Additionally, I think the public is now realizing how powerful the friend/follow metadata graph is. It's pretty easy/robust to infer traits about people for whom you don't have explicit information. Alan, Betty, and Charlotte filled out the quiz and all live in Florida; Dave has mutual friendships with Alan, Betty, and Charlotte; Dave probably lives in Florida. You can do this recursively and don't need to have concrete information about the majority of members in a graph in order to make pretty strong probablistic inferences about the traits of a large graph.