Towards Benchmarking the Utility of Explanations for Model Debugging

TL;DR

This paper advocates for developing a benchmark to evaluate the usefulness of post-hoc explanation methods in debugging text classifiers, emphasizing the importance of both effectiveness and efficiency in explanations.

Contribution

It introduces the concept of a benchmark for evaluating explanation utility in model debugging and outlines desirable properties for such a benchmark.

Findings

Identifies key properties for an explanation utility benchmark

Highlights the importance of evaluating both effectiveness and efficiency

Proposes a first step towards standardized evaluation of explanation methods

Abstract

Post-hoc explanation methods are an important class of approaches that help understand the rationale underlying a trained model's decision. But how useful are they for an end-user towards accomplishing a given task? In this vision paper, we argue the need for a benchmark to facilitate evaluations of the utility of post-hoc explanation methods. As a first step to this end, we enumerate desirable properties that such a benchmark should possess for the task of debugging text classifiers. Additionally, we highlight that such a benchmark facilitates not only assessing the effectiveness of explanations but also their efficiency.