Seeding a Simple Contagion

Abstract

This paper introduces a methodology for selecting seeds to maximize contagion using a coarse categorization of individuals. Within a large and flexible class of random graph models, I show how to compute a seed multiplier for each category---the average number of new infections a seed generates---and I propose randomly seeding the category with the highest multiplier. Relative to existing methods for targeted seeding, my approach requires far less computing power---the problem scales with the number of categories, not the number of individuals---and far less data---all we need are estimates for the first two moments of the degree distribution within each category and aggregated relational data on connections between individuals in different categories. I validate the methodology through simulations using real network data.