Agree to Disagree: When Deep Learning Models With Identical Architectures Produce Distinct Explanations

TL;DR

This paper reveals that explanations for deep learning models in healthcare are highly inconsistent due to training variability, raising concerns about their reliability and trustworthiness in critical applications.

Contribution

The study introduces a measure of explanation consistency and demonstrates significant variability in explanations for neural networks, contrasting with the robustness of kernel methods.

Findings

Explanation consistency for neural networks is around 33%.

Kernel methods show 94% explanation consistency.

Model explanations vary significantly with training hyper-parameters.

Abstract

Deep Learning of neural networks has progressively become more prominent in healthcare with models reaching, or even surpassing, expert accuracy levels. However, these success stories are tainted by concerning reports on the lack of model transparency and bias against some medical conditions or patients' sub-groups. Explainable methods are considered the gateway to alleviate many of these concerns. In this study we demonstrate that the generated explanations are volatile to changes in model training that are perpendicular to the classification task and model structure. This raises further questions about trust in deep learning models for healthcare. Mainly, whether the models capture underlying causal links in the data or just rely on spurious correlations that are made visible via explanation methods. We demonstrate that the output of explainability methods on deep neural networks can vary significantly by changes of hyper-parameters, such as the random seed or how the training set is shuffled. We introduce a measure of explanation consistency which we use to highlight the identified problems on the MIMIC-CXR dataset. We find explanations of identical models but with different training setups have a low consistency: $\approx$ 33% on average. On the contrary, kernel methods are robust against any orthogonal changes, with explanation consistency at 94%. We conclude that current trends in model explanation are not sufficient to mitigate the risks of deploying models in real life healthcare applications.