Back to the Baseline: Examining Baseline Effects on Explainability Metrics

TL;DR

This paper critically examines how the choice of baseline affects explainability metrics in AI, revealing trade-offs and proposing a new baseline that better balances information removal and OOD image generation.

Contribution

It identifies limitations of current baselines in explainability metrics and introduces a novel, model-dependent baseline that improves the trade-off between information removal and OOD images.

Findings

Existing baselines either remove information or produce OOD images

No current baseline satisfies both desirable properties simultaneously

The proposed baseline improves the trade-off between these properties

Abstract

Attribution methods are among the most prevalent techniques in Explainable Artificial Intelligence (XAI) and are usually evaluated and compared using Fidelity metrics, with Insertion and Deletion being the most popular. These metrics rely on a baseline function to alter the pixels of the input image that the attribution map deems most important. In this work, we highlight a critical problem with these metrics: the choice of a given baseline will inevitably favour certain attribution methods over others. More concerningly, even a simple linear model with commonly used baselines contradicts itself by designating different optimal methods. A question then arises: which baseline should we use? We propose to study this problem through two desirable properties of a baseline: (i) that it removes information and (ii) that it does not produce overly out-of-distribution (OOD) images. We first show that none of the tested baselines satisfy both criteria, and there appears to be a trade-off among current baselines: either they remove information or they produce a sequence of OOD images. Finally, we introduce a novel baseline by leveraging recent work in feature visualisation to artificially produce a model-dependent baseline that removes information without being overly OOD, thus improving on the trade-off when compared to other existing baselines. Our code is available at https://github.com/deel-ai-papers/Back-to-the-Baseline