Boundary-Aware Uncertainty for Feature Attribution Explainers

TL;DR

This paper introduces GPEC, a framework that quantifies uncertainty in feature attribution explanations by combining boundary-aware and approximation uncertainties, improving trustworthiness of explanations for complex black-box models.

Contribution

The paper proposes a novel geodesic-based kernel within GPEC to accurately measure uncertainty in explanations, adaptable to any classifier and attribution method.

Findings

GPEC's uncertainty estimates better identify unreliable explanations.

The framework is effective on tabular and image datasets.

The kernel captures decision boundary complexity theoretically and empirically.

Abstract

Post-hoc explanation methods have become a critical tool for understanding black-box classifiers in high-stakes applications. However, high-performing classifiers are often highly nonlinear and can exhibit complex behavior around the decision boundary, leading to brittle or misleading local explanations. Therefore there is an impending need to quantify the uncertainty of such explanation methods in order to understand when explanations are trustworthy. In this work we propose the Gaussian Process Explanation UnCertainty (GPEC) framework, which generates a unified uncertainty estimate combining decision boundary-aware uncertainty with explanation function approximation uncertainty. We introduce a novel geodesic-based kernel, which captures the complexity of the target black-box decision boundary. We show theoretically that the proposed kernel similarity increases with decision boundary complexity. The proposed framework is highly flexible; it can be used with any black-box classifier and feature attribution method. Empirical results on multiple tabular and image datasets show that the GPEC uncertainty estimate improves understanding of explanations as compared to existing methods.