causal inference

In causal effect estimation, the absence of standardized hyperparameter tuning evaluation criteria impedes reliable model selection and creates a substantial gap between commonly used metrics and true performance. This paper systematically investigates the interplay between hyperparameter tuning and evaluation, jointly analyzing estimators (T-/X-/R-Learner), base learners (random forests, gradient boosting, neural networks), and evaluation metrics (IPW, DR, PEHE) across four benchmark datasets. Key findings are: (1) thorough hyperparameter tuning eliminates performance differences among mainstream causal estimators; (2) the choice of evaluation strategy exerts greater influence on final performance than either the estimator type or base learner architecture; and (3) existing evaluation metrics underestimate the performance gain from optimal model selection by over 35% on average. These results demonstrate that hyperparameter tuning is the primary determinant of causal estimation accuracy, underscoring an urgent need for more robust, theoretically grounded evaluation paradigms in causal machine learning.

Estimating causal effects from observational data requires selecting appropriate control and instrumental variables that satisfy causal identification conditions—a challenging task often reliant on strong domain knowledge or ad hoc assumptions. Method: This paper proposes the first end-to-end joint learning framework that automatically identifies valid combinations of control and instrumental variables. Grounded in conditional independence testing, the method integrates nonparametric dependence measures with structural search optimization, ensuring statistical consistency in variable selection under mild regularity conditions. Contribution/Results: Unlike conventional approaches requiring prespecified variable sets or strong prior assumptions, our framework is fully data-driven. In simulations, it achieves significantly higher variable identification accuracy. Empirically, applied to the Job Corps study, its estimated treatment effect closely aligns with results from the randomized controlled trial—demonstrating both validity and robustness in real-world causal inference.

In high-dimensional observational settings of randomized controlled trials (RCTs), causal treatment effect estimation is prone to modeling and sampling selection biases. Method: We introduce ISTAnt—the first real-world visual causal benchmark grounded in ant-behavior RCTs—and theoretically prove that classification accuracy cannot serve as a proxy for causal estimation quality. We further propose representation-learning principles tailored for scientific causal inference. Contribution/Results: Through tripartite validation—rigorous theoretical analysis, controlled synthetic experiments, and real biological experiments—across 6,480 large-scale fine-tuned models built upon state-of-the-art vision backbones, we demonstrate that common deep learning practices (e.g., loss design, data sampling) induce substantial systematic biases, and classification performance exhibits no strong correlation with causal estimation accuracy. Our work establishes a reproducible benchmark, theoretical criteria, and practical guidelines for high-dimensional causal inference.

In clinical practice, relative risk (RR) is more clinically interpretable than absolute risk difference; however, mainstream heterogeneous treatment effect (HTE) methods—such as causal forests—rely on absolute risk differences for recursive partitioning, often overlooking RR heterogeneity. To address this, we propose the first RR-oriented variant of causal forest: a nonparametric node-splitting criterion grounded in generalized linear model comparisons, explicitly designed to maximize statistical power for detecting RR heterogeneity. Our method imposes no strong distributional assumptions and automatically identifies covariates—and their interaction structures—that drive RR heterogeneity. Simulation studies and empirical analyses demonstrate that the proposed framework substantially improves identification of clinically relevant subgroups, successfully capturing RR heterogeneity patterns missed by conventional causal forests. It achieves this while preserving computational feasibility, thereby enhancing both clinical interpretability and statistical power.

Direct effect estimation on a selected causal graph induces selection bias due to data reuse, invalidating confidence intervals. Method: We propose the first post-selection inference framework for fixed-population causal effect parameters, integrating resampling with graph-structure screening to depart from the conventional “select-then-infer” paradigm. Built upon the PC algorithm, our approach unifies conditional independence testing, Gaussian modeling, and joint estimation over multiple candidate graphs, and is modularly extensible to other causal discovery algorithms and distribution families. Contribution/Results: We establish asymptotic validity—specifically, asymptotically exact coverage—for confidence sets targeting the true causal effect. Empirical evaluations demonstrate that our method substantially improves reliability and robustness of causal inference under uncertainty, yielding well-calibrated confidence sets even after graph selection.

This study addresses the challenge of identifying and estimating causal effects under network interference, where an individual’s treatment may spill over and affect others’ outcomes. The authors propose a solution based on a linear outcome model that yields unbiased and consistent estimates of both binary and continuous treatment effects when the interference structure is known or partially known. The approach accommodates both fixed and random interference network specifications and innovatively eliminates interference-induced bias while remaining compatible with standard linear regression software. It also conveniently allows for the incorporation of random effects and heteroskedasticity- and autocorrelation-consistent (HAC) standard errors. Numerical simulations and empirical analyses demonstrate the method’s effectiveness in bias correction and practical applicability.

This study addresses the core challenge in causal inference of accurately identifying true causal effects in settings characterized by high-dimensional observational data and endogenous selection. Leveraging both experimental data from a large technology company’s new feature rollout and observational data from users’ self-selection into the feature, this work provides the first joint validation of causal machine learning methods in a real-world product environment. By integrating propensity score modeling, doubly robust estimation, and high-dimensional covariate adjustment, the research demonstrates that careful modeling substantially improves the accuracy of causal effect estimates. The findings not only confirm the practical feasibility of modern causal inference techniques but also distill a set of best practices for enhancing estimation credibility, offering an empirical benchmark and actionable guidance for high-dimensional causal inference.

This work addresses the frequent bias in causal machine learning applications within observational health research, often stemming from overlooked key assumptions. It proposes a clinical-context-oriented roadmap for applying causal machine learning, systematically integrating clinical domain knowledge with causal inference theory to construct an actionable analytical framework. The approach emphasizes evaluating the plausibility of causal assumptions under real-world data constraints and fosters transparent, rigorous, and interpretable causal analyses through close collaboration between clinicians and machine learning researchers. By grounding causal modeling in clinical reality and methodological rigor, the framework enhances both the scientific validity and practical utility of findings derived from observational studies.

This study addresses the limitations of traditional control-based causal inference methods—such as matching and difference-in-differences—in settings characterized by pervasive or structurally ambiguous spillover effects, where reliance on uncontaminated control units impedes accurate identification of both average direct and spillover effects. Within the potential outcomes framework, this work provides the first systematic comparison between control-based and prediction-based counterfactual approaches—including interrupted time series and machine learning control—in terms of their identification capabilities. Through simulation and empirical analyses, the authors demonstrate that in environments with widespread interference, prediction-based methods can more reliably estimate certain causal parameters over short horizons, circumventing the stringent assumption of unperturbed units and thereby offering a promising alternative for causal inference under complex interference.