This paper presents a novel application of graph neural networks for modeling and estimating network heterogeneity. Network heterogeneity is a concept characterizing the dependence of an individual’s outcome or decision on their diverse local network scenarios. Graph neural networks are powerful tools for studying this dependence. We delineate the convergence rate of the graph neural networks estimator, as well as its applicability in semiparametric causal inference with heterogeneous treatment effects. The finite-sample performance of our estimator is evaluated through Monte Carlo simulations. In an empirical setting related to microfinance program participation, we apply the new estimator to examine the average treatment effects and outcomes of counterfactual policies, and to propose a Pareto frontier of strategies for selecting the initial recipients of program information in social networks.