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

Authors: Marco Zullich, Emily Schiller, Ivan Gentile, David Dembinsky, Adriano Lucieri, Kilian Göller, Steffen Seitz

Venue: Presented at the IJCAI-ECAI 2026 Workshop on Explainable Artificial Intelligence (XAI)

Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)

Cite as: arXiv:2610.12493 [cs.AI] (v1)

Submitted: 24 September 2026

Abstract

Explainable Artificial Intelligence (XAI) tools have often been presented as methods to increase trust in machine learning models by making aspects of their algorithms more human-understandable. However, evidence accumulated over the last decade points to an overall failure of this objective, instead promoting overreliance on incorrect predictions, which can subsequently lead to algorithmic aversion. Rather, XAI appears to function primarily as a debugging tool, useful for obtaining anecdotal insights into, for example, features learned by the model.

One of the main issues underlying this limitation is that XAI methods often oversimplify the inner workings of the model, resulting in misalignment between the explanation and the model itself—a behavior referred to as unfaithfulness or infidelity. Faithfulness can be interpreted as the degree of approximation between an explanation and the model. Computing faithfulness in a rigorous manner would thus make it possible to quantify one of the various elements of trustworthiness of an explanation. However, existing metrics present significant limitations, making this quantification either infeasible or computationally expensive.

In this position paper, we outline five major shortcomings of current faithfulness metrics and propose four possible avenues for research to bolster their reliability.

Position and Contributions

This paper argues that the evaluation of faithfulness for feature-importance explanations remains structurally misaligned with the very properties it purports to measure. The authors catalog five principal deficiencies in contemporary faithfulness metrics and chart four research directions intended to improve their reliability. The work is positioned within the growing 2026 discourse on reducing overreliance on XAI outputs and reframing explanation quality assessment as a rigorous, quantifiable enterprise rather than a heuristic one.

Metadata

• arXiv ID: arXiv:2610.12493 [cs.AI]
• Version: v1
• Workshop: IJCAI-ECAI 2026 Workshop on Explainable Artificial Intelligence (XAI)

For the full text and any accompanying supplementary materials, consult the official arXiv listing.

via ArXiv AI