XAI-TRIS: Non-linear image benchmarks to quantify false positive post-hoc attribution of feature importance
This work addresses the lack of rigorous validation for XAI methods, which is a critical issue for researchers and practitioners relying on interpretability in deep learning applications, though it is incremental in providing benchmarks rather than new methods.
The authors tackled the problem of evaluating post-hoc feature attribution methods in explainable AI (XAI) by creating benchmark datasets with known ground truth explanations for non-linear classification scenarios, and they found that popular XAI methods often perform no better than random baselines or edge detection methods, with explanations varying widely across model architectures.
The field of 'explainable' artificial intelligence (XAI) has produced highly cited methods that seek to make the decisions of complex machine learning (ML) methods 'understandable' to humans, for example by attributing 'importance' scores to input features. Yet, a lack of formal underpinning leaves it unclear as to what conclusions can safely be drawn from the results of a given XAI method and has also so far hindered the theoretical verification and empirical validation of XAI methods. This means that challenging non-linear problems, typically solved by deep neural networks, presently lack appropriate remedies. Here, we craft benchmark datasets for three different non-linear classification scenarios, in which the important class-conditional features are known by design, serving as ground truth explanations. Using novel quantitative metrics, we benchmark the explanation performance of a wide set of XAI methods across three deep learning model architectures. We show that popular XAI methods are often unable to significantly outperform random performance baselines and edge detection methods. Moreover, we demonstrate that explanations derived from different model architectures can be vastly different; thus, prone to misinterpretation even under controlled conditions.