The Accuracy Paradox: Why a 95% Accurate AI Can Still Be Wrong 84% of the Time
中文摘要
即使AI诊断准确率达95%，因罕见病样本稀少，其阳性预测值可能仅为16%，导致84%的误报率，凸显了高准确率背后的统计陷阱。
English Summary
Despite 95% overall accuracy, AI can suffer an 84% false discovery rate in rare disease detection due to base rate fallacies, highlighting critical statistical limitations in model reliability.
原文节选
We’ve all seen the headlines. A new AI model drops, boasting a spectacular 95% overall accuracy at detecting a rare disease, flagging… Continue reading on Medium »