Auditing Institutional Heterogeneity for Generative AI in Patient Education: A Large-Scale Study of 102 US Transplant Handbooks

Yubo Li, Rema Padman, Ramayya Krishnan (2026) — Carnegie Mellon University. arXiv preprint (cs.CY).

📄 Full text (arXiv)

Li, Padman and Krishnan audit 102 US transplant-center patient handbooks that serve as grounding corpora for generative AI patient-education assistants. They show large institutional heterogeneity in the underlying education materials, undermining the premise that grounding a genAI assistant in local content yields consistent guidance: patients at different institutions can receive materially different AI-mediated answers to the same question. The study extends grounding-quality concerns familiar from retrieval-augmented-tutoring-algorithm-kite into health education, and connects to AI-driven medical training work such as medeasy-ai-standardized-patients, adaptive-virtual-patient-psychotherapy-training and medgame-llm-medical-education-gamification. The equity implication (institution-dependent quality of AI-mediated education) parallels equity-in-ai-education.

Related Pages

• retrieval-augmented-tutoring-algorithm-kite

• medeasy-ai-standardized-patients

• adaptive-virtual-patient-psychotherapy-training

• medgame-llm-medical-education-gamification

• equity-in-ai-education

Citation

APA: Yubo Li, Rema Padman, Ramayya Krishnan (2026). Auditing Institutional Heterogeneity for Generative AI in Patient Education: A Large-Scale Study of 102 US Transplant Handbooks. arXiv:2607.22606.