Prediction Intervals for Individual Treatment Effects in a Multiple Decision Point Framework using Conformal Inference

2025-12-09

AI Summary

This study addresses the problem of quantifying uncertainty in individual treatment effects (ITEs) across multiple sequential decision points, enabling personalized dynamic interventions in domains such as healthcare and education. We propose the first conformal prediction framework for time-varying causal inference—extending conformal inference to multi-stage settings for the first time. Our method relies solely on a weak non-exchangeability assumption to guarantee rigorous lower bounds on prediction interval coverage probability, thereby relaxing stringent requirements of strong exchangeability or stationarity imposed by prior approaches. Integrating sequential causal modeling with the micro-randomized trial (MRT) design, we develop a calibration strategy tailored for longitudinal ITE uncertainty quantification. Empirical evaluation on simulated MRTs and real-world data from the Intern Health Study demonstrates that our prediction intervals achieve coverage rates precisely matching the theoretical lower bound—substantially outperforming existing cross-sectional methods.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Calibration & Uncertainty Quantification
• Planning, Routing, and Scheduling: Scheduling under Uncertainty

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurately quantifying uncertainty of individual treatment effects (ITEs) across multiple decision points is crucial for personalized decision-making in fields such as healthcare, finance, education, and online marketplaces. Previous work has focused on predicting non-causal longitudinal estimands or constructing prediction bands for ITEs using cross-sectional data based on exchangeability assumptions. We propose a novel method for constructing prediction intervals using conformal inference techniques for time-varying ITEs with weaker assumptions than prior literature. We guarantee a lower bound for coverage, which is dependent on the degree of non-exchangeability in the data. Although our method is broadly applicable across decision-making contexts, we support our theoretical claims with simulations emulating micro-randomized trials (MRTs) -- a sequential experimental design for mobile health (mHealth) studies. We demonstrate the practical utility of our method by applying it to a real-world MRT - the Intern Health Study (IHS).

Problem

• Constructs prediction intervals for time-varying individual treatment effects
• Uses conformal inference with weaker assumptions than prior methods
• Applies to multiple decision points in personalized decision-making contexts

Innovation

• Conformal inference for time-varying individual treatment effects
• Weaker assumptions than prior exchangeability-based methods
• Guaranteed coverage lower bound dependent on non-exchangeability