Are Objective Explanatory Evaluation metrics Trustworthy? An Adversarial Analysis

TL;DR

This paper introduces SHAPE, a novel adversarial explanation method for XAI that challenges existing evaluation metrics, revealing their limitations and the need for human-involved assessments.

Contribution

The paper proposes SHAPE, a new explanation technique based on causal notions, and demonstrates its ability to deceive current objective evaluation metrics in XAI.

Findings

SHAPE outperforms GradCAM and GradCAM++ in tests.

SHAPE is comparable to RISE in effectiveness.

Existing metrics can be fooled by adversarial explanations.

Abstract

Explainable AI (XAI) has revolutionized the field of deep learning by empowering users to have more trust in neural network models. The field of XAI allows users to probe the inner workings of these algorithms to elucidate their decision-making processes. The rise in popularity of XAI has led to the advent of different strategies to produce explanations, all of which only occasionally agree. Thus several objective evaluation metrics have been devised to decide which of these modules give the best explanation for specific scenarios. The goal of the paper is twofold: (i) we employ the notions of necessity and sufficiency from causal literature to come up with a novel explanatory technique called SHifted Adversaries using Pixel Elimination(SHAPE) which satisfies all the theoretical and mathematical criteria of being a valid explanation, (ii) we show that SHAPE is, infact, an adversarial explanation that fools causal metrics that are employed to measure the robustness and reliability of popular importance based visual XAI methods. Our analysis shows that SHAPE outperforms popular explanatory techniques like GradCAM and GradCAM++ in these tests and is comparable to RISE, raising questions about the sanity of these metrics and the need for human involvement for an overall better evaluation.