Faithfulness Without Anchors: A Position Paper on the Structural Misalignments of Explainable AI Faithfulness Metrics for Feature Importance

A position paper argues that Explainable AI (XAI) faithfulness metrics for feature importance suffer from structural misalignments, often oversimplifying model behavior and leading to unfaithful explanations. The paper suggests that faithfulness should be measured as the degree of approximation between an explanation and the model to better quantify trustworthiness.