Published:

86 comments

Given that most people – including political journalists – don’t really understand either data or data science, it’s a field ripe for under- or over-reaction. Or, indeed, both at the same time.

Consider the way in which the Conservative Party under-appreciated what was required to build a proper targeting model in last year’s snap General Election. As I recounted in The Rusty Machine, ConservativeHome’s post-mortem dissection of the troubled Tory campaign, there was a disastrous assumption that the successful micro-targeting which worked so well in 2015 could be repeated without any run-up or advance preparation.

This was cargo cult thinking, akin to the Pacific islanders who believed in the wake of the Second World War that clearing a runway and lighting signal fires would cause American planes to come back, laden with food and materiel. In exactly the same way, CCHQ gathered the same people around the same table in the same building; but without the prior hard work of two years spent gathering data, analysing it, using it to hone messages and so on, they suffered a disappointment born of confusing cause and effect.

Part of the problem was that ‘data’ had been taken for granted. The right approach had been powerfully effective two years before, but a combination of complacency by non-scientist campaigners and the reluctance of the relevant industry to explain how it does what it does led some to erroneously assume the feat could be repeated at the flick of a switch. As just another tool, it would just work when needed, so what was there to worry about? Oops.

At the other end of the scale are the people who treat ‘data’ as a black box, onto which they can project all sorts of breathtaking and/or sinister powers. For them, it fulfils a god-of-the-gaps role, particularly when it comes to explaining or excusing defeats which they might otherwise find it too uncomfortable to accept as real, fair or deserved. How much easier life is to understand if you haven’t lost or failed to persuade people, but have instead been robbed through sharp practice of some sort.

Ironically, one example of this tendency came in an exaggeration of the data-based effectiveness of that very same Conservative campaign. At that precise time at which CCHQ’s plans were unravelling, due in part to data failures, there were a series of reports warning about their use of what were ominously termed ‘dark ads’. Particularly – the horror – adverts targeting people on Facebook with messages that they might find convincing.

The term ‘dark ads’ concealed, in reality, an updating of age-old campaigning practice: just as when you knock on someone’s door or ring them up you aim to give them reasons to vote for you which you believe will be compelling, political parties have started to target adverts towards voters’ interests and concerns. The mysterious nature of exactly what ‘data’ is, and widespread ignorance about how technology actually works, made it easy to describe Facebook adverts as if they were somehow underhand. “Campaigning in secrecy is enormously destructive of the basic principle of democracy”, claimed one expert – as if perfectly legitimate campaigning media such as targeted mailings or private conversations haven’t done exactly the same thing in a lower-tech way for ages.

The agenda of such reports during the election was pretty clear – to cast a degree of doubt on what many at the time still expected to be a sizeable victory for ‘Theresa May’s Team’. Fittingly, the source in The Observer quoted criticising such adverts was a left-wing campaigner complaining that they were being outbid in their attempts to broadcast their own targeted adverts on Facebook to the same voters. That hint that perhaps this wasn’t quite the sinister, one-sided magical victory machine that it was being made out to be is compounded by the fact that the Conservative campaign was, quite famously, not a success.

That stubborn fact appears to have been quietly forgotten by those who pushed the heavy implication that targeted advertising was somehow overthrowing democracy. Carole Cadwalladr wrote the original story warning about Tory ‘dark ads’ being used in the Delyn constituency, but she doesn’t seem to have returned to the topic of Delyn since the Labour majority there inconveniently grew 11 days after she published her warning.

Instead, the hype continues for the supposed powers of firms like Cambridge Analytica. It’s still unclear what, if any, role CA played in Leave.EU – the clown-car project of Arron Banks, not the actual Leave campaign – but The Observer still attempts to blur two campaigns who loathed each other into one by ominous references to “the Brexit campaign”. CA’s work with the Trump campaign is also displayed as evidence of their wizardry, while the same people try not to talk very much about their earlier work for Ted Cruz, whose campaign failed miserably even with the help of the data magicians.

In both of the above circumstances – under-stating and over-stating the power of new technologies – everyone could do with engaging a little more scepticism.

A campaigner who tells you that winning elections is all about gut-feel, the unquantifiable touch of the electoral guru, and that data is useless, is largely bullshitting to preserve his mystique, like a cabby who tells you Uber will never be any use because GPS will never match The Knowledge. Smart new campaigning tactics are useful, regardless of whether they’re disapproved of by traditionalists and vested interests as an ugly scientific incursion into a gentleman’s art.

Equally, an alarmist who makes grandiose claims that “micro-targeted ‘dark advertising’ on Facebook is a fundamental threat to democracy itself” (as The Guardian does today) is largely bullshitting to try to produce a silver bullet which explains away outcomes they dislike as somehow illegitimate. Targeting isn’t infallible, nor is it only available to those on one side of any particular battle – campaigns using it can and do lose, and it’s now normal for both sides of any given campaign to compete to use it most effectively. In other words: it’s a tactic, a means to deliver a message.

And then there are the people who are getting rich by doing all this targeting and analysing in the first place. When they tell their would-be customers that their service constitutes an overpowering force for dominating the minds of voters, they are largely bullshitting to sell their services. That’s part of their job. When the Conservative Party wanted to call a snap election, under the misapprehension that they could just hire in an accurate data operation overnight, none of the salesmen said “we won’t be able to do that in time, are you sure?”

There’s no requirement for anyone to undermine their own sales, of course; rather, buyers and observers ought to ask tougher questions and be slower to leap to convenient conclusions.

Christopher Wylie’s account of how CA built its model (told to Carole Cadwalladr, notably) is undoubtedly important, not least because it offers insights into the approaches such firms use and because of his revelations about the alleged illegal harvesting of the Facebook data of millions of American voters. But Wylie is also someone who hardly loses out when the world learns that his work is immensely powerful, that he is particularly stupendous at it, and that he (now) has a conscience as well. It’s right to report what he has to say, and particularly to investigate allegations of major data breaches, but those doing so have a duty to treat such claims with at least a touch of scepticism – particularly when he’s saying things the listener desperately wants to hear.

86 comments for: There must be a middle-ground between taking data-driven campaigning for granted and wild alarmism about its dangers

Leave a Reply

You must be logged in to post a comment.