How Good is your Explanation? Algorithmic Stability Measures to Assess the Quality of Explanations for Deep Neural Networks
This work addresses the need for reliable evaluation metrics in explainable AI, which is crucial for building trust in AI systems, though it is incremental as it builds on existing stability concepts.
The paper tackles the problem of objectively assessing the quality of explanations for deep neural networks by proposing two new measures, mean generalizability (MeGe) and relative consistency (ReCo), based on algorithmic stability. Experiments show that 1-Lipschitz networks produce explanations with higher MeGe and ReCo than common networks while maintaining similar accuracy.
A plethora of methods have been proposed to explain how deep neural networks reach their decisions but comparatively, little effort has been made to ensure that the explanations produced by these methods are objectively relevant. While several desirable properties for trustworthy explanations have been formulated, objective measures have been harder to derive. Here, we propose two new measures to evaluate explanations borrowed from the field of algorithmic stability: mean generalizability MeGe and relative consistency ReCo. We conduct extensive experiments on different network architectures, common explainability methods, and several image datasets to demonstrate the benefits of the proposed measures.In comparison to ours, popular fidelity measures are not sufficient to guarantee trustworthy explanations.Finally, we found that 1-Lipschitz networks produce explanations with higher MeGe and ReCo than common neural networks while reaching similar accuracy. This suggests that 1-Lipschitz networks are a relevant direction towards predictors that are more explainable and trustworthy.