#729: From Practice Partner to Co-Learning: Designing Role-Fluid AI Partnership for Skills Development

Building on research in human–AI collaboration and appropriate reliance, we propose a conceptual design model of a role-fluid, AI-driven practice–learning partner to support novices’ skill development in dynamic, context-dependent tasks. The model integrates LLM-generated personas with agent-based modeling. Using a multiple-case, mixed-methods approach, we examined its implementation and feasibility across three domains: responsive teaching, transitional care, and multimodal qualitative data coding. The findings inform the design of novice–AI collaborations that promote learning and sustained skill development, rather than merely cognitive offloading or short-term task completion.