A Fresh Look at Sanity Checks for Saliency Maps

TL;DR

This paper introduces two improved versions of the Model Parameter Randomisation Test (MPRT) to address methodological issues, enhancing the reliability of saliency map evaluations in XAI.

Contribution

It proposes Smooth MPRT and Efficient MPRT, modifications that improve the robustness and interpretability of model parameter randomisation tests.

Findings

Enhanced metric reliability with proposed modifications

Reduced noise impact through sampling in Smooth MPRT

Avoided biased similarity measures with Efficient MPRT

Abstract

The Model Parameter Randomisation Test (MPRT) is highly recognised in the eXplainable Artificial Intelligence (XAI) community due to its fundamental evaluative criterion: explanations should be sensitive to the parameters of the model they seek to explain. However, recent studies have raised several methodological concerns for the empirical interpretation of MPRT. In response, we propose two modifications to the original test: Smooth MPRT and Efficient MPRT. The former reduces the impact of noise on evaluation outcomes via sampling, while the latter avoids the need for biased similarity measurements by re-interpreting the test through the increase in explanation complexity after full model randomisation. Our experiments show that these modifications enhance the metric reliability, facilitating a more trustworthy deployment of explanation methods.