Social network analysis: “The Friendster Autopsy”

A group of researchers has provided an autopsy of Friendster:

Friendster was once the hottest thing in social networking. Google wanted to buy it for $30 million back in 2003, but — burdened by technical glitches and a more nimble competitor in Facebook — it was pretty much dead in the U.S. by 2006. That said, it trudged along for a few more years, helped by a relatively strong following in southeast Asia. Then, around 2009, a site redesign crushed it…

What they found was that by 2009, Friendster still had tens of millions of users, but the bonds linking the network weren’t particularly strong. Many of the users weren’t connected to a lot of other members, and the people they had befriended came with just a handful of their own connections. So they ended up being so loosely affiliated with the network, that the burden of dealing with a new user interface just wasn’t worth it.

“First the users in the outer cores start to leave, lowering the benefits of inner cores, cascading through the network towards the core users, and thus unraveling,” Garcia told us during an online chat…

The researchers describe heart of successful networks in terms of what that they call K-cores. These are subset of users who not only have a lot of friends, but they have “resilience and social influence,” Garcia says. As these K-cores disintegrated, the whole Friendster thing fell apart.

This sounds like a classic social network analysis: are there enough strong bonds in a large network to keep it together? As they note, it is not just about the number of users involved but rather the relationships and interactions between key groups that have an outsized influence on the entire network.

I suspect Facebook (and other social network people) has some idea about this idea already. What I would want to hear about is how then Facebook helps keep these “k-cores” or hubs or key nodes together. Are the algorithms for news feeds geared toward these people or groups? How are things tweaked to help these k-cores grow and continue to have an outsized influence?

Why “Your Facebook friends have more friends than you”

Here is an overview of an interesting quirk on Facebook: your Facebook friend likely has more friends than you do.

It’s just the digital reflection of what’s known to sociologists as the “friendship paradox.” In 1991, sociologist Scott Feld found that, generally speaking, any person’s friends tend to be more popular than they are. The reason, he said, is fairly simple: people are more likely to be friends with someone who has more friends than someone who has fewer friends.

This is true on Facebook as well, the study found. A small number of people are isolated and don’t appear on many lists, but popular people show up again and again.

Another interesting result of the study finds that Facebook users tend to get more messages, friend requests, likes and photo tags than they give, pointing to the existence of a few Facebook “power-users” driving the site’s activity.

Keith Hampton, a professor at Rutgers University and the lead author of the report, said that power users make up around 20-30 percent of Facebook’s users, and that there are three specialties within these power users. Some users send a lot of friend requests, while others most frequently “like” posts and pictures. A third kind of power user tends to make a lot of photo tags.

If you put this in social network terms, there are certain people who are nodes in the Facebook network. These nodes have more friends and are centers of information, comments, pictures, likes on Facebook, between different groups and users.

If we know this is how the world works, you could imagine how this information could be put to use. Perhaps Facebook puts information from these nodes more often in your news feeds. Perhaps marketers hope to specifically target these people as they can have a wide reach. Perhaps other users could look to connect with these nodes, knowing that these people could help them get to information (like jobs? social events?) that a less connected user could not.

I was thinking about this as I was trying to explain network behavior to some students in class recently. Are users of Facebook aware of where they fall within their networks, meaning are they nodes themselves or far from the center of activity? If they are aware of this, does this change their behavior? Would it be beneficial for Facebook to show users where they fall in their network with the chance that it might boost their online activity levels?