GenAI in group settings, or how such engagements foster social bonds.

Solution

• Proposed approach: Interactive workshops using three GenAI models (Copilot, Midjourney, Suno) to explore how people with dementia engage with generative outputs and co-create prompts in group settings.
• Novelty:
  1. Demonstrates that people with dementia can actively shape GenAI prompts and outputs, highlighting their capacity for co-creation.
  2. Shows how group-based GenAI interactions foster shared realities and social connectedness through negotiation and meaning-making.
  3. Provides design guidelines for accessible, privacy-preserving, and co-creation-oriented GenAI interfaces tailored to dementia care contexts.
• Procedure and key techniques:
  • Conducted six workshops with 17 participants with mild to moderate dementia, using visual cards as tangible prompts for generating text, images, and music via GenAI models.
  • Facilitated iterative prompting and discussion cycles to co-create outputs and interpret their meaning.
  • Analyzed verbal and non-verbal behaviors using thematic analysis to identify patterns of engagement and social connectedness.

Results

• Concrete findings:
  • Participants actively co-constructed prompts and evaluated GenAI outputs, seeking verisimilitude or engaging in imaginative storytelling.
  • GenAI-supported activities fostered social connectedness through negotiation of shared realities and reciprocal engagement.
  • Over time, participants developed partial understanding and critical awareness of GenAI capabilities.
• Advantage over baselines: GenAI’s flexibility and responsiveness enabled iterative co-creation and negotiation, surpassing static, non-AI-generated content in fostering engagement and shared realities.
• Experiments / evaluation:
  • Workshops involved text (Copilot), image (Midjourney), and music (Suno) generation, supported by visual cards.
  • Data collection included video and audio recordings, thematic analysis of transcripts, and exit interviews with caregivers.
• Limitations and future work:
  • Findings may not generalize to people with advanced dementia or diverse cultural contexts.
  • Further research is needed to disentangle the contributions of different GenAI modalities and explore applications in naturalistic settings.

Summary

This study demonstrates that Generative AI can foster social connectedness among people with dementia by enabling collaborative prompting and co-creation of outputs. Through interactive workshops, participants actively shaped GenAI outputs, negotiated shared realities, and developed critical awareness of the technology. The findings highlight GenAI’s potential to support agency, personhood, and relational engagement in dementia care. Design guidelines emphasize scaffolding, perceptibility, fail-safes, and user-led personalization to ensure accessibility and ethical use. Future research should explore broader applications of GenAI in everyday care practices and investigate its impact across diverse populations and contexts.