Artificial intelligence can tell if you are gay or straight just by looking at a photo


A controversial new study has claimed that artificial intelligence can tell if a person is gay or straight simply by analysing a photo of their face.

The study, titled Deep neural networks are more accurate than humans at detecting sexual orientation from facial images, was carried out by Michal Kosinski and Yilun Wang from Stanford University who said they were “really disturbed” by their findings.

The authors say an AI system had an 91 percent accuracy rate when picking out homosexual men and an 83 percent accuracy for homosexual women.

The study used an algorithm to analyse 35,326 images of 14,776 people taken from an online dating site to try and find “links between characteristics and facial features that might be missed or misinterpreted by the human brain”.

The algorithm analysed what are classed as “fixed” features such as the shape of the nose and eyes, as well as “transient” features which are described as “grooming style”.

When the algorithm analysed a photo of a homosexual man and then a photo of a hetrosexual man, the researchers say it identified the homosexual man as “more likely to be gay” 81 percent of the time.

However, when the algorithm was able to analyse five or more photos of the same person, the accuracy rate increased to 91 percent.

The study then compared the accuracy of the algorithm to that of human judges.

By comparison, the human judges were accurate only 54 percent of the time for women and 61 percent for men.

The study concluded that homosexual men and women have “gender-atypical facial morphology, expression, and grooming styles”.

It also states that human faces “contain much more information about sexual orientation than can be perceived and interpreted by the human brain”.

The researchers claimed that homosexual men have longer noses, narrower jaws and less facial hair compared to their hetrosexual counterparts.

“Lesbians tended to use less eye makeup, had darker hair, and wore less revealing clothes (note the higher neckline)—indicating less feminine grooming and style. Furthermore, although women tend to smile more in general, lesbians smiled less than their heterosexual counterparts”.

The researchers say that their findings have exposed a potential threat to the safefty and privacy of gay men and women, especially as governments around the world are increasingly developing and deploying facial recognition technology.

They highlighted countries such as Iran and Saudi Arabia where homosexuality is punishable by death.

Via: The Economist & Guardian


Comments are closed.