Graph Neural Networks for Causal Inference Under Network Confounding

TL;DR

This paper introduces a method using graph neural networks to address high-dimensional network confounding in causal inference from observational network data, leveraging low-dimensional structure when interference diminishes with distance.

Contribution

It proposes a novel application of GNNs for causal inference under complex network confounding, extending beyond existing low-dimensional assumptions.

Findings

GNNs effectively adjust for network confounding in high-dimensional settings.

Low-depth GNNs are justified when interference decays with network distance.

The approach enables feasible estimation in complex network causal models.

Abstract

This paper studies causal inference with observational data from a single large network. We consider a nonparametric model with interference in both potential outcomes and selection into treatment. Specifically, both stages may be the outcomes of simultaneous equations models, allowing for endogenous peer effects. This results in high-dimensional network confounding where the network and covariates of all units constitute sources of selection bias. In contrast, the existing literature assumes that confounding can be summarized by a known, low-dimensional function of these objects. We propose to use graph neural networks (GNNs) to adjust for network confounding. When interference decays with network distance, we argue that the model has low-dimensional structure that makes estimation feasible and justifies the use of shallow GNN architectures.