DAVID STILLWELL fidgets with his empty takeaway cup as we talk. Sitting in this quiet cafeteria at the University of Cambridge, the ongoing firestorm of US politics feels a million miles away. But with Donald Trumpās surprise victory in the 2016 US presidential election, the fire found its way to him, thrusting the young researcher into the spotlight.
āItās uncomfortable,ā he says, uncomfortably. āPlenty of investigative journalists have wanted to have off-the-record conversations about what companies are doing and whether weāve helped them.ā
Advertisement
The conversations he is referring to concern what some consider a form of pervasive mind control. Stillwell played a key role in exposing ways that firms and governments can exploit our online data, mining it to create individual psychological profiles they can use to fine-tune adverts and political messages for maximum impact, ushering in an era of unprecedented digital persuasion of the masses.
It started in summer 2007. On a whim ā having just finished a psychology degree at the University of Nottingham, UK ā Stillwell made a Facebook app called . It let people take a test that describes personality types according to the āBig Fiveā traits, which include degrees of agreeableness, conscientiousness and extroversion.
Months later, some researchers asked Stillwell if they could use his data. But he hadnāt collected any. He had only set it up because āI thought it would be cool,ā he says. Then he wised up and started gathering data. It would prove a career-making move. When Michal Kosinski, then at the University of Cambridge, approached him a year later, Stillwell had a data gold mine of more than a million Facebook profiles paired to personality types (that number now tops 6 million).
In 2013, Stillwell, Kosinski and colleague Thore Graepel dropped the bombshell that machine-learning techniques made it possible toĀ . And accurately too. Just nine likes is enough to predict your personality traits as well as a colleague could. With 65 likes, as well as a friend; with 125 likes, a family member. Most people have around 225 likes, so organisations that possess this sort of data can . Not only that, it only takes a few Facebook likes to predict your age, gender, intelligence, sexuality, political and religious views, relationship status and a host of other things. In short, the internet knows just what pushes your buttons.
Laid bare
Why was that a big deal? For marketers, discovering someoneās characteristics to any great degree typically involved asking them to fill out a questionnaire, making it impractical. But if it could be done automatically, our psychologies are . Nowadays, even your , your personality predicted from your tweets.
What does it mean to have such knowledge in corporate hands? āWe can be hacked,ā says Stillwell. āManipulated, persuaded or encouraged.ā He gives the example of a smooth-talking car seller. āThe guy sizes you up and starts giving you the spiel he thinks will perfectly match you. As heās talking, you are either smiling or looking disinterested and heās using that to adjust his pitch as he goes along.ā Thatās the level of targeting we are talking about, says Stillwell. And it can be done online, by an algorithm, when you donāt have your psychological guard up.
Stillwell is now deputy director of The Psychometrics Centre at the University of Cambridge. His and colleagues involved 3.5 million people. They found that those targeted with online advertising based solely on a single Facebook like were 40 per cent more likely to click on an online advert and 50 per cent more likely to follow through with a purchase than those seeing untailored advertising. When such messaging can be scaled to target many millions of people at the press of a button, and with no regulatory oversight, that for some is an alarming degree of influence.
Nudges at vast scales might sway democracies. Cambridge Analytica, a firm that taps into big data, approached Stillwellās team in 2013. It had been spending big buying personal data from various sources and wanted help using it to influence US politics. The chance to work with this trove of data was tempting ā āwe discussed possibilities for monthsā ā but ultimately Stillwell couldnāt reconcile himself with Cambridge Analyticaās political ambitions.
āIf we think we have it bad today, mass pursuasion could make it a lot worseā
And that was that ā until Trumpās win shocked the world. Cambridge Analytica wasted no time in announcing that its social media ads tailored to individual votersā personality types were key. Many were scandalised: targeted campaign ads are legit, but automated personality profiling had the whiff of foul play.
When I ask Stillwell about the election result, he shrugs. āGiven the timescales involved, I didnāt think it likely that Cambridge Analytica would have been able to come up with something so genius it would have swung an entire election,ā he says. āWe spent four years on it and didnāt come up with anything that incredible.ā
Think of Trump as a dry run, however. With the amount of money and personal data that companies like Cambridge Analytica are feeding into the hopper, there is no doubt that personality-based profiling will be perfected. āJust because it didnāt matter in 2016 doesnāt mean it wonāt matter next time,ā says Stillwell.
So having lifted the lid of Pandoraās box, I want to know what worries Stillwell the most. Being drip-fed personalised messages without being aware that someone is trying to change how we think is insidious, he says. āThe power balance right now is weighted towards those who hold the data, and we really donāt know how itās being used,ā he says.
Later, Brett Frischmann, a law professor at Villanova University in Pennsylvania, goes further, telling me over the phone that we are on a slippery slope. āIf we think we have it bad with filter bubbles today ā with people being fed radically different takes on world events finding it harder and harder to find political common ground ā then [mass] persuasion could make it a lot worse. These tools could be used for a form of brainwashing.ā
Considering how much Stillwell knows about all this, Iām curious to what extent he moderates his own online habits. āThere is some horrific stuff Iām interested in that I still Google now and again,ā he deadpans. I laugh. He laughs. Awkward silence. āI donāt think my data is being used in a way that makes me feel uncomfortable,ā he adds. In a world of hand-wringing over big-data exploitation, Stillwell manages to find a silver lining. āIn the 1950s, shopkeepers knew you and sold stuff in a way that made it relevant to you ā it was a personalised, one-to-one relationship. Then we moved to a world of internet stores where, in the name of efficiency, we give everyone the same experience.ā
Stillwell believes that automated psychological profiling is helping to bring back that personal touch. āItās not necessarily all about people taking advantage of us. We can also be treated like individuals again.ā
Profile
David Stillwell is deputy director of The Psychometrics Centre at the University of Cambridge, and lecturer in big data analytics and quantitative social science
This article appeared in print under the headline āThe internet knows you all too wellā
