As I See It: The Politics of Self-Disclosure
March 27, 2017 Victor Rozek
It is quite possible that two of the most unexpected and consequential events of recent times had their origins with a student in Warsaw. Perhaps if he hadn’t been accepted to Cambridge University, the fates of two nations would not now be spinning off in unforeseen directions. But he was, and in Cambridge, the ripples of his fascination with psychometrics grew into an unintended tsunami. His name is Michal Kosinski, and there are a lot of people angry with him.
Psychometrics is the reality-based branch of psychology, at least in so far as it is data-driven. It is rooted in a model developed in the 1980s, which was designed to provide insight into human beings and predict their behavior based on five personality traits. Crafty minds arranged them to make optimal use of their acronymic potential: Openness, Conscientiousness, Extroversion, Agreeableness, and Neuroticism, or OCEAN.
Using extensive questionnaires, researchers used these five attributes to predict subjects’ receptivity to new experiences; their degree of perfectionism, sociability, and cooperativeness, and the ease with which they get upset. But although the model yielded a high degree of accuracy, it was limited by its cumbersome data gathering methodology. Then the Internet exploded, and some kid named Zuckerberg had the idea of a generation.
At the time, Kosinski was a research psychologist and it didn’t take him long to see the solution to his data gathering constraints. The process he perfected is capably documented by Hannes Grassegger and Mikael Krogerus writing for Motherboard, but suffice it to say that Kosinski “developed a method to analyze people in minute detail based on their Facebook activity.”
The key was tracking their “likes,” and it took remarkably few data points to begin putting meat on digital bones. An average of 68 Facebook “likes” would allow Kosinski to accurately predict skin color, sexual orientation, and political affiliation. But it didn’t stop there, according to Grassegger and Krogerus. “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.”
Even people who valued their privacy and tended to be selective about their online disclosures, were not necessarily immune. Kosinski made this sobering claim: with 70 “likes” he knew more about someone than their friends knew; 150 “likes” disclosed more than their parents knew; and with 300 “likes” he knew subjects more intimately than their partners did.
One of the unintended consequences of research, however, is that others may use it for purposes not originally anticipated.
It turns out that the model Kosinski developed could be reverse-engineered, in the sense that it could not only create psychological profiles, but could search for them as well. If you wanted to identify mothers worried about education in London, or clean water advocates in Newcastle, or fuming middle class white males, or dog-loving singles, or left-handed gay Eskimos for that matter, you could find them.
All those little innocent “likes,” years of them, like dots on a pointillist’s canvass, meaningless individually, but collectively revealing an unambiguous picture. A seemingly minor expression of appreciation for an article about addiction; or a reaction to a street-crime video; or musical preferences; or political partialities; or a response to videos of starving children, or puppies at play, all revealed a nugget of valuable information about the user. And those seemingly benign quizzes that tell you which animal you are most like, or what color your personality resembles; from a user’s perspective, nothing more than mouse clicks, but in Kosinski’s hands they became the building blocks of individual personalities, preferences, and predictable behaviors.
“What Kosinski had invented,” say Grassegger and Krogerus, “was sort of a people search engine.”
It wasn’t long before Kosinski started getting murky inquiries about his model from the corporate sector, notably something called Strategic Communication Laboratories (SCL). Kosinski had never heard of SCL so he did what any curious psychologist would do: he Googled it. The company, it turns out, was in the “election management” business, and boasted that it could provide tailored marketing based on psychological modeling to “influence elections.”
Kosinski freaked, no doubt in a dignified way appropriate to psychologists. This was not the purpose of his work and he wanted no part of it. But enough of it had apparently leaked that others were distorting the model for their own purposes.
The first indication that the genie had left the bottle came a year later in the form of a British referendum. SCL is one of those Russian-stacking-doll type of corporations, structurally designed to obfuscate rather than illuminate ownership and affiliations. But one of its spawns is a company called Cambridge Analytica that was hired to convince British voters it would be a swell idea to leave the European Union. It proposed to do this by micro-targeting undecided voters with personalized messages, developed by “measuring people’s personality from their digital footprints, based on the OCEAN model.”
Deliberately or coincidentally, Kosinski’s work had been co-opted.
Virtually none of the election prognosticators predicted the Brexit result. At worst, they thought the vote would be close, but most indicators suggested Brittan would remain in the EU. That opinion, however, was based on demographics, which supposes that all people – or at least the majority – belonging to a certain group will automatically vote and think alike. But Cambridge Analytica based its approach not on demographics, but psychometrics. Voters whose profiles indicated a fear of immigrants, for example, were likely bombarded with scary immigrant news feeds and calls for restoration of cultural pride. Profiles indicating economic distress, might receive feeds that stressed job loss and the fiscal costs of membership in the EU. Essentially, any voter whose profile suggested an openness to leaving the EU, received posts that reinforced (and possibly exaggerated) what they were already inclined to believe.
The second unexpected event took place on the other side of the Atlantic. Prohibitive long shot Donald Trump hired Cambridge Analytica to sell his unpredictable and often contradictory messages to the voters. His inconsistencies, argue Grassegger and Krogerus, “suddenly turned out to be his great asset: a different message for every voter.”
Cambridge Analytica “divided the US population into 32 personality types, and focused on just 17 states,” although the company claims to have the ability “to predict the personality of every single adult in the United States of America.” Several of the targeted states were believed to be firmly in the Clinton camp, and thus received less attention from the Democratic candidate. And while the election results cannot be attributed to any single factor, it is reasonable to assume some strategies had more influence than others.
But how did a self-absorbed American political outsider get the idea to hire an obscure election management company from England? Turns out that the decision was not serendipitous. The biggest investor in Cambridge Analytica is billionaire Trump enthusiast Robert Mercer. And, sitting on its board of directors is none other than the president’s alt-right counselor and strategist, Steve Bannon.
With each passing day, more and more details of our personal lives are being gathered, stored, and analyzed by people whose interests may or may not coincide with ours – or our nation’s. Voter manipulation is just one unsavory outcome of unrestrained data collection. We can be sure there will be others.