Computers don't know your personality all that well, but neither do your friends

This image was removed due to legal reasons.

This week, a study claiming that computers could assess your personality more accurately than your friends made headlines.

Lots of people bought the story. Gizmodo, say. Or Time. The Register. Science Magazine, even.

Buried in the data is the fact that neither your friends — nor computers — seem to be very good at deciphering your personality at all. Which might mean these methods might not be the best way to fully examine one's identity.


A couple of years back, a group of scientists developed an app dubbed MyPersonality. Facebookers who used it would answer a 100-question survey that assessed their personalities on five traits: openness, conscientiousness, extraversion, agreeableness and neuroticism.

If I'd been one of the more than 4 million people who used the app, my data could then be mined by these researchers. (The study looked at a subset of 86,000 users.) And if I had friends who used MyPersonality, they might be asked to fill out a 10-question survey about me, on those same five traits. Because the scientists have access to my data, they could also look at what pages I liked, and feed my Likes into an algorithm that calculated, for instance, my degree of openness or neuroticism. (This was based on a model that equated things like running or liking Barack Obama with a certain degree of conscientiousness or agreeableness, which seems somewhat arbitrary.)

They'd then pit my friend's and the computer's assessment of me against my own to see how well they matched.

This image was removed due to legal reasons.

The methods, in an of themselves, seem clever, but here's the thing. The correlation coefficient — the metric they're using to measure who is "better" at judging a personality— is .49 for the average person. For the computer, it's .56. That's not terribly high when the scale runs from 0 to 1. If a friend or a computer really "knew" you, you'd expect the correlation coefficient to be closer to 1.

That doesn't mean your friends or computers don't get you at all. The 10-question survey, only asked a few questions about you, so it's possible your friends are actually more attuned to you than reported.


Some of the insights about people the researchers found from the Facebook data will undoubtedly be of interest to marketers. For instance, people who rated high on "openness" tended to like Salvador Dalí and TED talks. Extroverts gravitated toward partying, Jersey Shore star Snookie and dancing. They could infer political preferences, and such — valuable information for companies and organizations that want to target us.

But there is a silver lining. The findings also suggest that, at least for now, it's not easy for a human or computer judge to know all the facets of your personality. That makes sense. After all, personality isn't simple or static. It shifts depending on whether you're around family, the mean girls from high school or your boss.


At this point, distilling all the nuances that make you you into strings of data a machine can crunch to profile you just isn't possible. And that's a good thing.

In the future that might be different, or so the researchers warn:

Understandably, people might distrust or reject digital technologies after realizing that their government, internet provider, web browser, online social network or search engine can infer their personal characteristics more accurately than their closest family members. We hope that consumers, technology developers, and policy makers will tackle those challenges by supporting privacy-protecting laws and technologies, and giving the users full control over their digital footprints.


Daniela Hernandez is a senior writer at Fusion. She likes science, robots, pugs, and coffee.

Share This Story

Get our newsletter