The twittersphere avoided making a bit of a mistake this morning. Wikileaks had obtained a new version of the BNP membership list, which they released (the BNP claim this list is a fake). Prior to release it was claimed that a peer of the realm was on the list and immediately post release that peer was named. Only it turns out it wasn’t him, someone who styled himself Lord with a very similar name was the man on the list. Fortunately the released list was detailed enough that this could be checked, someone had the wit to check before blindly repeating the name. Once they’d done this they started correcting the false rumour (in what looks like quite a vigorous manual effort). It’s worth noting here that the fact-checker appears to be a trained journalist.
But it could so easily have been very different. It could have been very difficult to establish the rumour was false, it could have been that the diligent fact checker stopped to finish his cup of tea before tweeting his correction, the rumour could have been re-tweeted by someone with many followers. All of these things could have happened but didn’t, will this be true the next time?
On the plus side, twitter rumours do appear to be traceable back to source and it’s very easy to find the individual rumour-mongers and put them right. This is certainly true for non-malicious rumourmongering (that’s to say where people have not made a special effort to propagate a rumour, nor hide their tracks).
There is a scientific link here, modelling of all sorts of networks has long been a respectable scientific field. For example, there’s Per Bak’s forest fire model and work that follows on from there. More recently there’s been work focussing more explicitly on computer networks and social networks. To a physicist Twitter represents an example of a simple system which has nodes (with ingoing and outgoing links) and messages that are propagated between the nodes. The nodes could be trees in a forest and the thing passed could be fire, or the nodes could be computers in a network with the message being network traffic; the nodes could be scientific papers with the messages citations of other papers. The physics doesn’t care about the detail of these things, it cares about a small number of parameters in the system: how many links in and out of a node? What’s the probability of a message being transmitted from one node to the next?
So there’s an interesting bit of network analysis to do here. How fast can a rumour propagate on Twitter? What fraction of people refrain from tweeting a false rumour to stop it propagating? What’s the best way to squash a false rumour?
Having watched the no doubt frenzied activities involved in squashing today’s rumour. One useful tool would be an automated rumour-quashing robot. It would search for tweets containing the rumour (probably based on a manually selected keyword) and tweet the originator with a rebuttal.
Think before you tweet!