Brexit: The Uncivil War showed us how the EU Referendum was won with Data ScienceAshley NewBlockedUnblockFollowFollowingJan 11Something that wasn’t discussed as much as it should have after the release of Channel 4 / HBO’s Brexit The Uncivil War last week was how the extensive use of data science was the main reason that the Leave campaign won against all the odds.
In this article, I discuss how data science is changing the political landscape.
Traditional polling at the time of the 2016 EU referendum was almost unanimously showing that Remain was most likely going to win.
Even leave proponents like Nigel Farage were posturing that they would continue the fight to leave if defeated (and for another referendum) on the lead up to the result, on the basis that the result would be close.
The result when it arrived, came as a shock to everyone, politicians and political commentators alike.
How traditional polling got it wrongBeing a staunch advocate of remain myself, I remember watching the result from my parents’ home in the South of France late into the night, going to bed at about 3am with the colour drained from my face.
Upon waking, my worst fears had been confirmed and we were leaving the EU.
My partner cried upon hearing the news, born in Germany but a British Citizen, she had gone to one of the only EU schools in the UK.
She felt like her identity was being stripped away from her.
The next day the political fallout was massive and is something we are still dealing with today.
What was widely discussed after the result was how the polls got it all wrong, something that had happened repeatedly in recent years and happened again 5 months later when Donald Trump was elected into the White House.
Which again shocked every serious political commentator, and according to Michael Wolff, even Donald Trump himself.
Even the result of the 2017 UK election which resulted in a hung parliament was unexpected.
Not long after this, I wrote how traditional polling was dead, highlighting that the only poll that predicted a hung parliament result correctly was an experimental YouGov poll that used more modern data science techniques.
In short, traditional polls regularly ask a statistically relevant amount of the people (around 2,000) their opinions on various matters.
This selection is representative of different demographics within the electorate, for example by age, gender and income.
The pollsters then gauge fluctuations in opinions across the country by these demographics.
YouGov’s experimental poll used more modern techniques, they conducted larger interview samples (of around 50,000 people) without caring much for getting a fair demographic spread.
They then used multi-level regression and post-stratification to identify clusters of voters within the population.
Essentially breaking down the electorate into far more granular demographics and weighting the fluctuations based on that.
While this is an important topic in and of itself, the question that is more important here is not how traditional polling models got it wrong but why they got it wrong.
What Brexit: The Uncivil War showed us about the use of Data Science in the EU ReferendumLast week, Channel 4 aired their much talked about Brexit: The Uncivil War.
For those that haven’t seen it yet, it is a feature-length drama based on accounts from extensive interviews with key players of the events from the moment the Brexit referendum was announced until the result.
The drama, written by James Graham, follows Benedict Cumberbatch beautifully playing Dominic Cummings, a controversial political strategist and how he masterminded the Leave campaign.
Interestingly it portrayed him as a person who was purely focused on the job of winning the referendum rather than what the outcome would look like and whether he believed it was right.
His political objective seemed to be that he didn’t agree with the way the British political class was running Britain and wanted to disrupt it.
Others may call this bringing down the establishment, or in America, draining the swamp.
He saw the Brexit referendum as the vessel to fulfilling this objective.
If all accounts are accurate, Brexit: The Uncivil War is supposed to be an unbiased portrayal of how the Leave campaign won the EU referendum.
Since The Uncivil War has been released, most critics of Leave seem to be focused on two things; that the drama whitewashes over how the Leave campaign broke electoral law on spending.
And how the campaign relied on racist and anti-immigration sentiment to win.
What hasn’t been more widely discussed is how the use of data science led the whole campaign and is what gave Leave an edge over the Remain campaign, and quite possibly won it.
And more importantly, what this means for the future of politics.
What Dominic Cummings points out to all the critics is that he wasn’t deciding the messages, he let the data decide.
That’s how they got to the core messages that would hit the electorate hardest; that Turkey was imminently joining the EU and that the £350 million we send to the EU should be used for an underfunded NHS.
Both of which haven’t transpired or proved accurate.
In Brexit: The Uncivil War, we see Dominic meet Zack Massingham, the founder of AggregateIQ, a Canadian political consultancy and technology firm.
The Leave campaign is supposed to have spent £3.
5 million on AggregateIQ, 50% of its £7 million of the allowance, of which it breached by spending £7,449,079.
That is a rather big percentage of their allowance, so they must have banked a lot of this firm.
Furthermore, Brexit: The Uncivil War sinisterly portrays Dominic’s meeting with Zack as him saying that they could tap into “3 million extra voters the other side have no idea exist”.
Dominic is then seen saying he’s going to “Hack the system [electorate]” to some flabbergasted old guard politicians.
In Brexit The Uncivil War, it shows us how Dominic had a team of Data Scientists from AggregateIQ, put away in a shaded side room, as a sort of separate project to the rest of the Leave campaign.
The first thing they are seen doing is create a campaign to learn what different clusters of the electorate thought about different issues around the EU.
They did this by giving people the chance to win an aggregate bet, one that would be statistically almost impossible to win (a 1 in 5,000,000,000,000,000,000,000 chance).
Then they used the finding of this research to create behaviourally micro-targeted messages to different segments in the electorate.
It was found that during the referendum, they paid for over 1 billion Facebook adverts, with various pro-leave messages.
For example, segments of less “racist” voters may have received pictures of Boris Johnson stating, “I’m pro-immigration, but above all, I’m pro-controlled immigration”.
While others received messages such as “TURKEY HAS A POPULATION OF 76 MILLION.
TURKEY IS JOINING THE EU.
Older people were targetted with adverts on how the money we send to the UK could be spent on the NHSPeople who were animal lovers would have received adverts like theseAnd supposed tea lovers or patriots would have received messages like theseIf they clicked on the relevant advert, they would then receive a horde of continuous adverts around the same topic, reinforcing this point of view.
In some reports, it was suggested that the majority of the advertising spend was spent in the final week before the EU referendum vote.
In Brexit The Uncivil War, the Leave campaign team are even seen asking why the adverts are not being talked about on TV and in news cycles, to which they figured the other side had no idea that this was going on.
On 23 June 2016, 17,410,742 people voted to leave the EU against 16,141,241 who voted to remain.
Why is Data Science changing the political landscape?Of course, anyone who even lightly follows the western news cycle will have likely heard of the Cambridge Analytica scandal.
Cambridge Analytica was set up and run by Robert Mercer (A billionaire AI expert) and Steve Bannon (An alt-right media Svengali), Donald Trump’s financial backer and campaign chief executive.
Cambridge Analytica whistle-blower Christopher Wyle has highlighted how AggregateIQ has close links to Cambridge Analytica.
This scandal showed us that through the misuse of Facebook user data they could psychometrically profile the entire electorate.
Using a viral personality app called myPersonality they were able to cross-reference personality types against what people had liked on Facebook to profile people with a high degree of accuracy.
The data that was harvested and used to feed the algorithms that profiled the electorate became a vital weapon in the EU referendum and US election.
Wyle even described how Steve Bannon called these tools weapons in his “psychological warfare”, with the objective to bring about nationalist populist movements all over the world.
While some people may object to this form of political manipulation and see this as scary.
On a positive, Cambridge Analytica has now closed down due to the intense political pressure following the fallout from the election of Donald Trump, Facebook have cracked down on the third-party data leak issue that enabled these data science tools to be built, and GDPR protection of European citizens means you have more control over how your personal data is used now.
Next time this happens, it probably won’t be as easy or go as unnoticed.
Unfortunately, other firms like Cambridge Analytica still exist and will continue to do so while there is sufficient demand.
Politicians spend a serious amount of money on campaigns, of course, they are going to use the latest techniques and technologies to try and win.
Expect this to be the new normalYou may baulk at this fact if you haven’t been happy with the result of the EU referendum or the election of Donald Trump.
This is just because the national populist movements got there first.
Since we have moved into the digital age, we share more and more data about ourselves and create more avenues for information to reach us.
Modern marketing and advertising have become increasingly about creating relevant, personalised messages to continually more granular target segments, politics is just another client in this market.
Data science will find ways of piecing our information together to create profiles of us to ensure the right messages reach us.
Political campaigns will find ways to research our emotional states about particular issues and create messages that elicit a particular voting response in us.
Platforms will continue to exist that will enable those targeted campaign messages to reach us.
If Brexit: The Uncivil War showed me anything, it was the failure of the Remain campaign to embrace the latest tools and technology that were available from the explosion in data science.
They too could have been targeting the voters the Leave campaign were targeting and the ones that didn’t bother to go out and vote.
Like traditional pollsters, they used the same old techniques that have been used for decades and failed to get it right.
It’s time for politics to step into the 21st CenturyWe now live in what’s called the post-truth era, it has become less about arguing the pros and cons in a rational debate, but about launching “grenades” of information (no matter if they are true or false) that resonate with the electorate in an emotional way which then requires the opposition to “put out fires”.
This can be a two-way street, I am certainly not proposing that lying is acceptable, it would be perfectly possible to build better fact-checking tools that would enable real-time checks on false information in the future to help deal with that.
But I am saying campaigns on every side of the argument need to learn from Cambridge Analytica and AggregateIQ that the old way in politics is gone, data science is here to stay.
The winner is the one that stays ahead in the game.
You can follow me on Twitter @NoWayAsh Please feel free to share your feedback, comments and opinions with me.