Six degrees of Twitter

May 4, 2010

Its hard to be original. When I first signed up with Twitter, I liked it straight away, somewhat to my surprise, as I have never liked Facebook, and saw Twitter and Facebook as the twin spearheads of trivia. One pleasant Twitter surprise was the spare simplicity. Only does one thing. No ads. Strict 140 character limit. That kind of discipline can be good for artistic expression. An idea jumped into my head. Twitter Haikus ! Ten minutes later I realised that Twitter is swarming with haikus. Those of you on Twitter can just search on #haiku; others can see a constant stream of twitter haikus here and more nice stuff here .

Half an hour later I thought … how about a Twitter Novel … twenty first century Dickens ? However … I guess I don’t need to carry on.  For your enjoyment here is the stream from the currently ongoing Twitter version of Romeo and Juliet.

A week later I had noticed that when I got a new folllower, it was kinda fun clicking at random on their followers or followees, to see where I ended up. Pretty soon I found myself having “small world” thoughts – how many links separate any two Twitterers ? Could be a game, or a research project, thought I.

Scooped again. Through TechCrunch, to which I am mildly addicted, I learned of a report by a company called Sysomos which apparently gives us the answer. They analysed 5.2 billion connections and concluded that Twitter users are on average separated by 4.67 steps.

However, I couldn’t figure out what they really did. They say they analysed Twitter “friendships” but Twitter doesn’t have “friends”. It has followers and followees. They can’t be using followees, as any fule can search for a famous person and follow them. Bingo. One step. But they obviously don’t mean followers either, because the article refers to the phenomenon of coming across one of your followers after a few friendship links. This article suggests that a reasonable definition of “friend” on Twitter is someone you have directed a post to (with “@”) at least twice. But who knows if thats what Sysomos did ?

I have just sweated through two stressful peer review experiences. But getting cross with the Sysomos report reminded me how good the peer reviewed literature is. You couldn’t get away with not defining your methods at MNRAS.

Is this six-degrees-small-world just pop nonsense or is there really something to it ? Yes, a graph with six links can connect extremely large numbers of nodes. But that doesn’t tell us much about how real world networks actually develop. Don’t networks divide into islands ? There is a good wikipedia page on the subject and at least one quite famous serious paper ( Watts and Strogatz 1998) but this web article casts doubt on the whole thing.

I’d tell you the answer If only I had any energy left after solving quasars, marking exams, and reading nine million grant proposals.