The XAI Disagreement Problem concerns the fact that various explainability methods yield different local/global insights on model behavior. Thus, given the lack of ground truth in explainability, practitioners are left wondering “Which explanation should I believe?”. In this work, we approach the Disagreement Problem from the point of view of Functional Decomposition (FD). First, we demonstrate that many XAI techniques disagree because they handle feature interactions differently. Secondly, we reduce interactions locally by fitting a so-called FD-Tree, which partitions the input space into regions where the model is approximately additive. Thus instead of providing global explanations aggregated over the whole dataset, we advocate reporting the FD-Tree structure as well as the regional explanations extracted from its leaves. The beneficial effects of FD-Trees on the Disagreement Problem are demonstrated on toy and real datasets.

Post-hoc global/local feature attribution methods are progressively being employed to understand the decisions of complex machine learning models. Yet, because of limited amounts of data, it is possible to obtain a diversity of models with good empirical performance but that provide very different explanations for the same prediction, making it hard to derive insight from them. In this work, instead of aiming at reducing the under-specification of model explanations, we fully embrace it and extract logical statements about feature attributions that are consistent across all models with good empirical performance (i.e. all models in the Rashomon Set). We show that partial orders of local/global feature importance arise from this methodology enabling more nuanced interpretations by allowing pairs of features to be incomparable when there is no consensus on their relative importance. We prove that every relation among features present in these partial orders also holds in the rankings provided by existing approaches. Finally, we present three use cases employing hypothesis spaces with tractable Rashomon Sets (Additive models, Kernel Ridge, and Random Forests) and show that partial orders allow one to extract consistent local and global interpretations of models despite their under-specification.

Machine Learning (ML) based defect prediction models can be used to improve the reliability and overall quality of software systems. However, such defect predictors might not be deployed in real applications due to the lack of transparency. Thus, recently, application of several post-hoc explanation methods (e.g., LIME and SHAP) have gained popularity. These explanation methods can offer insight by ranking features based on their importance in black box decisions. The explainability of ML techniques is reasonably novel in the Software Engineering community. However, it is still unclear whether such explainability methods genuinely help practitioners make better decisions regarding software maintenance. Recent user studies show that data scientists usually utilize multiple post-hoc explainers to understand a single model decision because of the lack of ground truth. Such a scenario causes disagreement between explainability methods and impedes drawing a conclusion. Therefore, our study first investigates three disagreement metrics between LIME and SHAP explanations of 10 defect-predictors, and exposes that disagreements regarding the rankings of feature importance are most frequent. Our findings lead us to propose a method of aggregating LIME and SHAP explanations that puts less emphasis on these disagreements while highlighting the aspect on which explanations agree.