One of the more perplexing things about the Trump presidency is why it exists in the first place: he took office having lost the popular vote by a wide margin, and with one of the smaller electoral college margins in memory. The win also defied virtually all of the pre-election polling and commentary: almost no one (except Michael Moore) predicted the outcome correctly, and on his victory lap, Trump himself admitted that he thought he was going to lose. So a lot of us have tried to figure out what happened (I continue to think the election was about white supremacy, though that doesn’t explain how Trump got the white supremacists to the ballot box; I’ve also wondered about the libertarian candidates and Clinton’s staggering failure to take the Good News about the auto bailout to the rust belt states). Others have wondered about the Comey letter, Russian hacking, and so on. Now there’s another possibility: the Trump campaign’s use of big data, as reported in this chilling article on Motherboard.
The gist of the article is this: Trump was using a shadowy data analytics company called Cambridge Analytica to micro-target voters. Guess who’s on the board at Cambridge Analytica? That’s right: Steve Bannon. The company’s CEO explains the process:
“First, Cambridge Analytica buys personal data from a range of different sources, like land registries, automotive data, shopping data, bonus cards, club memberships, what magazines you read, what churches you attend. Nix [the CEO] displays the logos of globally active data brokers like Acxiom and Experian—in the US, almost all personal data is for sale. For example, if you want to know where Jewish women live, you can simply buy this information, phone numbers included. Now Cambridge Analytica aggregates this data with the electoral rolls of the Republican party and online data and calculates a Big Five personality profile. Digital footprints suddenly become real people with fears, needs, interests, and residential addresses”
The key to this is the Big Five personality profile, and the technique appears to be based on a discovery a few years ago that you could determine someone’s personality profile from analysis of Facebook metadata – specifically, people’s “likes.” The paper, which has been my standard citation when I want to scare people about how much big data can infer from relatively little, showed that not only could you determine someone’s race, gender, and so forth with high accuracy from their FB likes, but you could also peg their personality type:
“For example, users who liked the ‘‘Hello Kitty’’ brand tended to be high on Openness and low on ‘‘Conscientiousness,’’ ‘‘Agreeableness,’’ and ‘‘Emotional Stability.’’ They were also more likely to have Democratic political views and to be of African-American origin, predominantly Christian, and slightly below average age.”
That result got less press at the time than some of the other results, but in many ways it’s more significant, since personality type is predictive of future behavior.
The Obama campaign’s analytics operation was widely credited as being significant in its success; Obama knew down to the household (and sometimes individual) level how people had voted in the past, and was able to make reasonable inferences about what was likely to get them to the polls. But as far as I know, they didn’t have personality type. Assuming the Motherboard piece is correct, the Trump campaign then carried that to the next level. It’s entirely plausible: research during the Republican primary showed that the best predictor of a Trump voter was authoritarian personality. A campaign that knew whose personality leaned that way could be very effective in turning out the authoritarian-personality voter.
It’s time to take a step back and think about how frightening this all is from the point of view of a functioning democratic politics. If it’s right, Trump was able to give voters, via social media, exactly and only the kind of message that their personality was likely to respond positively to. Researchers have, for a while, worried about the ways that social media might affect elections: Cass Sunstein was worrying fifteen years ago about how people on the Internet (pre-social media!) would tend to congregate with the like-minded, a process that would cause them to radicalize in their own views, increasing group polarization. More recently, Facebook showed that a “don’t forget to vote” reminder made a small but statistically significant difference in turnout; Zeynep Tufekci and Jonathan Zittrain had the simultaneous realization that this would enable the company to tip a close election by carefully selecting who was reminded to vote, and that the subterfuge would be essentially undetectable. The Trump campaign’s micro-targeting is the same kind of worry, but in a way it’s more disturbing, not only because it may have brought us President Trump (frightening enough!) but because that type of analytics will immediately become standard for all major national campaigns.
I don’t think it’s controversial to say that Trump is the most polarizing president in a long, long time. This sort of micro-targeting is likely to make that problem worse, because it reduces the need for candidates to present themselves fully to the public. In elections dominated by “low information voters” (or at least, where there’s enough low-information voters to decide the election), those voters will only get information likely to nudge them toward a candidate. They will never see aspects of the candidate that might not nudge them in that direction, or that might push them away. This is all speculative, but consider the curious case of the white suburban women who voted for Trump. Could they have been under-informed about his misogyny, or were they subtly told that mass media made it all up? Or what about the religious conservatives who voted for him? Were they willing to hold their noses at the ballot box for the sake of overturning Roe v. Wade, or did they simply not really hear much about his history of philandering? We will never know, but these sorts of questions, and their future iterations, are now going to be permanently with us. It’s also worth remembering in this context that Trump has repeatedly spewed venom about mass media, and told his supporters that media is fundamentally dishonest and not to be trusted. Thus prepared to distrust fact-checkers, these voters are ready to believe the “alternative facts” Trump presents them, perhaps especially in the form of gentle nudges via social media.
The late Benedict Anderson famously proposed that the modern nation-state was enabled by mass media: by creating a space through which citizens of disparate geographic and other characteristics understood each other, mass media enabled them to form an “imagined community,” whereby they viewed themselves as having similar characteristics, even if they really didn’t. It was this that fundamentally glued the modern nation-state together, and enabled the functioning of liberal democracy. Whatever one thinks of Anderson’s thesis, personality-type micro-targeting presents its opposite. Did that hand the presidency to Trump? Again, we will never know for sure, but the question is a deeply troubling one with potentially game-changing consequences for polities like the U.S.