Causal Inference in Networks: Community Informed Design

Networks are becoming increasingly prevalent in experimentation. As such, methods need to be developed to account for the potential correlation and interference that may occur in networks. Social networks frequently exhibit homophilous community structure, meaning that individuals within observed or latent communities are more similar to one another. This motivates our development of community aware experimental design. It is probable that connections between individuals likely flow along within community edges rather than across community edges. We develop a design that conditions on communities and show that as the community detection problem gets more difficult or if the community structure does not affect the causal question, the proposed design maintains its performance.