SHAP-IQ: Unified Approximation of any-order Shapley Interactions

Published: 21 Sept 2023, Last Modified: 14 Jan 2024NeurIPS 2023 posterEveryoneRevisionsBibTeX
Keywords: Explainable Artificial Intelligence, Feature Interaction, Shapley Interaction, Shapley Value
TL;DR: A novel model-agnostic, sampling-based approximation technique for any Shapley Interaction Index that satisfies the linearity, symmetry and dummy axiom.
Abstract: Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models.
Supplementary Material: zip
Submission Number: 14013