Any explanation that faithfully explains this type of model needs to be in agreement with this invariance property.

In Section B, we discuss feature importance methods in more detail.

Estimating personalized effects of treatments is a complex, yet pervasive problem.

Explainable AI (xAI) methods are important for establishing trust in using black-box models. However, recent criticism has mounted against current xAI methods that they disagree, are necessarily false, and can be manipulated, which has started to undermine the deployment of black-box models. Rudin (2019) goes so far as to say that we should stop using black-box models altogether in high-stakes cases because xAI explanations "must be wrong". However, strict fidelity to the truth is historically not a desideratum in science. Idealizations -- the intentional distortions introduced to scientific theories and models -- are commonplace in the natural sciences and are seen as a successful scientific tool. Thus, it is not falsehood qua falsehood that is the issue. In this paper, I outline the need for xAI research to engage in idealization evaluation. Drawing on the use of idealizations in the natural sciences and philosophy of science, I introduce a novel framework for evaluating whether xAI methods engage in successful idealizations or deceptive explanations (SIDEs). SIDEs evaluates whether the limitations of xAI methods, and the distortions that they introduce, can be part of a successful idealization or are indeed deceptive distortions as critics suggest. I discuss the role that existing research can play in idealization evaluation and where innovation is necessary. Through a qualitative analysis we find that leading feature importance methods and counterfactual explanations are subject to idealization failure and suggest remedies for ameliorating idealization failure.

While machine learning (ML) models are increasingly used due to their high predictive power, their use in understanding the data-generating process (DGP) is limited. Understanding the DGP requires insights into feature-target associations, which many ML models cannot directly provide, due to their opaque internal mechanisms. Feature importance (FI) methods provide useful insights into the DGP under certain conditions. Since the results of different FI methods have different interpretations, selecting the correct FI method for a concrete use case is crucial and still requires expert knowledge. This paper serves as a comprehensive guide to help understand the different interpretations of FI methods. Through an extensive review of FI methods and providing new proofs regarding their interpretation, we facilitate a thorough understanding of these methods and formulate concrete recommendations for scientific inference. We conclude by discussing options for FI uncertainty estimation and point to directions for future research aiming at full statistical inference from black-box ML models.

Interpretability methods are valuable only if their explanations faithfully describe the explained model. In this work, we consider neural networks whose predictions are invariant under a specific symmetry group. This includes popular architectures, ranging from convolutional to graph neural networks. Any explanation that faithfully explains this type of model needs to be in agreement with this invariance property. We formalize this intuition through the notion of explanation invariance and equivariance by leveraging the formalism from geometric deep learning. Through this rigorous formalism, we derive (1) two metrics to measure the robustness of any interpretability method with respect to the model symmetry group; (2) theoretical robustness guarantees for some popular interpretability methods and (3) a systematic approach to increase the invariance of any interpretability method with respect to a symmetry group. By empirically measuring our metrics for explanations of models associated with various modalities and symmetry groups, we derive a set of 5 guidelines to allow users and developers of interpretability methods to produce robust explanations.

Mean decrease in impurity (MDI) is a popular feature importance measure for random forests (RFs). We show that the MDI for a feature $X_k$ in each tree in an RF is equivalent to the unnormalized $R^2$ value in a linear regression of the response on the collection of decision stumps that split on $X_k$. We use this interpretation to propose a flexible feature importance framework called MDI+. Specifically, MDI+ generalizes MDI by allowing the analyst to replace the linear regression model and $R^2$ metric with regularized generalized linear models (GLMs) and metrics better suited for the given data structure. Moreover, MDI+ incorporates additional features to mitigate known biases of decision trees against additive or smooth models. We further provide guidance on how practitioners can choose an appropriate GLM and metric based upon the Predictability, Computability, Stability framework for veridical data science. Extensive data-inspired simulations show that MDI+ significantly outperforms popular feature importance measures in identifying signal features. We also apply MDI+ to two real-world case studies on drug response prediction and breast cancer subtype classification. We show that MDI+ extracts well-established predictive genes with significantly greater stability compared to existing feature importance measures. All code and models are released in a full-fledged python package on Github.