Instrumental and Proximal Causal Inference with Gaussian Processes

2026-03-02

AI Summary

This work addresses the challenge of reliably quantifying epistemic uncertainty in causal effect estimation under unobserved confounding, where existing instrumental variable and proximal causal methods often fall short. The authors propose a deconfounded Gaussian process (DGP) framework that integrates Gaussian processes into causal inference, leveraging posterior mean and variance to deliver accurate point predictions and well-calibrated estimates of epistemic uncertainty, respectively. This approach unifies kernel-based estimation with Bayesian uncertainty quantification and enables systematic model selection via marginal log-likelihood. Empirical evaluations demonstrate that DGP consistently achieves superior predictive performance and more reliable uncertainty estimates across multiple benchmarks, significantly outperforming current methods in terms of empirical coverage frequency and accuracy–rejection curves tailored to decision-aware settings.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Causality
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Instrumental variable (IV) and proximal causal learning (Proxy) methods are central frameworks for causal inference in the presence of unobserved confounding. Despite substantial methodological advances, existing approaches rarely provide reliable epistemic uncertainty (EU) quantification. We address this gap through a Deconditional Gaussian Process (DGP) framework for uncertainty-aware causal learning. Our formulation recovers popular kernel estimators as the posterior mean, ensuring predictive precision, while the posterior variance yields principled and well-calibrated EU. Moreover, the probabilistic structure enables systematic model selection via marginal log-likelihood optimization. Empirical results demonstrate strong predictive performance alongside informative EU quantification, evaluated via empirical coverage frequencies and decision-aware accuracy rejection curves. Together, our approach provides a unified, practical solution for causal inference under unobserved confounding with reliable uncertainty.

Problem

• instrumental variable
• proximal causal learning
• unobserved confounding
• epistemic uncertainty
• causal inference

Innovation

• Deconditional Gaussian Process
• Epistemic Uncertainty
• Instrumental Variable
• Proximal Causal Learning
• Causal Inference