I like a total of 34 pages on Facebook, among them The Atlantic Health, my high-school alumni page, and a bacon-themed restaurant in a city where I no longer live.

What does this reveal? Not very much, to my untrained eye—or at least not much of substance, beyond a healthy appreciation for cured meat. But to a computer, it could say plenty. In a study published Monday in the Proceedings of the National Academy of Sciences, researchers from Stanford University and the University of Cambridge found that by analyzing a person’s Facebook likes, a computer could accurately judge several aspects of their personality—more accurately, in some cases, than their friends or family.

Around 86,000 volunteers filled out a 100-question survey on Facebook using the MyPersonality app, a tool developed by Cambridge’s Psychometrics Center, and gave the app access to their likes. From there, the researchers determined which likes correlated most strongly to each of the personality traits collectively known in psychology as the “Big Five”: openness, conscientiousness, extraversion, agreeableness, and neuroticism.

“Exploring the Likes most predictive of a given trait shows that they represent activities, attitudes, and preferences highly aligned with the Big Five theory,” the researchers wrote. “For example, participants with high openness to experience tend to like Salvador Dali, meditation, or TED talks; participants with high extraversion tend to like partying, Snookie (reality-show star), or dancing.”

After submitting their answers, participants then had the option to pass shorter versions of the MyPersonality test on to their Facebook friends, who could score them for the same traits. Around 17,600 people were judged by one person they knew, and around 14,400 by two.

Familiarity, the researchers found, was no match for cold, hard statistical analysis. The more likes a person had, the better the computer was at making its judgments: With 10 likes, it was more accurate than a co-worker; with 70, more than a friend or roommate; 150, more than a parent or sibling; and with 300, more than a spouse. (The study authors estimate that the average Facebook user likes 227 pages, though in 2013, social-media analytics company Socialbakers put the number at 40 worldwide and 70 for users in the U.S.)

The study builds on previous research, published in March 2013, that used Facebook likes to predict a host of demographic factors, including sexual orientation, political leaning, age, intelligence level, and even parents’ marital status. Predictors of high intelligence, for example, included liking thunderstorms, The Colbert Report, science, and curly fries. (While some connections were obvious, the study noted, “other pairs are more elusive; there is no obvious connection between curly fries and high intelligence.”) The 2013 study also looked at personality traits, compiling a list of the most accurate predictors for each one—users who liked Hello Kitty, for example, tended to be more open, less conscientious, and less agreeable.

This latest study, like the last one, shows that “big data and machine-learning provide accuracy that the human mind has a hard time achieving,” said co-author Michael Kosinski, a computer-science professor at Stanford. “Humans tend to give too much weight to one or two examples, or lapse into non-rational ways of thinking.”

Well, yes and no. Kosinski and his colleagues measured what’s known in psychology as “self-other agreement,” or the level of similarity between a person’s assessment of themselves and the assessments of others—in this case, a computer model as well as other people. The computer, because it came closer to identifying the participants’ self-reported personality traits, showed a greater level of self-other agreement than did the human judges. That doesn’t mean, though, that the computer’s judgments are “more accurate,” as the study title declares. As a 2008 paper in the Journal of Research in Personality explained: “The difficulty with self-other agreement is that it assumes that the self is an accurate judge of his or her own personality, which is only true when people are willing and able to provide accurate self-judgments.”

That’s a doozy of an assumption in any context, but especially on Facebook, where declaring oneself a fan of something can be as much a performance as an act of self-expression. To borrow from the researchers’ examples: Liking meditation in real life says you’re probably mellow and into self-reflection, or trying to be; liking “meditation” on Facebook probably means that you want people to see you as such. Liking parties in real life indicates that you’re probably a social being; liking “partying” on Facebook probably means that you want to be known as a social butterfly, or that you haven’t updated your page since high school.

“We all rely on cues to make snap judgments when we meet new people, and those judgments can often be accurate, at least in broad strokes,” Jennifer Ouellette wrote in Me, Myself, and Why: Searching for the Science of Self. “Those cues may also include our ‘stuff’: Our choices in fashion, jewelry, tattoos, and key chains all provide clues about who we are, whether we intend them to do so or not.”

And our digital “stuff,” she added, is no different: “At the end of the day, Facebook is just one more tool we use for self-verification. We want to be known and understood by others in keeping with how we feel about ourselves.”

In that case, what the study may prove, more than anything else, is that it’s working—that our computers, more than the people close to us, know us as we want to be known. Whether they know us as we are, though, is another question altogether.