How Facebook 'Likes' Could Be Used to Make Personality-Based Hiring Decisions
That Facebook has collected a trove of information on its users should surprise no one. Many of us use the social network as a public journal of sorts, as well as a convenient, connected way to keep tabs on friends, loved ones and random acquaintances. As a result, Facebook knows our whereabouts, our preferences, who are friends are and the major events coming up on our calendars.
But how well does Facebook really know us? Put another way, how much can be accurately deduced about our offline psychological profiles from what we share on the social network? Can connecting the data points reveal real insight into our personalities?
The answer to the last question is an emphatic yes, at least according to a recent study published in the scientific journal PNAS, which found that one small sliver of our Facebook activity – what we ‘like’ on the platform – can predict personality traits with a startling degree of accuracy, beating predictions from our work colleagues, our friends, our family and even, in certain cases, our spouses.
This may be unsettling, but it also reveals the future of personality-based recruitment, says Michal Kosinski, a computer science research associate at Stanford University and one of the study’s authors. Using a “laughable amount of the digital footprint,” i.e. people’s public Facebook likes (“we didn’t analyze web browsing or search queries…there’s nothing kinky, intimate or outrageous here,” he says) it’s possible to decipher an individual’s level of openness, conscientiousness, extraversion, agreeableness and neuroticism.
From an employer perspective, the findings point to a new method for assessing candidates’ personalities quickly, cheaply, efficiently and most importantly, accurately. “Now we know it's better to look at your Facebook likes than to invite you in for a lengthy interview and ask you to fill out a questionnaire – a process that’s not just expensive, but very cheating prone,” Kosinski says.
The study’s researchers had more than 86,000 participants fill out a 100-item personality questionnaire, and then fed the results from 90 percent of the volunteers into a computer algorithm, along with their corresponding Facebook likes. The computer worked to draw links between openness, conscientiousness, extraversion, agreeableness, neuroticism and certain types of likes, and then used these connections to predict personality scores for the remaining 10 percent of the participants. With just 10 Facebook likes, it was able to generate a more accurate personality profile than a work colleague. With 65 likes it could beat a roommate or friend, 125 to best a family member and around 300 to do better than a husband or wife. (The average Facebook user has 227 likes.)
These findings comes on the heels of another study carried out by Kosinski and his colleagues, which showed that a computer program, again exclusively using Facebook likes, could predict extremely personal (and statistically valid) information about a person - everything from his or her race, to IQ, to sexuality to amount of drug use. As in this previous study, specific individual likes were highly correlated with certain traits. For example, “participants with high openness to experience tend to like Salvador Dalí, meditation, or TED talks,” the authors write, while “participants with high extraversion tend to like partying, Snooki (reality show star), or dancing.”
While the initial study “went wide” this one “focused on personality only,” explains Kosinski. “We dug really deep to optimize the model for that.” The same could be done for qualities such as IQ, political orientation or any of the other traits explored in previous study.
Dating, for one. The problem with online matchmaking is that people have a tendency to lie – about looks, age, occupation, but also about personality. “Using behavioral based assessments would improve the quality of the matches,” Kosinski says.
There’s an obvious potential for marketers, too. Personality affects what kinds of marketing we respond to, and companies (including Facebook) are already mining our in increasingly sophisticated ways. (As when universities analyze alumni’s social media activities in order to determine the best strategy to prompt them to donate.)
But the largest impact automated, accurate, and cheap personality assessment tools will have, Kosinski predicts, is on recruitment. Increasingly, personality is being trumpeted as “more important” than experience and harder skills, and corporations are experimenting with new, if relatively untested, ways to make hiring decisions based on the metric. (The Milwaukee Bucks recently hired a facial coding expert who claims he can decipher traits such as selfishness, resilience, and composure by studying the micro-expressions of potential players.)
Kosinski hopes that soon, companies will be able to evaluate candidates’ personality profiles compiled from their entire online behavior, encompassing everything from what they listen to on Spotify, to they search queries, to their purchasing habits. When you combine digital footprints from multiple sources – adding information about a person’s search history to what they’ve liked on Facebook – accuracy in predicting personality “will only go up,” he says. “It suggests that computers can not only beat us but they could probably beat us by large margins if we give them enough data.”
This brushes against uncomfortable terrain; as Kosinski's previous research illustrates, our online behavior can easily be parsed to decipher a panoply of personal details. “There is obviously a huge dimensional privacy worries – but I think to be honest, if we already know that we can predict accurately your sexual orientation and political and religious views, in a way who cares that we can also predict your personality profile,” Kosinski says. Instead of restricting the flow of data they share online, he believes individuals should push for greater transparency and control over who can access it.
Because at the end of the day, he insists, we will continue to reveal personal data points about our lives over social media. Like other big data evangelists, Kosinski preaches that the convenience of customization outweighs privacy concerns. “I’m willing to share [my digital footprint] with Facebook because in return, they offer me a very highly customized newsfeed, which I really enjoy,” he says. Similarly, he predicts job candidates will someday happily hand over detailed personality profiles culled from their online activity to employers so that they can be matched with a position tailored for their exact skillsets: “Like Netflix, except for job offerings."