American Statistical Association Risk Analysis Section Webinar
Navigating The Complex Landscape of Health AI Trust:
The Importance of Statistical Science
11:00 AM – 12:30 U.S. Eastern Time, June 18, 2025
Speaker: Dr. Constantin F. Aliferis (University of Minnesota)
Abstract: Decades of Nobel- and Turing- award winning progress in the science and
technology of AI comprising foundational discoveries, combined with recent developments in
generative AI, have fueled intense interest in health AI (HAI). At this point in time, HAI has a
wide biomedical literature basis supporting feasibility but has not found broadly its way to the
bedside. A critical reason is the current prominent gap of assurances (and resulting trust),
addressing safety, effectiveness, cost-effectiveness, and ethics.
In this webinar we will discuss the critical importance (and common pitfalls) of establishing
theoretical and empirical properties of HAI as the foundational precondition of assurances and
the importance of statistical science. We will compare theory with empirical studies, contrast
reductionist vs system-level properties, and highlight the importance of alignment of properties
to specific clinical problems. We will also discuss how AI method and AI model “labels” can play
a pivotal role for organizing and disseminating critical operating information for HAI. Finally we
will address how health practitioners can navigate the complex space of HAI properties.
Concepts to be covered:
- AI trust, assurances, adoption
- Characteristics of AI in health sciences vs healthcare vs other fields
- Health AI constituents
- Theory plus empirical need for managing AI risks
- AI labels
- Lifecycles of health AI methods and models
- AI embedding and health systems thinking
- The role of industry in AI and the importance of academia; Innovation vs regulation dilemma.
- Short review of recent ASA recommendations. Is statistics enough to ensure health AI
trust?
Registration: