Patterns of “Likes” posted by people on Facebook can unintentionally expose their political and religious views, drug use, divorce and sexual orientation, researchers said Monday.
A study of 58,000 U.S. Facebook users, reported in the Proceedings of the National Academy of Sciences, arises from an emerging discipline in which experts sift through extremely large digital data sets, such as collections of web searches or Twitter messages, for subtle patterns and relationships. It highlights the power and risks of digital demographics, which are the key to targeted online advertising, experts said.
“People who share the ‘Likes’ do not realize that they are sharing very private issues as well,” said psychologist Michal Kosinski of the U.K.’s Cambridge University, who led the study. “The predictions based on ‘Likes’ are very fine-tuned and very much on the personal level.”
Facebook’s billion users can click that they “Like” anything from a movie or a public figure to a comment a friend posts about his cat. There was a median number of 68 “Likes,” which are readily accessible online, posted by the users in the study.
“People who share the ‘Likes’ do not realize that they are sharing very private issues as well,” said psychologist Michal Kosinski at the U.K.’s Cambridge University, who led the study.
For this study, the researchers used demographic profiles, behavioral questionnaires and psychological tests volunteered by the users and then correlated that with “Likes” the volunteers had posted on the social-networking site. They set up a program to see whether these patterns could predict personal information about a group of people in the database based solely on the patterns of “Likes.”
While far from perfect, the record of Facebook “Likes” was in many ways as accurate as a personality test, researchers said.
The researchers found, for example, that “Likes” for Austin, Texas; “Big Momma” movies; and the statement “Relationships Should Be Between Two People Not the Whole Universe” were among a set of 10 choices that, combined, predicted drug use. Meanwhile, “Likes” for swimming, chocolate-chip cookie-dough ice cream and “Sliding On Floors with Your Socks On” were part of a pattern predicting that a person didn’t use drugs.
As a measure of the computer model’s accuracy, the researchers were able to distinguish between Democrats and Republicans in 85% of the cases; between black and white people in 95% of the cases; and between homosexual and heterosexual men in 88% of the cases.
“This study should set off privacy alarm bells,” said Jeffrey Chester, a privacy activist at the Center for Digital Democracy in Washington, D.C., a nonprofit group that lobbies on media and telecommunications issues.
Read More: wsj.com