Counterfactual Delayed Feedback Learning
Keywords: Counterfactual, Delayed Feedback, HTE
Abstract: Estimation of heterogeneous treatment effects has gathered much attention in recent years and has been widely adopted in medicine, economics, and marketing. Previous studies assumed that one of the potential outcomes of interest could be observed timely and accurately. However, a more practical scenario is that treatment takes time to produce causal effects on the outcomes. For example, drugs take time to produce medical utility for patients and users take time to purchase items after being recommended, and ignoring such delays in feedback can lead to biased estimates of heterogeneous treatment effects. To address the above problem, we study the impact of observation time on estimating heterogeneous treatment effects by further considering the potential response time that potential outcomes have. We theoretically prove the identifiability results and further propose a principled learning approach, known as CFR-DF (Counterfactual Regression with Delayed Feedback), to simultaneously learn potential response times and potential outcomes of interest. Results on both simulated and real-world datasets demonstrate the effectiveness of our method.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13540
Loading