If you tell your Facebook friends you "like" artists such as Lady Gaga or brands such as Harley-Davidson, a surprising amount of information can be extracted from that, reveals a paper published Monday in Proceedings of the National Academy of Sciences.
"And you can extract it very accurately – you can get my gender, race, political views, religion, sexual orientation, personality, IQ and so on," said Michal Kosinski, a University of Cambridge researcher who led the study in an interview Monday.
- RELATED: 8 Facebook privacy flaps
Kosinski noted that Facebook "likes" are public and visible to everyone by default, including Facebook, your internet service provider, your web browser, the government, and many smaller apps that you provided consent to – some of which may have the potential to misuse that information.
Kosinski , his colleague David Stillwell, and Thore Graepal, lead researcher at Microsoft Research's Online Services and Advertising and Applied Games group, trained a computer using a group of 58,000 Facebook profiles provided by American volunteers and the 55,000 most popular "likes" on Facebook. Users had a median of 68 "likes."
Following training, when provided with the "likes" of thousands of other volunteers, the computer predicted many of the traits with surprising accuracy. It was right:
- 93 per cent of the time about whether the user was male or female.
- 95 per cent of the time about whether the user was African-American or Caucasian.
- 88 per cent of the time about whether a male user was homosexual or heterosexual.
- 85 per cent of the time about whether they were Democrat or Republican.
- 82 per cent of the time about whether they were Christian or Muslim.
- 75 per cent of the time about how old the user was.
Surprisingly, it was also able to predict, with much better accuracy than chance, traits such as:
- Whether the user smoked cigarettes (73 per cent).
- Whether the user's parents had split up by the time he or she was 21 (60 per cent).
- The user's intelligence, according to an IQ test (40 per cent).
- How open a user's personality is (43 per cent).
The researchers have set up a site called youarewhatyoulike.com demonstrating the computer's ability to extract information based on a user's Facebook "likes."
The study noted that many of the most predictive "likes" weren't obvious ones. For example, fewer than five per cent of users labelled as "gay" were connected with gay groups such as the "No H8 campaign." Instead, likes such as "Britney Spears" and "Desperate Housewives" were "moderately indicative of being gay."
Meanwhile, the "likes" most correlated with high intelligence were thunderstorms, The Colbert Report, science and curly fries. While the study suggests that people should be aware of the impact of their Facebook "likes" on their privacy, Kosinski said that in some ways, the ability to extract information about likes can be used for users' benefit.
For example, they may help businesses better tailor and personalize web services. The information may also provide a way for people such as human resources staff and researchers to predict personality traits easily and more cheaply than other methods.
"They can be used and I'm sure they will be used for general improvement of the quality of our lives," Kosinski said. "It's just there are some risks that we have to take into consideration and protect ourselves from."Suggest a correction