TL;DR本文研究二元分类器中的置信度值使用,发现对于决策者来说确定何时信任预测很困难,提出了置信度值与决策者自身置信度一致时决策最优且易于发现,提出了多重校准并验证其有效性在 AI 辅助决策场景中。
Abstract
Whenever a binary classifier is used to provide decision support, it
typically provides both a label prediction and a confidence value. Then, the
decision maker is supposed to use the confidence value to calibrat