From Santa Fe Institute Events Wiki
Statistical Inference for Complex Networks Workshop, December 3-5, 2008, Santa Fe NM
Cosma Shalizi (homepage)
Homophily, Contagion, and Confounding: Pick Any Three
People close to each other in social networks tend to resemble each other in many ways. Sometimes they were similar to each other before, and that's why they're close (homophily); sometimes being close made them more similar, because one of them influenced the other (contagion). I will discuss the difficulties of distinguishing these two processes, especially with snap-shot data and with unobserved characteristics, and explore how the presence of either process can systematic subvert (or "confound") causal inferences about social mechanisms.