Facebook | Three likes and you’re gay

With as little as just three likes, Facebook is able to tell if a user on its social media site is probably gay or not, and then use this to make money.

Researchers from Columbia Business School, the Network Science Institute and the Stern School of Business sought to asses how the site infers “users’ personal characteristics” – such as sexual orientation – by their activities, to target them for advertisers.

“Users… can be targeted based on these inferred propensities and their relationship to particular content or advertising campaigns,” they said.

The researchers found that by liking pages for topics such as Lady Gaga, Katy Perry, the Human Rights Campaign, True Blood and Harry Potter, a user is likely to be identified as gay.

“Online marketers increasingly depend on statistical inferences based on available information,” they said. “While some online users may benefit from being targeted based on inferences of their personal characteristics, others may find such inferences unsettling.”

The researchers noted that there are implications for users’ privacy. “Users who prefer not to share this personal status even with their friends may not want it to be predicted by the system. To many, privacy invasions via statistical inferences are at least as troublesome as privacy invasions based on personal data.”

The system is not 100% accurate and “a user who is in fact not gay may not want an incorrect inference to be drawn about him or her”.

The researchers recommend that Facebook implement a ‘‘cloaking device’’ that allows users to request to not to be identified as gay or by any other personal trait or characteristic.

“The cloaking device essentially tells the system: ‘do not draw inferences like this about me’—or more practically, ‘do not show me ads or content for the same reasons that you decided to show me this’,” they explained.

The method proposed by the researchers would, they said, give users a better degree of privacy and “control over the inferences made about them by statistical models”.