IDEAS home Printed from https://ideas.repec.org/p/arx/papers/2606.08853.html

AI-Assisted Variance Reduction in Randomized Experiments

Author

Listed:
  • David Arbour
  • Eli Ben-Michael
  • Avi Feller
  • Apoorva Lal
  • Lo-Hua Yuan

Abstract

Generative AI and large language models can produce realistic predictions of human behavior from rich, unstructured inputs with little to no task-specific training data. Recent work uses these ``digital twin'' predictions to supplement human responses in surveys and experiments. We study the special case of using AI-generated predictions to reduce variance in randomized experiments. We argue that doing so requires no new estimators and that researchers can simply include AI predictions as covariates in standard regression adjustment, analogous to adjusting for a prognostic score. A benefit of this approach is a ``do no harm'' property whereby the adjusted estimator reverts to the unadjusted difference in means when predictions are uninformative. Other methods, such as variants of prediction-powered inference, do not have this guarantee. We provide implementation guidance, including how to obtain continuous scores from discrete LLM outputs and how to use LLMs to featurize unstructured inputs as auxiliary covariates. We demonstrate these ideas in simulations and three empirical applications: a survey mega-study, an email marketing A/B test, and a large-scale technology platform experiment. Overall, efficiency gains are real if modest, with greater benefits in studies that contain substantial text and other unstructured data. We also confirm the do no harm property empirically. Given these gains and limited costs, we recommend adjusting for AI-generated predictions as a regular empirical practice.

Suggested Citation

  • David Arbour & Eli Ben-Michael & Avi Feller & Apoorva Lal & Lo-Hua Yuan, 2026. "AI-Assisted Variance Reduction in Randomized Experiments," Papers 2606.08853, arXiv.org.
  • Handle: RePEc:arx:papers:2606.08853
    as

    Download full text from publisher

    File URL: https://arxiv.org/pdf/2606.08853
    File Function: Latest version
    Download Restriction: no
    ---><---

    References listed on IDEAS

    as
    1. Haoge Chang & Joel A. Middleton & P. M. Aronow, 2024. "Exact Bias Correction for Linear Adjustment of Randomized Controlled Trials," Econometrica, Econometric Society, vol. 92(5), pages 1503-1519, September.
    2. repec:mpr:mprres:6573 is not listed on IDEAS
    3. Zhexiao Lin & Peng Ding & Fang Han, 2023. "Estimation Based on Nearest Neighbor Matching: From Density Ratio to Average Treatment Effect," Econometrica, Econometric Society, vol. 91(6), pages 2187-2217, November.
    4. George Gui & Seungwoo Kim, 2025. "Leveraging LLMs to Improve Experimental Design: A Generative Stratification Approach," Papers 2509.25709, arXiv.org.
    5. Ben B. Hansen, 2008. "The prognostic analogue of the propensity score," Biometrika, Biometrika Trust, vol. 95(2), pages 481-488.
    6. David Broska & Michael Howes & Austin van Loon, 2025. "The Mixed Subjects Design: Treating Large Language Models as Potentially Informative Observations," Sociological Methods & Research, , vol. 54(3), pages 1074-1109, August.
    7. Imbens,Guido W. & Rubin,Donald B., 2015. "Causal Inference for Statistics, Social, and Biomedical Sciences," Cambridge Books, Cambridge University Press, number 9780521885881.
    8. Argyle, Lisa P. & Busby, Ethan C. & Fulda, Nancy & Gubler, Joshua R. & Rytting, Christopher & Wingate, David, 2023. "Out of One, Many: Using Language Models to Simulate Human Samples," Political Analysis, Cambridge University Press, vol. 31(3), pages 337-351, July.
    9. Xinran Li & Peng Ding, 2020. "Rerandomization and regression adjustment," Journal of the Royal Statistical Society Series B, Royal Statistical Society, vol. 82(1), pages 241-268, February.
    10. Xinrui Ruan & Xinwei Ma & Yingfei Wang & Waverly Wei & Jingshen Wang, 2025. "Can language models boost the power of randomized experiments without statistical bias?," Papers 2510.05545, arXiv.org, revised Dec 2025.
    11. John J. Horton & Apostolos Filippas & Benjamin S. Manning, 2023. "Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?," NBER Working Papers 31122, National Bureau of Economic Research, Inc.
    12. Fangzhou Su & Peng Ding, 2021. "Model‐assisted analyses of cluster‐randomized experiments," Journal of the Royal Statistical Society Series B, Royal Statistical Society, vol. 83(5), pages 994-1015, November.
    13. Kevin Guo & Guillaume Basse, 2023. "The Generalized Oaxaca-Blinder Estimator," Journal of the American Statistical Association, Taylor & Francis Journals, vol. 118(541), pages 524-536, January.
    14. P L Cohen & C B Fogarty, 2024. "No-harm calibration for generalized Oaxaca–Blinder estimators," Biometrika, Biometrika Trust, vol. 111(1), pages 331-338.
    Full references (including those not matched with items on IDEAS)

    Most related items

    These are the items that most often cite the same works as this one and are cited by the same works as this one.
    1. Lin, Zhexiao & Han, Fang, 2025. "On regression-adjusted imputation estimators of average treatment effects," Journal of Econometrics, Elsevier, vol. 251(C).
    2. Gao, Mengsi & Ding, Peng, 2025. "Causal inference in network experiments: Regression-based analysis and design-based properties," Journal of Econometrics, Elsevier, vol. 252(PA).
    3. Zhao, Anqi & Ding, Peng, 2024. "No star is good news: A unified look at rerandomization based on p-values from covariate balance tests," Journal of Econometrics, Elsevier, vol. 241(1).
    4. Zhexiao Lin & Pablo Crespo, 2024. "Variance reduction combining pre-experiment and in-experiment data," Papers 2410.09027, arXiv.org, revised Mar 2026.
    5. Felipe A. Csaszar & Harsh Ketkar & Hyunjin Kim, 2024. "Artificial Intelligence and Strategic Decision-Making: Evidence from Entrepreneurs and Investors," Strategy Science, INFORMS, vol. 9(4), pages 322-345, December.
    6. Shenkman, E. & Parshakov, P. & Matkin , N., 2026. "Comparative analysis of large language models and human behavior in the Dictator Game," Journal of the New Economic Association, New Economic Association, vol. 71(2), pages 177-203.
    7. Jun Wang & Yahe Yu, 2024. "Improved estimation of average treatment effects under covariate‐adaptive randomization methods," Statistica Neerlandica, Netherlands Society for Statistics and Operations Research, vol. 78(2), pages 310-333, May.
    8. John R. Graham & Campbell R. Harvey & Manish Jha, 2026. "CFOs Meet LLMs," Papers 2606.13812, arXiv.org.
    9. Hongshen Sun & Juanjuan Zhang, 2025. "From Model Choice to Model Belief: Establishing a New Measure for LLM-Based Research," Papers 2512.23184, arXiv.org.
    10. Zhexiao Lin & Peng Ding, 2025. "Unifying regression-based and design-based causal inference in time-series experiments," Papers 2510.22864, arXiv.org.
    11. Koji Takahashi & Joon Suk Park, 2025. "Generative AI for Surveys on Payment Apps: AIs' View on Privacy and Technology," IMES Discussion Paper Series 25-E-13, Institute for Monetary and Economic Studies, Bank of Japan.
    12. Zhexiao Lin & Peng Ding & Fang Han, 2023. "Estimation Based on Nearest Neighbor Matching: From Density Ratio to Average Treatment Effect," Econometrica, Econometric Society, vol. 91(6), pages 2187-2217, November.
    13. Hui Chen & Antoine Didisheim & Mohammad & Pourmohammadi & Luciano Somoza & Hanqing Tian, 2025. "A Financial Brain Scan of the LLM," Papers 2508.21285, arXiv.org, revised Feb 2026.
    14. repec:osf:osfxxx:r3qng_v1 is not listed on IDEAS
    15. Matthew O. Jackson & Qiaozhu Me & Stephanie W. Wang & Yutong Xie & Walter Yuan & Seth Benzell & Erik Brynjolfsson & Colin F. Camerer & James Evans & Brian Jabarian & Jon Kleinberg & Juanjuan Meng & Se, 2025. "AI Behavioral Science," Papers 2509.13323, arXiv.org, revised May 2026.
    16. Shilei Luo & Zhiqi Zhang & Hengchen Dai & Dennis Zhang, 2026. "Behavioral Transfer in AI Agents: Evidence and Privacy Implications," Papers 2604.19925, arXiv.org.
    17. George Gui & Seungwoo Kim, 2025. "Leveraging LLMs to Improve Experimental Design: A Generative Stratification Approach," Papers 2509.25709, arXiv.org.
    18. Aliya Amirova & Theodora Fteropoulli & Nafiso Ahmed & Martin R Cowie & Joel Z Leibo, 2024. "Framework-based qualitative analysis of free responses of Large Language Models: Algorithmic fidelity," PLOS ONE, Public Library of Science, vol. 19(3), pages 1-33, March.
    19. Sugat Chaturvedi & Rochana Chaturvedi, 2025. "Who Gets the Callback? Generative AI and Gender Bias," Papers 2504.21400, arXiv.org.
    20. Olivier Toubia & George Z. Gui & Tianyi Peng & Daniel J. Merlau & Ang Li & Haozhe Chen, 2025. "Database Report: Twin-2K-500: A Data Set for Building Digital Twins of over 2,000 People Based on Their Answers to over 500 Questions," Marketing Science, INFORMS, vol. 44(6), pages 1446-1455, November.
    21. Anne Lundgaard Hansen & Seung Jung Lee, 2025. "Financial Stability Implications of Generative AI: Taming the Animal Spirits," Papers 2510.01451, arXiv.org.

    More about this item

    NEP fields

    This paper has been announced in the following NEP Reports:

    Statistics

    Access and download statistics

    Corrections

    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:arx:papers:2606.08853. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    If CitEc recognized a bibliographic reference but did not link an item in RePEc to it, you can help with this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: arXiv administrators (email available below). General contact details of provider: https://arxiv.org/ .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.