Problems with Shapley-value-based explanations as feature importance measures

TL;DR

This paper critically examines the use of Shapley values for feature importance in machine learning, highlighting mathematical issues and questioning their effectiveness for human-centered explanations.

Contribution

It reveals fundamental mathematical problems with Shapley-value-based explanations and argues they are not well-suited for human-centric interpretability goals.

Findings

Mathematical issues arise with Shapley-value-based explanations.

Mitigating these issues often requires complex causal reasoning.

Shapley values may not align with human interpretability needs.

Abstract

Game-theoretic formulations of feature importance have become popular as a way to "explain" machine learning models. These methods define a cooperative game between the features of a model and distribute influence among these input elements using some form of the game's unique Shapley values. Justification for these methods rests on two pillars: their desirable mathematical properties, and their applicability to specific motivations for explanations. We show that mathematical problems arise when Shapley values are used for feature importance and that the solutions to mitigate these necessarily induce further complexity, such as the need for causal reasoning. We also draw on additional literature to argue that Shapley values do not provide explanations which suit human-centric goals of explainability.