Christians and Muslims classified correctly in 82% of cases using Facebook data The Facebook “like” button can be sued to predict sexuality, political leanings and even intelligence, says a new study. Researchers at Cambridge University used algorithms to predict religion, politics, race and sexual orientation.
The research, published in the journal PNAS, forms surprisingly accurate personal portraits, researchers said.
The findings should "ring alarm bells" for users, privacy campaigners said.
The study used 58,000 volunteers who alongside their Facebook "likes" and demographic information also provided psychometric testing results - designed to highlight personality traits.
The Facebook likes were fed into algorithms and matched with the information from the personality tests.
The algorithms proved 88 percent accurate for determining male sexuality, 95 percent accurate in distinguishing African-American from Caucasian-American and 85 percent for differentiating Republican from Democrat.
Christians and Muslims were correctly classified in 82 percent of cases and relationship status and substance abuse was predicted with an accuracy between 65 percent and 73 percent.
The links clicked rarely explicitly revealed these attributes. Fewer than 5 percent of gay users clicked obvious likes such as gay marriage, for instance.
Instead, the algorithms aggregated huge amounts of likes such as music and TV shows to create personal profiles.
It also threw up some strange pairings.
"Curly fries correlated with high intelligence and people who liked the Dark Knight tended to have fewer Facebook friends," said research author David Stillwell.