PhD Thesis Defense: Dylan Moore

Optional ZOOM LINK

"Augmenting Sensemaking in Dementia Care and Education through AI-enabled 'Narrative Stewardship'"

Abstract

This thesis asks how AI systems might act as stewards of narrative knowledge. I develop this concept by investigating use cases where people rely on stories about experiences, relationships, and meaning to support sensemaking. I specifically focus on sensemaking scenarios where such context is both critical as well as tacit, distributed across communities, or otherwise hard for an AI system to query. Across three case studies, I show how AI can elicit, apply, and circulate such stories to support outcomes related to young adult learning, person-centered care, and family involvement in care.

In my first case study, I examined the problem of supporting student learning via a mix of AI and peer-to-peer support in extracurricular contexts. Through a multi-year study, I developed Apricot Stone City, a narrative-based learnersourcing platform. A mixed-method evaluation demonstrated significant knowledge gains and highlights how this approach might enable scalable, peer-driven learning through AI-enabled sensemaking.

In my second case study, I examined the problem of supporting person-centered caregiving practices for people living with dementia in long-term care facilities. Through a multi-year study, I developed CareInsights, an AI-enabled documentation infrastructure that facilitates and records salient examples of effective person-centered therapy. CareInsights then translates these into concise, actionable guidance to caregivers that can be embedded in existing workflows and tools. Extending this infrastructure to families, I next developed and evaluated Family In The Loop, which supports narrative-form information sharing between caregivers and family members of care recipients. Both projects involved sustained fieldwork and iterative design with frontline caregivers across seven long-term care facilities, and mixed-method evaluations. My findings demonstrate how this AI-enabled approach to information capture, practical guidance, and communication can support the sensemaking required for effective person-centered care.

Synthesizing across these case studies, I contribute the novel concept of narrative stewardship as a design orientation for the field of human–AI interaction. I define this concept with five properties—attentiveness, translation, generativity, accountability, and dignity—that encompass key objectives and considerations for AI systems that mediate narrative knowledge across communities that produce and depend on it.

Thesis Committee

• Elizabeth Murnane (Chair)
• Petra Bonfert-Taylor
• Lorie Loeb
• James Landay (Stanford)

Contact

For more information, contact Thayer Registrar at thayer.registrar@dartmouth.edu.