Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainability

TL;DR

This paper introduces Asymmetric Shapley values (ASVs), a new explainability framework that incorporates causal knowledge into model-agnostic explanations, enhancing interpretability, fairness testing, and feature selection in AI systems.

Contribution

The paper proposes ASVs, a flexible extension of Shapley values that integrates causal structures, addressing limitations of traditional methods and broadening explainability applications.

Findings

ASVs improve explanations by including causal information.

They enable testing for unfair discrimination.

Support for sequential explanations in time-series models.

Abstract

Explaining AI systems is fundamental both to the development of high performing models and to the trust placed in them by their users. The Shapley framework for explainability has strength in its general applicability combined with its precise, rigorous foundation: it provides a common, model-agnostic language for AI explainability and uniquely satisfies a set of intuitive mathematical axioms. However, Shapley values are too restrictive in one significant regard: they ignore all causal structure in the data. We introduce a less restrictive framework, Asymmetric Shapley values (ASVs), which are rigorously founded on a set of axioms, applicable to any AI system, and flexible enough to incorporate any causal structure known to be respected by the data. We demonstrate that ASVs can (i) improve model explanations by incorporating causal information, (ii) provide an unambiguous test for unfair discrimination in model predictions, (iii) enable sequentially incremental explanations in time-series models, and (iv) support feature-selection studies without the need for model retraining.