This article originally appeared on Motherboard.
In the wake of the 2016 presidential election, Democrats and pundits shocked by Donald Trump’s victory scrambled to explain how he could have possibly won. Some pointed to Cambridge Analytica, a data analytics firm employed by the Trump campaign that claims to have built extensive personality profiles, with over 5,000 data points, on every American, based on their online habits.
Cambridge Analytica says it uses them for so-called “psychographic targeting,” or crafting political advertisements and marketing campaigns that cater to people’s specific personality types. Some argued the firm’s shady techniques manipulated Americans into voting for Trump.
Exactly what Cambridge Analytica did, and how how much of an impact it had on the election has been debated in the press for the past year. But its tactics are still frightening to people concerned about online privacy, who believed they represent a future in which political campaigns and corporations exploit our unique vulnerabilities for their gain. It turns out, their fears were not unwarranted.
A massive new peer-reviewed study published in the journal Proceedings of the National Academy of Sciences (PNAS) last month indicates that psychographic targeting techniques are incredibly effective, and don’t require much knowledge about marketing to execute.
Psychological persuasion starts by building a profile about you using public information available online, like your tweets, Facebook Likes, or Instagram photos. Previous research shows that people’s psychological characteristics can be accurately predicted from their digital footprints.
The researchers conducted three field experiments that collectively reached 3.5 million individuals on Facebook, Instagram, and Audience Networks, a Facebook service that allows advertisers to reach people on other apps. In each of their experiments, the researchers used a single Facebook like to predict a person’s personality type.
They found that tailoring advertisements to people’s unique psychologies “significantly altered” their behavior, as measured by clicks and purchases. Their findings suggest that psychographic targeting makes it possible to influence the actions of large groups of people. The study demonstrates that we need to start thinking about what kinds of safeguards should be put in place to regulate online advertising.
“We need to think about the contexts in which we actually want or do not want to use this technology,” Sandra Matz, an assistant professor at Columbia University's Business School and the lead author of the study told me over the phone. She said it likely won’t be effective to make psychographic targeting illegal outright, but that she hopes online platforms like Facebook regulate its use.
“I don’t think it makes sense to shut down this technology all together,” she said. Matz explained that what she believes would be useful is to prohibit advertisers from exploiting people’s psychologies in certain situations. For example, you shouldn’t be allowed to create an ad that persuades people based on their personality to not vote. In the paper, she also outlines another example: It should be forbidden to target people who have addictive tendencies with gambling ads.
Matz says what’s crucial is presenting Facebook and other social platforms with clear guidelines they can follow, because tech companies already have demonstrated that they have trouble figuring out what's happening on their own sites because they run so many ads.
If we do not restrict the application of psychological targeting, “It will basically take a whole Facebook unit of thousands of employees” to regulate the practice, Matz said. “It’s just not feasible.”
Of course, Facebook is one of the most profitable companies in the world, and if it wanted to focus on regulating psychological marketing techniques, it certainly has the resources to do so. Though it would be a difficult feat, as Matz points out.
Matz stressed simply getting off Facebook or similar platforms won’t exempt you from psychological marketing tactics. “You can make the exact same inferences about people based on their credit card data or smartphone data,” she said.
Here’s how the study was carried out. Facebook doesn’t let advertisers target people directly based on their psychological traits, but it’s fairly easy to work around this. Marketers can target them based on what Pages they like, which can be incredibly illuminative. It’s not hard to deduce, for example, that at least some of the thousands of people who have chosen to like various Facebook pages about alleviating anxiety suffer from it. The study predicted people’s personality traits based on a single like per person—they didn’t even utilize comprehensive digital footprints.
For their first two experiments, the researchers sought to target people based on the traits extraversion and openness to experience. Based on their Facebook likes, they organized people into two groups: those with predicted high levels of each trait, and those with low levels.
So what kind of pages does a extroverted person like? According to the researchers, things like “Making People Laugh,” and the band Slightly Stoopid. Introverts like pages such as “Computers” and the television series Stargate.
In the first experiment, two sets of advertisements for a UK beauty retailer (meant to influence extraverted and introverted women respectively) ran on Facebook for seven days. The ads were seen by over 3.1 million women, attracted over 10,000 clicks, and resulted in nearly 400 purchases. In that experiment, the researchers found that users were about 1.5 times more likely to make a purchase after viewing an ad that matched their personality. Their findings held even after controlling for age.
“The effects are strong,” Matz told me. “But you’re not going to suddenly turn a Hillary voter into a Trump voter. It’s just shifting you in a slight direction.”
In the second experiment, ads were tailored for a crossword app based on a user’s predicted level of openness, meaning how much they like to try new things. That campaign reached over 80,000 users, and resulted in over 1,000 clicks. In that experiment, the researchers too found that targeting a user based on their personality is more effective, even after controlling for age and gender. The study found that users were 1.38 times more likely to click an ad if it was tailored to their personality type.
For the third experiment, the researchers essentially reverse-engineered their technique. They started with a predetermined audience segment designed by marketers for a “bubble app shooter game.” It predictably included users who already played similar games, like FarmVille and Bubble Popp. The researchers analyzed the overall personality type of the users, and found that they were highly introverted.
The study’s authors then promoted the app with a message designed around their personalities, and compared it to what the marketers used before. Previously, the marketers used the phrase “Ready? FIRE! Grab the latest puzzle shooter now! Intense action and brain-bending puzzles!” which is not exactly well-suited to introverts. The researchers instead went with “Phew! Hard day? How about a puzzle to wind down?”
Both campaigns ran on Facebook for seven days, and were seen by over 500,000 users. In all, 1,837 of them downloaded the app. The ad designed for introverts attracted significantly more engagement. Those who saw it were 1.2 and 1.3 times more likely to click (the researchers did the test twice).
Overall, Matz said, what is surprising is not that people respond to advertisements that are crafted based on their personalities. “The assumption was always that people of different personalities would behave differently,” she told me. “This is just the first time we can actually capture that.”
What was surprising was how easy it was to execute. “This technology, it’s feasible with existing advertising platforms. You can actually use something that’s already available and it has an effect,” Matz explained. “Everybody can run Facebook ads. It’s super simple, that’s the whole point.”