When a behavior (e.g., product adoption) may spread through a social network, it can be advantageous to consider network structure when deciding where to seed that behavior. But what if the network is not yet observed and doing so is costly? Some recent empirical work employs methods that only rely on limited network information (e.g., by exploiting the friendship paradox). We consider these and other more sophisticated stochastic seeding strategies that likewise involve sampling information about the network. In the first part, we develop nonparametric methods for empirically evaluating stochastic seeding strategies, including by reusing existing data. This draws on and contributes to the policy evaluation and importance sampling literatures. We show that the proposed estimators and designs can dramatically increase precision while yielding valid inference. We apply our proposed estimators to two field experiments on insurance marketing in rural China and anti-conflict interventions in New Jersey schools. In the second part, we develop novel seeding strategies that come with guarantees about the loss compared with seeding using complete knowledge of the network. These algorithms make a bounded number of queries of the network structure and provide tight approximation guarantees for arbitrary networks. We test our algorithms on empirical network data to quantify the trade-off between the cost of obtaining more refined network information, and the benefit of the added information for guiding improved seeding strategies. (The first paper is joint work with Alex Chin and Johan Ugander, and the second is joint work with Hossein Esfandiari, Elchanan Mossel, and M. Amin Rahimian.)
Dean Eckles is a social scientist and statistician. Dean is the KDD Career Development Professor in Communications and Technology at the Massachusetts Institute of Technology (MIT), an associate professor in the MIT Sloan School of Management, and affiliated faculty at the MIT Institute for Data, Systems & Society. He was previously a member of the Core Data Science team at Facebook. Much of his research examines how interactive technologies affect human behavior by mediating, amplifying, and directing social influence — and statistical methods to study these processes. Dean’s empirical work uses large field experiments and observational studies. His published papers appear in Proceedings of the National Academy of Sciences, Journal of the American Statistical Association, Science, and other peer-reviewed journals and proceedings in statistics, computer science, and marketing. Dean completed five degrees, including his PhD, at Stanford University.