New Member Double-Header

Building the Science of AI Health Coaching for Chronic Disease Management (Nirali Shah, PhD, PT)
This talk will focus on understanding how AI health coaching systems should be designed, evaluated, and personalized to support individuals living with chronic diseases. I will present findings from studies examining what people want from AI health coaches, including preferences for evidence-based disease support, collaborative behavioral coaching, personalization, and emotional support. I will then describe our development and evaluation of Sprout, an LLM-based health coach for people with systemic sclerosis, and what we have learned from examining how people interact with the system over time. Specifically, I will discuss how differences in both the amount and purpose of interactions challenge conventional approaches to conceptualizing and measuring engagement with digital interventions. I will also describe ongoing work to develop more evidence-grounded and contextually personalized AI health coaching through retrieval-augmented generation and longitudinal contextualization. Together, this work motivates a broader set of questions for the field: What constitutes meaningful engagement with conversational AI? Which therapeutic processes contribute to benefit? How should AI health coaches adapt to different individuals and over time? And how should we evaluate both their potential benefits and harms? I will conclude by discussing opportunities for interdisciplinary collaboration to address these emerging behavioral, methodological, and technical challenges.