They say you can’t judge a book by its cover, but a new study suggests a computer can accurately judge your personality by the things you “Like” on Facebook.
The computer’s judgments aren’t perfect, but they’re often just as good at predicting our personality traits as our coworkers, roommates, friends or family members, according to a new study.
FOR THE RECORD
An earlier version of this story misspelled the last name of Michal Kosinski, one of the study authors.
Only romantic partners are able to match the computer’s personality-divining prowess, the authors of the study say. But while our spouse’s knowledge of our personality has likely developed over years of dinners, phone calls and cohabitation, the computer program only needs a few seconds and a list of our Facebook Likes to come to the same conclusions.
“I would say that as humans, we are amazing at judging personalities,” said Michal Kosinski, a psychologist who works at the Stanford University Computer Science Department and one of three authors on the study. “What is striking is that computers can beat us at our own game using a simple model and big data."
According to the computer model, people who like pages for “Snooki” and “BEER PONG” are more likely to be outgoing and active, while those who like “Doctor Who” and “Wikipedia” are more likely to be shy and reserved. Liking “the Bible” and “smiling” suggests a person is on the cooperative end of the spectrum, while those who like “Atheism” and “CHANEL” are more competitive.
“When you look at the results, there are some indicators that are very obvious, and others that we wouldn’t have realized on our own,” said Youyou Wu, the first author on the paper, who worked on this research as a PhD student at the Psychometrics Centre at the University of Cambridge. (She is now an intern at Facebook.)
The data from the study comes from the myPersonality project, a Facebook application designed by David Stillwell of the University of Cambridge. Between 2008 and 2012, 6 million myPersonality users agreed to take a lengthy psychometrics test in exchange for getting feedback about where they fall in the spectrum of the five big personality traits -- openness, conscientiousness, extraversion, agreeableness and neuroticism.
myPersonality also collected information from participants’ Facebook friends who agreed to answer a 10-item questionnaire designed to measure how a given person’s personality was viewed by others.
For this study, published in the Proceedings of the National Academy of Sciences, the researchers used a sample of 86,220 people from this data set.
The authors began the study by feeding the computer the Facebook “Likes” of a few thousand people and telling the computer what their personality was like, based on the tests they completed. The computer came up with a model of how the Likes correspond with personality traits, which the researchers then tested with another batch of Facebook Likes and personality traits.
The more Facebook Likes a person has, the more accurately the computer model can discern his or her personality, the researchers report. The computer needed just 10 Facebook Likes to judge a participant’s personality more accurately than a co-worker, 70 likes to do better than a cohabitant, 150 likes to do better than a family member, and 300 likes to do about as well as a spouse.
“It definitely surprised me that computers can be as accurate as romantic partners,” said Wu. “We know from previous studies that romantic partners know us really well because we spend a lot of time together.”
It turns out the computer is better at judging some aspects of our personality than others. It tends to be very good at judging openness, and not so good at judging neuroticism.
“Openness is related to liking art and poetry and being liberal, and these are very visible on Facebook,” said Wu. “But neuroticism is largely invisible in the information people present in the digital environment.”
The researchers say they understand that a computer that is as good at or better than we are at predicting our friends and loved ones personality traits may be unnerving for some people. Both Wu and Krosinski noted that the technology itself is morally neutral, but they acknowledged it could be used to hurt people.
“It could be a cheap, quality assessment of of what career would make the most use of a person’s talents and resources,” said Krosinski. “On the other hand, you can imagine advertisers using it to craft marketing messages that are difficult for you to resist.”