Heterogeneous Treatment Effects under Network Interference: A Nonparametric Approach Based on Node Connectivity

2024-10-15

AI Summary

Network interference complicates causal inference, as individual outcomes depend on the treatment status of neighboring nodes; conventional methods estimate only the population-average treatment effect (PATE), failing to capture node-level heterogeneity in treatment effects. To address this, we propose the first framework for estimating node-level counterfactual means, transcending the limitations of population-level aggregation. We introduce KECENI—a novel nonparametric kernel estimator leveraging node connectivity—that achieves double robustness and asymptotic normality under network dependence. We establish its consistency and asymptotic normality under general network topologies. Applying KECENI to microfinance data, we systematically identify how network centrality, clustering coefficient, and other topological features differentially moderate individual treatment effects—an analysis previously infeasible. Our framework provides a fine-grained, interpretable tool for network causal inference, enabling heterogeneous effect estimation while rigorously accounting for interference.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

In network settings, interference between units makes causal inference more challenging as outcomes may depend on the treatments received by others in the network. Typical estimands in network settings focus on treatment effects aggregated across individuals in the population. We propose a framework for estimating node-wise counterfactual means, allowing for more granular insights into the impact of network structure on treatment effect heterogeneity. We develop a doubly robust and non-parametric estimation procedure, KECENI (Kernel Estimation of Causal Effect under Network Interference), which offers consistency and asymptotic normality under network dependence. The utility of this method is demonstrated through an application to microfinance data, revealing the impact of network characteristics on treatment effects.

Problem

• Estimating node-wise treatment effects in networks
• Addressing interference challenges in causal inference
• Analyzing network structure impact on treatment heterogeneity

Innovation

• Node-wise counterfactual means estimation
• Doubly robust non-parametric KECENI method
• Kernel-based estimation under network interference