The second speaker this morning at Digital Methods is Andreas Jungherr, who shifts our focus back to Twitter: he is interested in how we may use observations from this platform to understand what happens in society as such. What, if anything, may we read out of, for example, the patterns around an election which could help us predict the outcome of the election?
In the German election 2009, for example, Andreas found substantial activity around the Pirate Party, but this is an artefact of the specific demographics of Twitter in the country at the time rather than a sign of genuine pandemic interest in the party. In the same campaign, the volume of political news being shared during the campaign clearly shows the gradual growth of interest ahead of Election Day, and pinpoints key moments like debates and state elections in the run-up.
Looking at mentions of the key candidates highlights several other moments - though usually spikes related to media reporting, rather than party-organised campaign events. Similarly, in the context of major controversies, it is a handful of media-active events (the sharing of videos and satirical images) which generate most significant activity.
This does not augur well for any attempts to predict election outcomes from social media data. In 2013, even with growing social media use, the picture is skewed - the Pirates and the Euro-sceptic and not particularly Internet-affine AfD party are both overrepresented here. Further, there are massive day-to-day fluctuations in activity around the parties, making predictions from such unstable data almost impossible. What would be more interesting is to examine what offline events cause an online echo.