You might want to reconsider the next time you feel an urge to "like" something on Facebook. Your "like" click could be all the information that's needed to distribute some pretty personal details far and wide.
Researchers at the University of Cambridge and Microsoft say they're able to predict, with a remarkable degree of accuracy, information such as a person's age, gender, political affiliation or race.
"Your likes may be saying more about you than you realize," said David Stillwell, one of the researchers involved in the study.
The study took 58,000 Facebook users for its dataset. The users agreed to have their "likes" tracked through an app called myPersonality, which analyzes psychometric testing results.
The researchers also state they're able to predict traits like intelligence, happiness, emotional health and conscientiousness, albeit less accurately.
Ethnicity and gender were the two traits most consistently predicted accurately — between 93 and 95 percent — followed by religion, political convictions, male sexual orientation and then female sexual orientation.
One of the more novel statistics that the researchers were able to unearth, with 60 percent accuracy, was whether or not a user's parents had stayed together past his or her 21st birthday.
"A growing proportion of human activities, such as social interactions, entertainment, shopping, and gathering information, are now mediated by digital services and devices. Such digitally mediated behaviors can easily be recorded and analyzed, fueling the emergence of computational social science and new services such as personalized search engines, recommender systems, and targeted online marketing," said the study.
A full PDF of the report is available here.
The new findings will hopefully serve as a warning to Facebook users to police their privacy settings more vigilantly. With an estimated 2.7 billion new likes registered every single day, advertisers have more material than ever from which to glean information about you and to custom-tailor their ads accordingly.
