The Grauniad Dabatlog has produced a rather fancy network visualisation of the sources cited in Anders Behring Breivik’s personal manifesto/horse-shit compendium. This is great as I now don’t need to worry that I perhaps should have made one. It’s very pretty and you can click on stuff, and see that some of the sources are thinktanks and some of them are newspapers, and well, it’s very pretty and you can click on stuff. It also comes with a piece by Andrew Brown reprising his “Don’t be beastly to the creationists!” shtick but with Melanie Phillips, for some reason.
Unfortunately it’s almost completely intransparent, and gives little indication of what data is being visualised or on what basis, and there is really no obvious conclusion to draw from it. But did I mention pretty and click? If forced to take a view, I would reckon that the underlying data is probably a matrix of which sources appear together with others and the layout algo is a force-directed graph (aka the default in pretty much any visualisation toolkit), probably weighted by appearance count. There’s some sort of proprietary metric called “linkfluence” which appears to be given by(indegree/outdegree)*len(neighbourhood) or words to that effect.
As a result, the only information I got from it was that he linked to Wikipedia, the BBC, and big news sites a lot. Well yes; Wikipedia, bbc.co.uk, etc, generate a hell of a lot of web pages and people read them a lot. Obviously, to say the least, you need to normalise the data with regard to sheer bulk, or you’d end up concluding that Google (or Bing or Yahoo) was his inspiration because he did a lot of web searches, or that he was a normal man twisted by SMTP because he used e-mail.
In fact, I thought they actually did that until I realised that RSS.org is about the other RSS, the Indian extreme-right movement, not the popular Internet syndication standard. Harrowell fail. Anyway, it does show up rather nicely that the groups “European nationalists”, “Counter-Jihad”, and “American Right-Wing” overlap. However, I feel there’s something missing in the characterisation of MEMRI and various other sites as just “Think Tanks” as if they were just like, say, IPPR.
Also, an emergent property of the data is that there is an Axis of Barking running vertically through it: the nearer you are to the top of the diagram, the more extreme and crazy. MEMRI, FrontPage, Gates of Vienna, Melanie Phillips are near the top; the Wikipedia article on the Russo-Turkish War of 1878 is at the bottom. And the MSM is somewhere in the middle. (Although I do wonder if they allocated the sources to groups before or after running the force-directed graph.)
It seems to be one of those command the exciting world of social media with just one click! things.
Anyway, upshot. I want to avoid Project Lobster producing a diagram like this one. It’s too impressionistic and fluffy and reliant on basically aesthetic reasoning. (I think we’ve had this point before.) Of course, that’s partly the difference between the underlying data sets; it was at least thinkable if unlikely that there would be no grouping in Breivik’s sources, while presumably political lobbying is nonrandom and subject to intelligent design.