When Machine Learning Gets Personal: Evaluating Prediction and Explanation

ArXi:2502.02786v2 Announce Type: replace In high-stakes domains like healthcare, users often expect that sharing personal information with machine learning systems will yield tangible benefits, such as accurate diagnoses and clearer explanations of contributing factors. However, the validity of this assumption remains largely unexplored. We propose a unified framework to quantify how personalizing a model influences both prediction and explanation. We show that its impacts on prediction and explanation can diverge: a model may become or less explainable even when prediction is unchanged.