2026 Encoding Health Equity Summit - Panel Discussion: Translating Ethical AI Principles into Practice-Panel Discussion: Translating Ethical AI Principles into Practice

Video Transcription

Video Summary

The panel focused on the rapid rise of AI in healthcare and the need for new approaches to equity, governance, and monitoring. Tina Hernandez-Bressard opened by stressing that today’s models are deeply embedded in healthcare systems and evolve too quickly for static evaluation.<br /><br />Leo Celi proposed the LTARC framework: local, task-specific, agile, reflective, and community-partnered. He argued that evaluation must be continuous, context-sensitive, and genuinely shared with communities, not tokenistic.<br /><br />Judy Gichoya highlighted how algorithms can infer race and other sensitive traits from medical images, showing both the power and danger of AI. She emphasized that race is socially visible but often not actionable, and that AI should be judged by what it changes in real-world care.<br /><br />The panel agreed that fairness is not just about debiasing models, but also about access, deployment, workflow integration, and who benefits. They warned that AI can worsen disparities if introduced into inequitable systems without strong oversight.<br /><br />Another major theme was implementation: ambient scribes, agentic systems, and generative AI may improve efficiency, but only if they truly help clinicians and patients. The discussion ended with a call for community education, political will, and redesigning AI to support better, more equitable healthcare outcomes.

Keywords

AI healthcare

equity

governance

monitoring

fairness

LTARC framework

medical images

algorithmic bias

community-partnered evaluation