Inference on Local Variable Importance Measures for Heterogeneous Treatment Effects

2025-10-21

AI Summary

This paper addresses the interpretability challenge of heterogeneous treatment effects (HTE) in high-stakes domains such as medicine. Methodologically, it proposes a unified framework integrating local variable importance assessment with global statistical inference, grounded in semiparametric functional parameter theory. It constructs a locally interpretable variable importance measure and embeds robust machine learning estimators to preserve statistical validity despite increasing model complexity. The key contribution is the first unification of individual-level variable importance quantification and population-level significance testing within a single inferential framework—thereby substantially enhancing both the interpretability and statistical credibility of black-box HTE models. Empirically applied to infectious disease prevention strategies, the method successfully identifies covariates with statistically significant impacts on treatment efficacy for specific subpopulations, demonstrating its practical applicability and reliability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Causality
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Large-scale security measurements

Abstract

We provide an inferential framework to assess variable importance for heterogeneous treatment effects. This assessment is especially useful in high-risk domains such as medicine, where decision makers hesitate to rely on black-box treatment recommendation algorithms. The variable importance measures we consider are local in that they may differ across individuals, while the inference is global in that it tests whether a given variable is important for any individual. Our approach builds on recent developments in semiparametric theory for function-valued parameters, and is valid even when statistical machine learning algorithms are employed to quantify treatment effect heterogeneity. We demonstrate the applicability of our method to infectious disease prevention strategies.

Problem

• Developing inferential framework for local variable importance in treatment effects
• Validating variable importance when using machine learning algorithms
• Assessing heterogeneous treatment effects in high-risk medical domains

Innovation

• Local variable importance measures for heterogeneous treatment effects
• Inferential framework using semiparametric theory for function-valued parameters
• Valid with machine learning algorithms for treatment effect heterogeneity