It's relatively easy to learn a lot about Facebook users - from their
political views and gender to their intelligence, race and sexual orientation -
by following their clicks, new British research reports.
Just clicking that you "like" something on Facebook leaves a virtual but
lasting fingerprint of who you are, information that can be gathered and
analysed by marketers, credit agencies, companies, potential employers,
politicians or the government, the researchers said.
Facebook "likes" and other digital records, such as Google browsing
histories, are easily retrievable and can be used to create an accurate and
revealing portrait of a person, said Michal Kosinski, lead study author and
operations director of the Psychometrics Center at the University of Cambridge
"I can tell you with confidence that I can predict who you are without you
telling me anything at all, just from your Facebook 'likes,'" he added.
When you 'Like' something
Clicking "like" on Facebook postings allows users to publicly express their
positive association with online content, such as Facebook pages of restaurants,
products, photos, quotes, musicians, sports figures, actors, organisations and
Most people are unaware they are leaving a highly personal information trail,
Kosinski said. Unlike data that most people guard carefully - such as medical
history or financial information - "liking" something on Facebook seems casual
and relatively unimportant to the user. What's surprising is that sensitive
inferences can be drawn by organisations from seemingly non-sensitive data, he
Sophisticated data-gathering operations can analyse just about any
information someone shares, said Lillie Coney, associate director of the
Electronic Privacy Information Center, a public interest research organisation
in Washington, DC.
"The biggest problem for consumers is that they don't know when they click to
think three or four steps ahead about how that information could potentially be
used," explained Coney, who was not involved with the research.
How the research was done
The study, published in the journal PNAS, tapped data from
"myPersonality," a popular Facebook app that provides users with online tests
about their personality, intelligence, emotional stability and life
satisfaction. Apps are easy-to-use web applications.
Among all the users of myPersonality, about 58 000 agreed to give the
researchers access to their Facebook profile and social network data, including
tests they took using myPersonality.
The researchers developed a mathematical model and correlated what they
learned just by assessing the participants' "liking" behaviour to what the
researchers knew about the participants from their psychological tests and
Based just on Facebook "likes," the research model predicted:
- Gender, 93% of the time
- Race, (white versus black) with 95% accuracy
- Sexual orientation, (gay 88% of the time, and lesbian 75% of the time)
- Drug use, with 65% accuracy
- Political affiliation (Democrat vs. Republican), 85% of the time
- Religion, (Christian vs. Muslim) with 82% accuracy
- Relationship status, (single or with someone) 67% of the time
The authors also found that the research model was almost as accurate as a
short personality test would be in predicting the Facebook users' degree of
openness to new experiences.
The critical aspect of the research model was that it amassed large amounts
of seemingly innocuous information, such as favourite music or television shows,
love of animals or interest in friends' photos, to pinpoint a participant's
distinct characteristics, Kosinski said.
Coney said data brokers mine browsing histories and social media sites to
link a wide range of information to individuals, and sell their assessments to
potential employers, politicians and others.
"Somebody will pay to use this data; the accumulated 'likes' are something
people can sell," Coney said. "And, unfortunately, nobody is sending you notice
that somebody is using this information."
Kosinski said psychological assessment models like the one he created could
be used to potentially mine data from millions of Facebook users worldwide.
Kosinksi, who said he enjoys using Facebook and other online resources, urges
consumers to be careful. He added, "People should be aware that whatever they do
online can be used to infer traits and personality aspects way beyond what they
believe it can be used for. [Our research] shows this can happen."
He said he hopes his research will start a discussion that leads policy
makers and consumers to modify the technology so users have control over the
data they create.
His study received funding from Microsoft Research, where he works as a
consultant, and the Boeing Corporation.
Learn more about privacy issues in the information age from the Electronic Privacy Information