Causal Network Motifs: Identifying Heterogeneous Spillover Effects in A/B Tests

by   Yuan Yuan, et al.

Randomized experiments, or "A/B" tests, remain the gold standard for evaluating the causal effect of a policy intervention or product change. However, experimental settings such as social networks, where users are interacting and influencing one another, violate conventional assumptions of no interference needed for credible causal inference. Existing solutions include accounting for the fraction or count of treated neighbors in a user's network, among other strategies. Yet, there are often a high number of researcher degrees of freedom in specifying network interference conditions and most current methods do not account for the local network structure beyond simply counting the number of neighbors. Capturing local network structures is important because it can account for theories, such as structural diversity and echo chambers. Our study provides an approach that accounts for both the local structure in a user's social network via motifs as well as the assignment conditions of neighbors. We propose a two-part approach. We first introduce and employ "causal network motifs," i.e. network motifs that characterize the assignment conditions in local ego networks; and then we propose a tree-based algorithm for identifying different network interference conditions and estimating their average potential outcomes. We test our method on a real-world experiment on a large-scale network and a synthetic network setting, which highlight how accounting for local structures can better account for different interference patterns in networks.


page 1

page 2

page 3

page 4


Exploiting neighborhood interference with low order interactions under unit randomized design

Network interference, where the outcome of an individual is affected by ...

The Local Approach to Causal Inference under Network Interference

We propose a new unified framework for causal inference when outcomes de...

Causal Inference under Network Interference with Noise

Increasingly, there is a marked interest in estimating causal effects un...

Causal Inference Under Interference And Network Uncertainty

Classical causal and statistical inference methods typically assume the ...

Neighborhood Adaptive Estimators for Causal Inference under Network Interference

Estimating causal effects has become an integral part of most applied fi...

Causal inference with misspecified interference structure

Interference occurs when the potential outcomes of a unit depend on the ...

Randomization Test for the Specification of Interference Structure

This study considers testing the specification of spillover effects in c...

Please sign up or login with your details

Forgot password? Click here to reset