biostudies-literatureCurtis SDFederal Government | DHAC | National Health and Medical Research CouncilOncology Core CAHHS | NIHFederal Government | DHAC | National Health and Medical Research Council (NHMRC)Benjamin Baker EndowmentHHS | NIH (NIH)e2424203122https://www.ebi.ac.uk/biostudies/studies/S-EPMC12402999biostudies-literatureUnknown122(34)AI is now a cornerstone of modern dataset analysis. In many real world applications, practitioners are concerned with controlling specific kinds of errors, rather than minimizing the overall number of errors. For example, biomedical screening assays may primarily be concerned with mitigating the number of false positives rather than false negatives. Quantifying uncertainty in AI-based predictions, and in particular those controlling specific kinds of errors, remains theoretically and practically challenging. We develop a strategy called multidimensional informed generalized hypothesis testing (MIGHT) which we prove accurately quantifies uncertainty and confidence given sufficient data, and concomitantly controls for particular error types. Our key insight was that it is possible to integraProceedings of the National Academy of Sciences of the United States of AmericaMinimizing and quantifying uncertainty in AI-informed decisions: Applications in medicine.PMC1240299906973R21NS11301680049589APP119497080057309U01CA2306915P50CA062924-22RA37CA230400Kinzler KWDouville CIrajizad ETie JLennon AMGoggins MDobbyn LNguyen TVWang TLPopoli MPanda SLi ANehme NWang YFader APapadopoulos NHo-Pham LTBettegowda CVogelstein JTShih IMWolfgang CLTran TSOgihara ITran BNHXu HBai YPtak JGilbert LSilliman NHruban RHCurtis SDVogelstein BGibbs PO'Reilly EfalseMinimizing and quantifying uncertainty in AI-informed decisions: Applications in medicine.AI is now a cornerstone of modern dataset analysis. In many real world applications, practitioners are concerned with controlling specific kinds of errors, rather than minimizing the overall number of errors. For example, biomedical screening assays may primarily be concerned with mitigating the number of false positives rather than false negatives. Quantifying uncertainty in AI-based predictions, and in particular those controlling specific kinds of errors, remains theoretically and practically challenging. We develop a strategy called multidimensional informed generalized hypothesis testing (MIGHT) which we prove accurately quantifies uncertainty and confidence given sufficient data, and concomitantly controls for particular error types. Our key insight was that it is possible to integra2025-01-01T00:00:00Z2025 Aug2026-07-16T07:50:39.092Z2026-07-09T10:45:52.171ZS-EPMC124029994083340810.1073/pnas.2424203122