Normalized AOPC: Fixing Misleading Faithfulness Metrics for Feature Attribution Explainability

TL;DR

This paper introduces Normalized AOPC, a new metric for feature attribution faithfulness in neural networks, addressing limitations of the traditional AOPC by enabling reliable cross-model comparisons and interpretation.

Contribution

It proposes a normalization method for AOPC, improving the reliability and interpretability of faithfulness evaluations across different models.

Findings

Normalization radically changes AOPC results

Previous conclusions based on AOPC may be unreliable

Normalized AOPC provides a more robust evaluation framework

Abstract

Deep neural network predictions are notoriously difficult to interpret. Feature attribution methods aim to explain these predictions by identifying the contribution of each input feature. Faithfulness, often evaluated using the area over the perturbation curve (AOPC), reflects feature attributions' accuracy in describing the internal mechanisms of deep neural networks. However, many studies rely on AOPC to compare faithfulness across different models, which we show can lead to false conclusions about models' faithfulness. Specifically, we find that AOPC is sensitive to variations in the model, resulting in unreliable cross-model comparisons. Moreover, AOPC scores are difficult to interpret in isolation without knowing the model-specific lower and upper limits. To address these issues, we propose a normalization approach, Normalized AOPC (NAOPC), enabling consistent cross-model evaluations and more meaningful interpretation of individual scores. Our experiments demonstrate that this normalization can radically change AOPC results, questioning the conclusions of earlier studies and offering a more robust framework for assessing feature attribution faithfulness. Our code is available at https://github.com/JoakimEdin/naopc.