Making AI models more trustworthy for high-stakes settings

Researchers made a technique that improves the trustworthiness of machine-learning models, which could help improve the accuracy and reliability of AI predictions for high-stakes settings such health care.