The Explanation Game: Explaining Machine Learning Models Using Shapley Values

TL;DR

This paper critically examines Shapley value-based explanations for machine learning models, highlighting formulation issues, proposing a unified game-theoretic framework with confidence intervals, and connecting explanations to contrastive psychology.

Contribution

It introduces a general game formulation unifying existing methods, enabling confidence intervals and contrastive explanations, and presents the FAE framework for interpreting ML models.

Findings

Differences in game formulations significantly affect attribution results.

The proposed framework provides confidence intervals for feature attributions.

Application to real datasets demonstrates improved interpretability.

Abstract

A number of techniques have been proposed to explain a machine learning model's prediction by attributing it to the corresponding input features. Popular among these are techniques that apply the Shapley value method from cooperative game theory. While existing papers focus on the axiomatic motivation of Shapley values, and efficient techniques for computing them, they offer little justification for the game formulations used, and do not address the uncertainty implicit in their methods' outputs. For instance, the popular SHAP algorithm's formulation may give substantial attributions to features that play no role in the model. In this work, we illustrate how subtle differences in the underlying game formulations of existing methods can cause large differences in the attributions for a prediction. We then present a general game formulation that unifies existing methods, and enables straightforward confidence intervals on their attributions. Furthermore, it allows us to interpret the attributions as contrastive explanations of an input relative to a distribution of reference inputs. We tie this idea to classic research in cognitive psychology on contrastive explanations, and propose a conceptual framework for generating and interpreting explanations for ML models, called formulate, approximate, explain (FAE). We apply this framework to explain black-box models trained on two UCI datasets and a Lending Club dataset.