Technology & AI

Turns Out That Opposites Don’t Attract After All

A study of “digital footprints” suggests that you’re probably drawn to personalities a lot like yours.

January 17, 2017

| by Martin J. Smith


Two toucans resting together

Digital footprint-based assessments are nearly impossible to cheat, which means they’re more likely to reveal your true colors. | Reuters/Jason Lee

Both science and intuition suggest that we choose our friends and romantic partners because we share things in common, such as age, educational level, race, religion, attitudes, and general intelligence. But what role does personality play in attraction? Are conscientious people drawn to conscientious people? Extroverts to extroverts?

Until recently no one really knew, says Michal Kosinski, an assistant professor of organizational behavior at Stanford Graduate School of Business. “Psychologists looked at this issue for many years, but the results were quite clear: Friends and partners are not similar in terms of personality,” he says. “This was surprising, because we know intuitively that people choose partners and friends who have similar personalities. It made us wonder if, perhaps, the researchers were doing something wrong.”

Personality is usually measured using surveys in which people answer questions about themselves. But people who answer personality questionnaires often evaluate themselves in the context of their peers. For example, an introvert who’s the most outgoing among his introverted friends might describe himself as an extrovert — a subjective judgment rather than an absolute, objective measure. That’s called the “reference-group effect.”

Kosinski and three colleagues decided to approach personality assessment differently. They used big data gleaned from Facebook — all those “likes” that users click to show their approval of someone else’s posts or opinions, and their word choices in their own posts and responses — to develop a more accurate picture of user personalities. The result is a paper Kosinski co-authored with Wu Youyou and David Stillwell of the University of Cambridge, and H. Andrew Schwartz of Stony Brook University in New York, published recently in Psychological Science. It concludes that, as its title suggests, “Birds of a Feather Do Flock Together.”

“Digital footprints left behind while using Facebook work pretty great when it comes to measuring personality,” says Kosinski, who coordinates the myPersonality Project, a global collaboration between more than 100 universities studying Facebook digital footprints of 8 million volunteers. “People log in to Facebook every day, leave a lot of digital footprints, and don’t treat it as anything unusual.”

This is much better than reporting on one’s own behavior using a questionnaire or being observed in a lab, he says. “When people realize that they’re being studied, they stop behaving naturally and change their responses and behavior.”

By contrast, he says, years worth of data on one’s behavior on Facebook or other digital environments are “way more difficult to fake.” Kosinski is convinced that personality scores based on digital footprints are at least as accurate as those obtained using traditional methods.

The implications could be profound for areas of business where personality assessment is critical, such as hiring and human resources. For example, Kosinski says corporate leaders trying to build effective teams within their organizations understand that “in a team, you need some thinkers, some doers, some leaders, and some followers. All those labels are good for management nonscientists. But it wasn’t understood before now that you should or could be matching people based on personalities.”


When people realize that they’re being studied, they stop behaving naturally.
Michal Kosinski

In addition to avoiding the reference-group effect, Kosinski says digital footprint-based assessments offer other advantages. Not only are they nearly impossible to cheat or misrepresent, but they can be applied to large populations, and can be administered quickly and without much cost.

“Any personality assessment method will have advantages and disadvantages,” he adds. “But if used properly, the digital footprint-based personality assessment can be more accurate, which is great for the company and the individual. If measurement is not accurate, you might be hiring the wrong person for the job.”

This most recent study is an offshoot of Kosinski’s previous studies of digital personality assessment, but he says the results may turn out to be the most important conclusion they’ve drawn. “Without really planning to, we found this profound influence of the reference-group effect [i.e. self-assessment] on traditional test results, and the outcome might be that some things we have long believed about personality psychology might not be true. How many other established truths in psychology are wrong or inaccurate because of that effect?”

If nothing else, Kosinski says the study raises doubts about the old saying that opposites attract. They might sometimes, but those cases are the exceptions, not the rule. “As it turns out,” he says, “the great majority of our interactions are with people who are a lot like us.”

For media inquiries, visit the Newsroom.

Explore More