Shapley Flow: A Graph-based Approach to Interpreting Model Predictions

TL;DR

Shapley Flow introduces a graph-based method for interpreting machine learning models by assigning importance to edges in a causal graph, providing deeper insights into feature interactions and model behavior.

Contribution

It proposes a novel approach that generalizes Shapley values to directed acyclic graphs, enabling comprehensive and interpretable feature importance attribution.

Findings

Effectively explains the flow of importance in models

Extends existing methods to a graph-based framework

Facilitates understanding of intervention impacts

Abstract

Many existing approaches for estimating feature importance are problematic because they ignore or hide dependencies among features. A causal graph, which encodes the relationships among input variables, can aid in assigning feature importance. However, current approaches that assign credit to nodes in the causal graph fail to explain the entire graph. In light of these limitations, we propose Shapley Flow, a novel approach to interpreting machine learning models. It considers the entire causal graph, and assigns credit to \textit{edges} instead of treating nodes as the fundamental unit of credit assignment. Shapley Flow is the unique solution to a generalization of the Shapley value axioms to directed acyclic graphs. We demonstrate the benefit of using Shapley Flow to reason about the impact of a model's input on its output. In addition to maintaining insights from existing approaches, Shapley Flow extends the flat, set-based, view prevalent in game theory based explanation methods to a deeper, \textit{graph-based}, view. This graph-based view enables users to understand the flow of importance through a system, and reason about potential interventions.