How to create “humble” AI

An MIT-led team is designing artificial intelligence systems for medical diagnosis that are more collaborative and forthcoming about uncertainty. Led by Leo Anthony Celi and Sebastián Andrés Cajas Ordoñez, the researchers developed a framework that includes computational modules like the Epistemic Virtue Score to evaluate AI certainty. The goal is to prevent 'arrogant' AI from steering doctors toward incorrect decisions, especially in high-stakes environments like the ICU where automation bias is a known risk.