CISE Seminar: Mohsen Bayati, Stanford University

Causal Message Passing: A Method for Experiments with Unknown and General Network Interference
Randomized experiments are a powerful methodology for data-driven evaluation of deci- sions or interventions. Yet, their validity may be undermined by network interference. This occurs when the treatment of one unit impacts not only its outcome but also that of con- nected units, biasing traditional treatment effect estimations. In this talk, we introduce a new framework to accommodate complex and unknown network interference, moving beyond spe- cialized models in existing literature. Our framework, which we term causal message-passing, is grounded in a high-dimensional approximate message passing methodology and is specifi- cally tailored to experimental design settings with prevalent network interference. Utilizing causal message-passing, we introduce a practical algorithm to estimate the total treatment effect, defined as the impact observed when all units are treated compared to the scenario where no unit receives treatment. We demonstrate the effectiveness of this approach across several numerical scenarios, each characterized by a distinct interference structure.

The talk is primarily based on Causal Message Passing: A Method for Experiments with Unknown and General Network Interference, (2023), that is joint work with Sadegh Shirani. Should time allow, we will also discuss recent findings from joint work with Yuwei Luo, Will Overman, Sadegh Shirani, and Ruoxuan Xiong.