In December 2005, I joined Facebook. At the time, it was still limited to college students. Your status updates were only visible on your wall, and your friends had to come to your wall in order to post messages. There were silly groups to join, and poking people and throwing sheep were popular interactions. This was long before games, apps, pages, and marketing on Facebook. In the time since, I have been very vigilant about only adding friends that I actually know. Therefore, it is likely that many of my Facebook friends also know each other in real life, leading to a dense network.
Figure 1a: Graph Statistics
Figure 1b: Raw Network
The people with whom I share the most friends, and therefore have the highest Degree scores are my parents, my spouse, and my siblings. I was surprised that my parents (Mom and then Dad) beat out my husband in terms of mutual friends. Considering that my husband and I attended high school together and that he knows many of my friends, I believe this says something significant about Facebook use. My husband rarely uses Facebook, while my parents use Facebook quite a lot, with my mom being the more frequent user. As such, my parents have a greater desire to seek out connections with people that they know than my husband does. It is not surprising that my husband has more friends in common with me than my siblings, as he and I are of a similar age, while my siblings are younger and have their own pools of high school/college/work friends.
Looking at the initial graph of alters, it seems that there are four groups of Facebook friends in my network (see Figure 2). Using the Clauset-Newman-Moore method, six clusters of friends were identified. Upon inspection, these groups reflect different aspects of my life: high school, college, work, and family (my side, husband’s side, and ex’s side). Five of these clusters are linked together without me (see Figure 3). My husband is affiliated with his family group and is the bridge spanner between that group and the four groups to which it is attached. The lone isolate cluster is made up of my coworkers. This is not surprising, as my coworkers are not likely to know people from my personal life.
Who are the most popular, important, and influential people in these groups? These friends have high degree, betweenness centrality, and eigenvector centrality scores. High degree means that we share many friends, and are likely close to each other. High betweenness centrality reflects awareness of many different friends, and also indicates a close friendship in real life. Influence is indicated by high eigenvector centrality scores, where the person is friends with many popular people. For each group, these are the people to target with messages or invitations to events in order to maximize dissemination and participation among the members of the group.
The original approach was to look at each group individually, but it was difficult without the connections between groups. This led to looking at the network as a Friend Wheel (see Figure 4) and then as a Friend Pinwheel (see Figure 5). Looking at the Friend Pinwheel, the key players within the network are larger and more centrally-located than the others. The two large nodes closest to the center represent my mother and my husband. The next closest node is my college roommate, who has a much lower Degree than either my husband or my mother, but is friends with my mother, as well as many of my college friends. It is interesting that a cluster of vertices came forward from the High School group, rather than a single person. These friends were all in Marching Band, and are friends with each other, too. Using the Friend Pinwheel also allows the inter-group connections to be seen. The most interconnected group is Family, followed by High School Friends. Therefore, news and information will travel quickly through these dense networks.
If you would like to contact me, please see my LinkedIn profile.