There’s a stark contrast between the social graphs we have in real life and those we build online. To comprehend the difference, we need look no further than our own day-to-day interactions with the people around us, minus the computer, mobile device, or any other tool that connects us to the online world.
Our online social graph is a deeply interconnected web of relationships, where the majority of our conversations involve people spanning many different social groups, including our family, friends, peers, etc. The graph generated from these types of interactions doesn’t include the concept of clustering.
This is where the real-life social graph differs. In our day-to-day interactions, we consciously form clusters for the people in our lives, interacting with each group in different ways and through various methods. We develop social boundaries for these clusters, sharing different pieces of ourselves with each group. For instance, you may be more candid with your friends about a physical relationship you are involved in than you are with your immediate family members.
Unfortunately, user clustering is one area where the online social graph is lacking. Since different criteria—anything from physical location to topics of interest—determine a person’s status in a cluster, it is incredibly difficult to programmatically categorize his relationship to someone else and maintain any level of privacy or security.