Constellations of Memory: An Exploratory Design Probe for Human–AI Collaborative Autobiographical Sense-Making

Autobiographical storytelling is a reflective practice through which people make sense of personal experiences, identity, and change. Recent advances in generative AI introduce new possibilities for supporting such practices collaboratively, while also raising ethical and experiential questions around agency, authorship, and the sharing of sensitive personal data. This paper presents Constellations of Memory, an exploratory VR prototype that investigates how human–AI collaboration might support autobiographical sense-making. The prototype uses a small set of narrative "constellation" nodes (Beginning → Setback → Breakthrough → Conclusion) to scaffold reflection, while the AI alternates between a facilitative role (guiding reflection through prompts and follow-up questions) and a co-authoring role (synthesising user-provided fragments into a coherent autobiographical narrative). Rather than delivering a finished system, we employ a Wizard-of-Oz implementation as a design probe to explore interaction possibilities and design tensions in AI-mediated autobiographical storytelling. Based on a researcher-led walkthrough, we examine how narrative scaffolding, user agency, and shifting AI roles shape sensitive autobiographical storytelling. We surface key ethical and experiential considerations, including trust, bias, and shared authorship, and position this work as an early exploration intended to inform future participatory and empirical research on responsible human–AI collaboration.