Preference-based Conditional Treatment Effects and Policy Learning
This work addresses the challenge of flexible modeling of heterogeneous treatment effects with complex outcomes for researchers and practitioners in causal inference, offering a novel framework that unifies various applications.
The paper tackles the problem of estimating conditional treatment effects and learning policies when outcomes are only rankable under a preference rule, introducing the Conditional Preference-based Treatment Effect (CPTE) framework. It provides new identifiability conditions for previously unidentifiable estimands and demonstrates clear performance gains in experiments.
We introduce a new preference-based framework for conditional treatment effect estimation and policy learning, built on the Conditional Preference-based Treatment Effect (CPTE). CPTE requires only that outcomes be ranked under a preference rule, unlocking flexible modeling of heterogeneous effects with multivariate, ordinal, or preference-driven outcomes. This unifies applications such as conditional probability of necessity and sufficiency, conditional Win Ratio, and Generalized Pairwise Comparisons. Despite the intrinsic non-identifiability of comparison-based estimands, CPTE provides interpretable targets and delivers new identifiability conditions for previous unidentifiable estimands. We present estimation strategies via matching, quantile, and distributional regression, and further design efficient influence-function estimators to correct plug-in bias and maximize policy value. Synthetic and semi-synthetic experiments demonstrate clear performance gains and practical impact.