I’ve been playing around with Twitter.com’s new social graph functions in their API, and for a quick experiment pulled the network formed by all twitterers with 30,000 or more followers (the “twitterati” elite). When I get a chance to do some more in-depth network analysis, I’ll follow up to this post, but here’s a quick look at the network. I identified 45 nodes (twitterers) with 30K+ followers, and 458 edges (follower links) among them. The following graph (sociogram- link opens to full size image) shows their connections, with nodes colored relative to each node’s betweenness centrality (directed).
Betweenness centrality usually connotates control or influence over network resources - which, of course, would be “tweets” in our example. Offhand, I’d have to interpret this as the relative ability of a twitterer to decide whether or not to retweet (pass on) a juicy bit of news or information to the rest of the network.
Of course, we should look at access too - the following graph colors and ranks each node by its relative closeness centrality which usually indicates relative connectedness and access to network resources. I’d interpret this as showing who in the network is most likely to hear about a given bit of news tweeting its way throughout the network.
For betweenness, our top twitterer was stephenfry, followed by zappos. For closeness, our leaders were mchammer and again zappos. Overall, the same list of 8 or 10 twitters lead on both measures.
Keep in mind that the network graphs above and measures I’ve discussed only include connections to other 30K+ twitterers; if we included the millions of followers’ links outside of this core group we wouldn’t really be able to make much visual sense of it. That’s a bigger project that I may tackle some other time!
Some other quick observations: it’s a tight network with no natural clusters. The nature of twitter is such that there really isn’t much cost for following everyone else of note, so pretty much everybody in this core is a degree or two away from each other.
These charts were generated using the wonderful, free yed graphing software.
Previous Post: 
Next Post: 