Budgeted Active Experimentation for Treatment Effect Estimation from Observational and Randomized Data

Jiacan Gao, Xinyan Su, Mingyuan Ma, Yiyan Huang, Xiao Xu, Xinrui Wan, Tianqi Gu, Enyun Yu, Jiecheng Guo, Zhiheng Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:33592-33631, 2026.

Abstract

Estimating heterogeneous treatment effects is central to data-driven decision-making, yet industrial applications often face a fundamental tension between limited randomized controlled trial (RCT) budgets and abundant but biased observational data (OBS) collected under historical targeting policies. Although observational logs offer the advantage of scale, they may suffer from severe policy-induced imbalance and overlap violations, rendering standalone estimation unreliable. We propose a budgeted active experimentation framework that iteratively collects informative randomized samples for causal effect estimation via active sampling. By leveraging observational signals, we develop an acquisition function targeting uplift estimation uncertainty, domain discrepancy, and overlap deficits to select the most informative units for randomized experiments. We establish finite-sample deviation bounds, asymptotic normality via martingale CLTs, and minimax lower bounds showing near-optimality in the linear representation setting. Experiments on synthetic datasets support our theoretical findings, and further extensions to industrial neural network-based uplift modeling scenarios show that active sampling can improve sample efficiency over random sampling under limited RCT budgets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-gao26z, title = {Budgeted Active Experimentation for Treatment Effect Estimation from Observational and Randomized Data}, author = {Gao, Jiacan and Su, Xinyan and Ma, Mingyuan and Huang, Yiyan and Xu, Xiao and Wan, Xinrui and Gu, Tianqi and Yu, Enyun and Guo, Jiecheng and Zhang, Zhiheng}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {33592--33631}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/gao26z/gao26z.pdf}, url = {https://proceedings.mlr.press/v306/gao26z.html}, abstract = {Estimating heterogeneous treatment effects is central to data-driven decision-making, yet industrial applications often face a fundamental tension between limited randomized controlled trial (RCT) budgets and abundant but biased observational data (OBS) collected under historical targeting policies. Although observational logs offer the advantage of scale, they may suffer from severe policy-induced imbalance and overlap violations, rendering standalone estimation unreliable. We propose a budgeted active experimentation framework that iteratively collects informative randomized samples for causal effect estimation via active sampling. By leveraging observational signals, we develop an acquisition function targeting uplift estimation uncertainty, domain discrepancy, and overlap deficits to select the most informative units for randomized experiments. We establish finite-sample deviation bounds, asymptotic normality via martingale CLTs, and minimax lower bounds showing near-optimality in the linear representation setting. Experiments on synthetic datasets support our theoretical findings, and further extensions to industrial neural network-based uplift modeling scenarios show that active sampling can improve sample efficiency over random sampling under limited RCT budgets.} }
Endnote
%0 Conference Paper %T Budgeted Active Experimentation for Treatment Effect Estimation from Observational and Randomized Data %A Jiacan Gao %A Xinyan Su %A Mingyuan Ma %A Yiyan Huang %A Xiao Xu %A Xinrui Wan %A Tianqi Gu %A Enyun Yu %A Jiecheng Guo %A Zhiheng Zhang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-gao26z %I PMLR %P 33592--33631 %U https://proceedings.mlr.press/v306/gao26z.html %V 306 %X Estimating heterogeneous treatment effects is central to data-driven decision-making, yet industrial applications often face a fundamental tension between limited randomized controlled trial (RCT) budgets and abundant but biased observational data (OBS) collected under historical targeting policies. Although observational logs offer the advantage of scale, they may suffer from severe policy-induced imbalance and overlap violations, rendering standalone estimation unreliable. We propose a budgeted active experimentation framework that iteratively collects informative randomized samples for causal effect estimation via active sampling. By leveraging observational signals, we develop an acquisition function targeting uplift estimation uncertainty, domain discrepancy, and overlap deficits to select the most informative units for randomized experiments. We establish finite-sample deviation bounds, asymptotic normality via martingale CLTs, and minimax lower bounds showing near-optimality in the linear representation setting. Experiments on synthetic datasets support our theoretical findings, and further extensions to industrial neural network-based uplift modeling scenarios show that active sampling can improve sample efficiency over random sampling under limited RCT budgets.
APA
Gao, J., Su, X., Ma, M., Huang, Y., Xu, X., Wan, X., Gu, T., Yu, E., Guo, J. & Zhang, Z.. (2026). Budgeted Active Experimentation for Treatment Effect Estimation from Observational and Randomized Data. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:33592-33631 Available from https://proceedings.mlr.press/v306/gao26z.html.

Related Material