Dynamic Interference Modeling For Estimating Treatment Effects From Dynamic Graphs
Keywords: Causal Inference, Treatment effect estimation, Dynamic Graph, Interference
TL;DR: We propose new approaches to address a challenging issue: Estimating treatment effect under dynamic interference.
Abstract: Estimating treatment effects can assist decision-making in various areas, such as commerce and medicine. One application of the treatment effect estimation is to predict the effect of an advertisement on the purchase result of a customer, known as individual treatment effect (ITE). In online websites, the outcome of an individual can be affected by treatments of other individuals, as people often propagate information with their friends, a phenomenon referred to as interference. Prior studies have attempted to model interference for accurate ITE estimation under a static network among individuals. However, the network usually changes over time in real-world applications due to complex social activities among individuals. For instance, an individual can follow another individual on one day and unfollow this individual afterward on an online social website. In this case, the outcomes of individuals can be interfered with not only by treatments for current neighbors but also by past information and treatments for past neighbors, which we refer to as \emph{dynamic interference}. In this work, we model dynamic interference for the first time by developing an architecture to aggregate both the past information of individuals and their neighbors. Specifically, our proposed method contains a mechanism that summarizes historical information of individuals from previous time stamps, graph neural networks that propagate information about individuals within every time stamp, and a weighting mechanism that estimates the importance of different time stamps. Moreover, the model parameters should gradually change rather than drastically because information of every individual gradually changes over time. To take it into account, we also propose a variant of our method to evolve the model parameters over time with long short-term memory. In our experiments on multiple datasets with dynamic interference, our methods outperform existing methods for ITE estimation because they are unable to capture dynamic interference. This result corroborates the importance of dynamic interference modeling.
Primary Area: causal reasoning
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.
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: 9375
Loading