A Perspective on Explanation Types and Content from Model and Algorithm Selection

Explainable AI (XAI) research has made substantial progress in developing methods for generating explanations, but less attention has been paid to understanding what explanatory information users need to make informed decisions. This position paper argues that explanation content and perceived interpretability should be understood as context- and user-dependent resources for decision support. Building on work on interactive model and algorithm selection, it proposes comparative human interaction as a means of eliciting, assessing, and refining users’ explanation preferences.