Does Facebook know you better than your own mother?
Researchers have developed a program that can accurately predict your personality – and all it needs is your Facebook profile.
AP Photo/Paul Sakuma
Since its inception, the Internet has served as a bastion of anonymity: with digital avatars and online profiles users can interact socially while obscuring their true identities. But thanks to new sophisticated algorithms, your online presence might say more about you than you think.
User data has long been used to zero in on demographic information – browser history, for example, can be a predictor for gender and political affiliation – but new studies show that a person’s online habits can even reveal his or her personality. Now, a computer – using only Facebook “Likes” – can get to know you better than your closest friends and family, according to a new study published in the journal PNAS.
“The evidence increasingly shows that [data mining] can say a lot about a person,” says Michal Kosinski, a Stanford University professor who co-authored the study. “Our most recent study shows that our computer can beat your close relative in determining your personality. But computers can also predict your sexual orientation, your political views, and so on.”
Kosinski’s team used the Big Five, or OCEAN, model, to examine personality traits in subjects. This model considers five distinct personality factors; openness, conscientiousness, extraversion, agreeableness, and neuroticism. A person with a high openness factor might be more imaginative or adventurous. People who are more conscientious tend to be efficient and organized. Extraversion considers a subject’s level of energy and tendency to experience positive emotions. Highly agreeable subjects tend to be compassionate, and generally work well in groups. Neuroticism measures how easily a subject experiences negative emotions like anger and anxiety (being neurotic is not the same thing as having a neurosis).
Kosinski’s new research links these personality values to Facebook Likes. For example, people who score highly in openness tend to like “Salvador Dalí” and “TED Talks,” while extroverts tend to like “dancing” and “Snookie.” In the experiment, over 70,000 participants self-assessed their own personalities with an OCEAN model questionnaire. Then, the computer generated a personality profile of each subject based on their Facebook Likes. A series of human judges – coworkers, siblings, and spouses – were also tasked with creating a profile of their corresponding participant.
The results were startling. By mining just 10 Likes, the computer could predict a participant’s personality more accurately than a coworker. At 70 Likes, the computer was more accurate than a friend or roommate. At 150 Likes, more accurate than a sibling or parent. And with 300 Likes at its disposal, the computer could even beat out spouses for accuracy. It seems to suggest that almost anything we do online, right down to our Likes, is telling of who we really are.
In a different study published by Personality and Social Psychology Bulletin, researchers at York University in Toronto found that avatars – digital representations of people used in online interactions – are similarly linked to personality. Certain traits in an avatar’s design – like hair color, clothing, and facial expression – elicit certain (and surprisingly accurate) impressions of the creator’s personality.
"I think we went into the study not knowing what to expect," Katrina Fong, lead author of the study, said. "People consider their avatars to be a true expression of their personalities, so you might expect them to be connected. But avatars can also be a great venue for personal exploration, so we weren’t sure. There are studies showing that certain things, like a person’s bedroom, can provide fairly good information about their personality, so in that sense we were hopeful that we would find a correlation."
Fong's study suggests that digital avatars can have actual social consequences. In other words, an online interaction between avatars could convey impressions accurately enough to predict real-life friendships. In the same way, the Like-mining algorithm developed by Kosinski’s team has a wealth of potential uses.
“I think, first of all, it’s important to mention that this technology has immensely positive implications for humanity,” Kosinski said. “More fair and accurate psychological assessments – these can be rolled out to very nearly everyone at low cost. If you can automate it and have computers test personality more accurately, you could bring psychological assessments to those who really need it – the people at the bottom of the ladder. There are amazing benefits this technology can bring us.”
According to Wu Youyou, a University of Cambridge Ph.D candidate who co-authored the study with Kosinski, “data-driven decisions” could improve lives. By supplementing human judgement in everything from job recruiting to dating, their research could touch many aspects of human life. If it all sounds very dystopian, don’t blame it on the researchers – they have concerns about digital privacy as well.
“Obviously, as with any technology, it’s morally neutral,” Kosinski said. “You can use it for the benefit of humanity or you can hurt people. So I think we need to be very cautious about how we use this. If it turns out that a given company, or institution, or government is using this technology wrong – even inadvertently – the negative outcomes could be rolled out to a great number of people.”
Ultimately, Kosinski says, it should be up to each user to decide how their information is distributed. But that doesn’t mean we should stop sharing our information.
“Policies and technologies should be designed so that people have full control over their digital footprint,” Kosinski said. “People should be able to refuse access to their information. In the UK, there were sites that allowed users to opt out of cookies – but if you did opt out, they kicked you off the site. I think access to Google should be a basic human right. I also believe that people would be willing to share their data simply because when we share our data, the services become better. If I share my data with Google, the search results become better – not just for me, but for everyone else. It’s a community service in that way.”
[Editors note: A previous version of this story misspelled the name of a lead researcher.]