There’s a very long article that I’ll attempt to summarise so you don’t have to sacrifice one of your limited supply of hours here on Earth. It’s fairly complicated but almost certainly quite important. And like most things these days, it’s got a whole bunch of Trump in it.
Cambridge Analytica is a company that claims to have played an instrumental role in both Brexit and Trump. Their prowess at swinging elections is built on a number of innovations.
First, in psychometrics. Back in the 1980s, university researchers identified key personality traits that enabled them to slice and dice the great sea of humanity. Apparently, we all display more or less of the Big Five traits:
And for once – says the copywriter who’s gnashed his teeth on more than a few naming projects – the damn acronym is both neat and meaningful. I bet there was some serious whooping and high-fiving going on when that was first worked out.
Anyway, back to the spooky stuff. These OCEAN personality traits were a fairly innocuous semi-academic exercise (used mainly for assessment and recruitment of the ‘right kind of people’) until they were accidentally weaponised on Facebook in 2008.
Before Facebook, supplying the info meant filling in a long and pretty weird questionnaire. But researchers created a MyPersonality app on Facebook that made it simple to answer questions and get an instant rating based on the Big Five. Significantly, people could also opt in to share their Facebook data with researchers.
People just love filling in those personality questionnaires, don’t they?
The app was such a big hit that soon researchers found themselves with an unexpected and unprecedented dataset. Millions of people had not only shared insight into their characters based on OCEAN categories, they’d also shared all sorts of details about their Facebook activity.
When researchers mapped the two sets of data against each other, multiple pennies started dropping. They realised you could accurately infer character type from fairly basic details of online behaviour. And you could predict a whole lot else besides. By 2012:
“On the basis of an average of 68 Facebook “likes” by a user, it was possible to predict their skin color (with 95 percent accuracy), their sexual orientation (88 percent accuracy), and their affiliation to the Democratic or Republican party (85 percent). But it didn’t stop there. Intelligence, religious affiliation, as well as alcohol, cigarette and drug use, could all be determined. From the data it was even possible to deduce whether someone’s parents were divorced.”
Now it was simply a question of feeding in more data and continuing to refine the model. Before long they could “evaluate a person better than the average work colleague, merely on the basis of ten Facebook “likes.” Seventy “likes” were enough to outdo what a person’s friends knew, 150 what their parents knew, and 300 “likes” what their partner knew. More “likes” could even surpass what a person thought they knew about themselves.”
It seems we’re all way more predictable than we’d like to think.
Soon the lead researcher started to have serious misgivings about where the project was heading. If you could accurately predict personality and core motivations from simple online behaviours, it wasn’t hard to imagine a not-too-distant future where absolutely everyone was absolutely knowable.
In the wrong hands, that kind of insight could be frighteningly powerful.
But by the time he pulled the plug on the project, it was already too late. A company called SCL – the parent company of Cambridge Analytica – had been following the research closely. When their attempt to buy the data failed, they simply went about building their own version.
By creating their own personality quizzes and adding data from all sorts of different sources – including land registries, shopping data and anything from data brokers like Experian – the company was able to build a picture of whole populations in more depth and more detail than ever before. Especially in the US, where data protection laws are much weaker than in Europe.
The question is, though, how useful can all that Big Data be?
This video gives you a good idea. In a slick and faintly terrifying nine minutes, the CEO of Cambridge Analytica explains how they’ve profiled the personality of every adult in the US – and how they used that data to help Ted Cruz (a pretty unappetising candidate) become the only serious contender to Trump for the Republican nomination.
They went on to use the same techniques to help Trump gain the presidency. Rather than relying on the blunt tools of demographics and geography, psychometrics was added to the mix. This not only allowed voters to be categorised into more meaningful groups, it also meant messages could be honed to resonate with different personality types.
Imagine a group of undecided voters in a swing district. Are you a fearful type? Here’s a shadowy fella climbing through a kitchen window. A traditional type? Here’s a grandfather teaching a cute kid how to use a hunting rifle. Both approaches aim to nudge people towards a pro-gun Republican stance, but do it in ways that best resonate with the audience.
The campaign used sponsored Facebook posts, pinpoint-targeted right down to individual streets and even buildings. The individual messages were tested and refined in real time too, constantly optimised to make the biggest impact on voters. And according to this recent article that may give you nightmares, it looks like AI was used to super-charge the whole process, pumping the handcart to Hell even faster than anyone ever thought possible.
Of course, anything Trump-related can seem instantly sinister. And it really doesn’t help that the CEO of Cambridge Analytica looks like Tom Hiddlestone auditioning for the role of an IT nerd arch-villain in an X-Men film. Or that Steve Bannon, Trump’s deeply creepy strategy guy, is on the company’s board.
But perhaps the techniques themselves are just a long-overdue step forward in targeted advertising. For years, we’ve been told that online advertising allows for hugely intelligent targeting of audiences. But still, most banners and the like seem comically dumb. “Aha, you’ve just bought a laptop. You are clearly someone who likes buying laptops. I will keep feeding you laptop ads. Even though you won’t need another laptop for at least five years.”
So while the Cambridge Analytica approach can appear downright dystopian in the political context, it will be interesting to see how it plays out in consumer advertising.
Will targeting people based on their psychometric profile create more powerful ads for everyday things like cars, trainers, chocolate and pensions? Or does it only really work in areas where emotions run as high as in the recent US elections?
Either way, I’m expecting my first brief from a psychometric AI planner-bot any day soon.