🤖 AI Summary
This paper addresses causal inference from a single large-scale network observation, confronting the high-dimensional network confounding challenge arising from simultaneous network interference and treatment selection endogeneity. To overcome the limitation of conventional methods—which rely on low-dimensional summaries of confounders—we propose two key innovations: (i) the first integration of graph neural networks (GNNs) into a causal inference framework, enabling end-to-end modeling of high-dimensional network confounding; and (ii) a “network-distance-decaying interference” assumption, imposing a low-dimensional structural constraint that ensures theoretical interpretability of shallow GNNs. Combining nonparametric causal modeling with simultaneous equations estimation, our approach delivers consistent estimation of the local average treatment effect under endogenous peer effects. Extensive experiments on synthetic and real-world network data demonstrate superior confounding control and estimation accuracy compared to existing methods.
📝 Abstract
This paper studies causal inference with observational data from a single large network. We consider a nonparametric model with interference in potential outcomes and selection into treatment. Both stages may be the outcomes of simultaneous equation models, which allow 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.
