Last week we presented a talk at the Data & Politics conference in Berlin. It was a fascinating day, with a wide range of talks covering the Obama elections, how the German political ecosystem is dealing with fake news, bot etc, how data analytics is used in politics (cue lots of mentions of Cambridge Analytics) and more – this is a good storify sequence summary. These are the 3 key points we discussed.
- What we had seen happen in the Brexit and US elections,
- What we could see happening so far in Germany
- Some thoughts about using data analytics in election strategies
For what happened during Brexit and the US elections, we have previously covered that topic in this blog post. Suffice to say that the winners were visible on social media, but for a variety of reasons did not seem to be seen on polling data and, to an extent, the UK and US political pundit industry seems to “learned nothing and forgotten nothing” from this so far.
As to Germany, the things we noted in our talk was that “Germany is different” as far as use of social media data for election prediction is concerned, for the following reasons:
- Lower use of social media in the population overall than UK or US, and there is still a greater trust of the mainstream media
- Far less participation on social media by political parties, even (especially) the far left and far right – whereas in UK and US the “edge” groups were the heaviest users by far, both for beter economics and getting their message through to their followers
- Strong regulatory steps taken to manage the use of bots, fake news, spamming and so on after seeing what happened in the US.
Right now our models are not showing much conclusive as it is very early days, though we did show the audience some emerging evidence of a shift to the right of the main centre-right parties (see diagram at the top), which also happened in Holland and we believe will be a feature in EU elections this year (see our ingoing EU hypothesis model over here).
With respect to campaign lessons for future social media data analytics, we looked at a number of areas in the Brexit and US elections in terms of:
- Using social media data analytics to monitor elections – what we have learned works and doesn’t
- What social media approaches worked in both elections – and what didn’t
- What processes and practices underpinned the leading social media campaigns
- Conversely, what could be learned from those that didn’t work
- How do Fake News, Bots, Trolls, Spammers etc impact elections and what can be done to neutralise them.
With respect to using social media analytics, what became clear is they are useful, but to optimise usage it’s necessary to have an overall integrated approach to analytics, mainly to focus effort and message. Also, it must be integrated into an overall strategy. Where things seems to have gone right was when analytics was placed very high up in the strategic decision making, above more “touchy feely” approaches. (For a forthright view of Brexit analytics I recommend Dominic Cummings various blog posts – He architected Vote Leave’s ramp up from no organisation to winning in 10 months, and there is no beating around any bushes there).
As to what Social Media approaches worked well in both elections, a number of common factors emerge. Firstly, monitoring needs to be constant and responses fairly fast to influence the unfolding of the never-ending story. Secondly, as far as messaging goes emotions trump (sic) facts every time. Thirdly, Test, test and test the message again – and once you have it right, barrage the channels. Quantity has a quality all of its own. Trump used a “Wildean” strategy (from Oscar Wilde – the only thing worse than being talked about is not being talked about) in which messaging was put out to dominate the newsfeeds, driving out other stories (an garnering a huge amount of free advertising.)
Common processes that seemed to underpin the successful campaigns were around ensuring speed of action and response, driving testing and message optimisation (it seemed to be as much about stopping interference from more traditional advisors as using analytics) and a relentless focus on getting the message out. Also, putting analytics very high up in the decision making process was a common factor. Both in Brexit and the Trump campaigns we see reports that they had to continually withstand attempts to change analytic driven decisions in order to fit with recommendations by the more traditional approaches.
The biggest problems in the less successful campaigns seems to be primarily around simply not being willing to believe and act on what the data was telling them (due to Confirmation and other biasses), and also being slow to respond to moves by the opposition. Part of this was that their campaigns by and large were focused on more traditional approaches, and the online analytics was low in the decision altering pecking order.
With Bots, Fake News, Trolls etc it seems that there has been as much media ink spilled about them as about the elections. Their impact is to magnify “your” message, drown the opponents message, motivate your supporters and cow the opponents’ by making them feel more and less “with the zeitgeist” respectively. Bots drive volume by broadcasting, Fake news by viral spread and Trolls reach. To neutralise them, one has to:
- First , understand their message and who it is aimed at.
- Second, understand their dynamic
- Third, be able to put out messaging to counter them – here, one interesting technique used in Climate change arguments is Innoculation – to put a sample of the opponents argument in yours
It was hard to get too much across in 30 minutes unfortunately, but we’ve been asked to write a more detailed briefing paper of what we learned, on this last topic of campaign strategy especially, so if anyone wants to get a summary copy, send us an email at email@example.com
Overall, it was a pleasure and a privilege to have been invited, and we learned far more I think than we could give back, and our thanks to the organisers – Initiative D21 – for allowing us to participate, and enjoy a bit of Fruehling im Berlin as well.