There are two kinds of social graphs in the world, explicit and implicit. The explicit ones are the most rare and give their owners disproportionate market values.
The two I am thinking of are Facebook (private market value around $40B) and LinkedIn, with a public market cap of ~$6B. What makes their graphs explicit is that two individuals deliberately and mutually describe their connection with one another.
Implicit “soft” graphs include Twitter and Google (via Gmail). In Twitter’s case, the opt-in is only one way; only one user declares interest in another, and it need not be reciprocated. Gmail’s graph is inferred from your behavior – they know with whom you’ve interacted, but you haven’t explicitly told Gmail about your relationship with a person.
Explicit graphs can be mined more easily, since they begin with hard data. As I’ve mentioned in the past, it’s data, not algorithms that will be hard for competitors to replicate in the future.
So, explicit graphs should be understood as truly personal, and social. Implicit graphs should be understood as a compilation of our interests.
Knowing what we like is old science, going back to the advent of credit cards. Who we like is new information. Thus the valuations.