Multi-objective Explanations of GNN Predictions

TL;DR

This paper introduces a multi-objective optimization approach to generate explanations for GNN predictions that balance interpretability and relevance, using a model-agnostic search algorithm validated across multiple datasets.

Contribution

It proposes a novel multi-objective explanation method for GNNs that balances simulatability and counterfactual relevance without needing model access.

Findings

Pareto optimal explanations outperform single-objective baselines.

Explanations are robust and sensitive to confounders.

Counterfactuals support algorithmic recourse and fairness.

Abstract

Graph Neural Network (GNN) has achieved state-of-the-art performance in various high-stake prediction tasks, but multiple layers of aggregations on graphs with irregular structures make GNN a less interpretable model. Prior methods use simpler subgraphs to simulate the full model, or counterfactuals to identify the causes of a prediction. The two families of approaches aim at two distinct objectives, "simulatability" and "counterfactual relevance", but it is not clear how the objectives can jointly influence the human understanding of an explanation. We design a user study to investigate such joint effects and use the findings to design a multi-objective optimization (MOO) algorithm to find Pareto optimal explanations that are well-balanced in simulatability and counterfactual. Since the target model can be of any GNN variants and may not be accessible due to privacy concerns, we design a search algorithm using zeroth-order information without accessing the architecture and parameters of the target model. Quantitative experiments on nine graphs from four applications demonstrate that the Pareto efficient explanations dominate single-objective baselines that use first-order continuous optimization or discrete combinatorial search. The explanations are further evaluated in robustness and sensitivity to show their capability of revealing convincing causes while being cautious about the possible confounders. The diverse dominating counterfactuals can certify the feasibility of algorithmic recourse, that can potentially promote algorithmic fairness where humans are participating in the decision-making using GNN.