What happened
Computer scientists built an AI system that approaches diagnosis the way experienced doctors do—forming hypotheses, testing them against evidence, and backtracking when something doesn't fit—rather than just pattern-matching symptoms to diagnoses. In experiments, it was more accurate and easier to explain its reasoning than earlier AI approaches to clinical diagnosis.
Why it matters
This shows AI can mimic the iterative, self-correcting reasoning that makes human experts good at hard problems, which matters if AI is going to handle domains where pattern-matching alone fails and you need to know why the system reached its conclusion.