The structure of social networks reflects important social processes, such as status, communities, and inequality---but separating signal from noise is nontrivial in social data. Statistical inference provides one point of entry: unpacking the assumptions behind random graph models, we can understand what tools will help us uncover meaningful social structure from noise. The empirical design of networks problems, however, is often overlooked but fundamentally changes the space of problems available. I will discuss single-graph (N=1) problems and some of the surprising challenges behind many-graph (N >> 1) problems. Using random graph models, statistical inference, and empirical design, I will present a brief overview of how to understand networked processes in real systems---from families and firms to food webs and Facebook. Speaker(s): Abbie Jacobs (University of Michigan, School of Information and Complex Systems)