Much has been written about Donald Trump's enthusiasm for Twitter, a category of analysis now basically its own genre. Mostly we focus on what his tweets tell us about his impulse control, or what they tell us about his disregard for the truth, or what they tell us about actual developments in national security. Most recently, though, people are wondering what they tell us about his marriage.
Researchers from ITMO University in Saint Petersburg, Russia, and the National University of Singapore built an algorithm that culls patterns from social media posts to divine whether a user is married or single. It has a success rate of 86 percent. It also identified Trump as single. Yet we all know Trump is married, courtesy of memes of a sad-looking woman trailing after him at events.
People have latched onto the error. The algorithm is cool, but most of the reason it's getting this much attention is the Trump thing, which isn't really its larger point; Trump isn't mentioned in the paper the researchers published. Plus, its 86-percent success rate is partly a result of its combined use of Twitter, Instagram, and Foursquare simultaneously, and the Trump-specific analysis used Twitter alone (though that still included multimedia).
The algorithm nonetheless represents an intriguing new vehicle for the ever-evolving field of relationship science. There was already existing research on how a change in relationship status can trigger a change in personality, as well as on personality profiling in social media more generally. The scientists wanted to expand. They did so by running data from multiple social platforms as opposed to the more customary single one, on the grounds that social media users today tend to use more than one themselves. We're all aware of how our curated social media personas don't accurately reflect our real ones, and of how we shift our personal brand from platform to platform. By combining those overlapping sources of data, this algorithm gets a markedly fuller picture.
The logic is simple: To take location, for instance, if you're single, you're more likely to spend time in certain places at certain hours of the day. "For location data, the most valuable places are night clubs, universities, beaches, different sports places," first author Kseniya Buraya, a researcher at ITMO, explains. Twitter alone might not provide much info about your whereabouts, but it's likely Instagram and Foursquare could.
According to Aleksandr Farseev, who worked on the new algorithm and has been investigating what's now called Multi-Source User Profiling since 2014, their topics of interest when analyzing tweets were number of emotional words and topics, tweets about feelings in general, tweets with negative words, and tweets with slang. With the possible exception of the slang part, you can see how Trump, whose tweets are often negative and nearly always emotional, ended up in the category he did.
The barometer of personality change during a long-term relationship has historically been pronoun use. In the early days, people tend to describe themselves and their partner as separate entities—"me and her," "me and you."
"But when the commitment gets to a certain point, people spontaneously start to use plural pronouns," says Brent Mattingly, an assistant professor of psychology at Ursinus College who was not involved with developing the algorithm. "'We,' 'us,' 'ours.' The way they think about themselves begins to change to where it's not just them, it's them plus the partner. There's this communal identity that happens."
You will recognize this as the tactic couples use to get away with giving you one wedding gift instead of two. Once people reach this level of commitment, they think less egotistically and are more inclined to table their own self-interests for the sake of the relationship. They might even jettison some of their negative attributes—things like smoking or drinking or gossiping—a process that Mattingly and his colleagues dubbed "self-pruning."
They also absorb positive new ones, which Mattingly et al call "self-expansion." This usually applies to hobbies, personal interests, and so forth. If you're in a committed relationship with someone interested in, say, fashion, then you'd be expected to gradually develop a better sense of fashion yourself, or at least stop Scotch-taping your ties. It refers to all the classic self-improvement stuff like volunteering and exercise, but it also just refers to any new thing you and your partner experience together.
Of course, not all personality changes wrought by serious relationships are good changes. Self-expansion is when you pick up positive habits, but "self-adulteration" is when you pick up shitty ones. And just as self-pruning refers to dropping bad habits because of your partner's influence, "self-contraction" was coined for dropping nice ones. The next time your friend group vents about that one person who's recently disappeared into their relationship and stopped making an effort to hang out, you may helpfully inform everyone of the official terms for these changes.
"People in established relationships who don't show these [changed] patterns, it's possible they don't feel the level of commitment you'd expect from their status," Mattingly says. "There might be external factors that are keeping the relationship going, yet psychologically they don't really feel that sense of unity."
A June 2016 analysis of Trump's tweets confirmed that his #1 most-used keyword was indeed the pronoun "I." And while it seems reasonable to just figure a candidate discussing their platform would have to say "I" that frequently, Hillary Clinton didn't use it nearly as much when campaigning during the same period. Trump's favorite topic is himself, and his self-image does not seem to change with time, regardless of relationship status or pretty much anything else. This is a shame, because according to Mattingly, relationship science also shows that we're more effective at resolving conflicts when we've benefitted from a little self-expansion.
The other irony of saying Trump tweets like a single person is that his tweets do not, in fact, come from one single person. Some come from iPhone-wielding aides who will one day pen lucrative tell-alls, and the researchers said their algorithm's misreading of Trump's marital status can be explained by the fact that it wasn't just analyzing him, it was also analyzing his handlers, some of whom are presumably single. But you could say that of pretty much any politician—including Barack Obama, whom the algorithm correctly identified as married (again using only Twitter). As with Trump, it analyzed factors from a month-long period, including volume of tweets and emotional weight of vocabulary. Trump's now-infamous penchant for tweeting in the wee hours helped kick him over to the "bachelor" category here—like with check-ins at loud, crowded places, being active late at night is read as a single-person activity.
Interestingly, the style of the language here—if someone's social media persona has a distinctive tone or syntax—could prove as telling as the content. Erica Slotter, an assistant professor at Villanova University's Department of Psychology (also unaffiliated with the algorithm) who studies the factors of romantic relationships that make our self-concepts malleable, is currently running a new lab study that may shed more light on this. "It's not just what you say, it's how you're saying it," Slotter says. "We modify our vocal patterns to be closer to [those of] our desired partner. We try to make ourselves similar to them."
Her preliminary data shows that when single people are presented with prospective new partners in situations like dating apps—and when, crucially, the single person is interested—they shift their language style to match the syntactical structure of the party they're trying to impress. But Trump, whose ghostwritten tweets exhibit competency in spelling and grammar and whose actual tweets exhibit whatever we call the opposite of those things, admittedly has more pressing uses for Twitter than winning over the woman to whom he is already married.