Revisiting the robustness of post-hoc interpretability methods

TL;DR

This paper introduces a new approach and metrics for fine-grained robustness assessment of post-hoc interpretability methods in AI, revealing their robustness is linked to coarse-grained performance.

Contribution

It proposes a novel approach and two metrics for detailed robustness evaluation of interpretability methods, addressing limitations of existing coarse-grained assessments.

Findings

Robustness correlates with coarse-grained performance.

New metrics enable fine-grained robustness analysis.

Interpretability methods vary significantly at the sample level.

Abstract

Post-hoc interpretability methods play a critical role in explainable artificial intelligence (XAI), as they pinpoint portions of data that a trained deep learning model deemed important to make a decision. However, different post-hoc interpretability methods often provide different results, casting doubts on their accuracy. For this reason, several evaluation strategies have been proposed to understand the accuracy of post-hoc interpretability. Many of these evaluation strategies provide a coarse-grained assessment -- i.e., they evaluate how the performance of the model degrades on average by corrupting different data points across multiple samples. While these strategies are effective in selecting the post-hoc interpretability method that is most reliable on average, they fail to provide a sample-level, also referred to as fine-grained, assessment. In other words, they do not measure the robustness of post-hoc interpretability methods. We propose an approach and two new metrics to provide a fine-grained assessment of post-hoc interpretability methods. We show that the robustness is generally linked to its coarse-grained performance.