This article is adapted from Voices in the Code: A Story About People, Their Values, and the Algorithm They Made, out Sept. 8 from Russell Sage Foundation Press. In May 2021, I got a call I never expected. I was working on a book about A.I. ethics, focused on the algorithm that gives out kidneys to transplant patients in the United States. Darren Stewart--a data scientist from UNOS, the nonprofit that runs the kidney allocation process--was calling to get my take: How many decimal places should they include when calculating each patient's allocation score? The score is an incredibly important number, given it determines which patient will get first chance at each donated organ.

Despite deep learning (DL) success in classification problems, DL classifiers do not provide a sound mechanism to decide when to refrain from predicting. Recent works tried to control the overall prediction risk with classification with rejection options. However, existing works overlook the different significance of different classes. We introduce Set-classifier with Class-specific RIsk Bounds (SCRIB) to tackle this problem, assigning multiple labels to each example. Given the output of a black-box model on the validation set, SCRIB constructs a set-classifier that controls the class-specific prediction risks with a theoretical guarantee. The key idea is to reject when the set classifier returns more than one label. We validated SCRIB on several medical applications, including sleep staging on electroencephalogram (EEG) data, X-ray COVID image classification, and atrial fibrillation detection based on electrocardiogram (ECG) data. SCRIB obtained desirable class-specific risks, which are 35\%-88\% closer to the target risks than baseline methods.