Artificial Intelligence manages to know sexual orientation with just a photo


An artificial intelligence algorithm developed at Stanford University in California has been able to identify through photographs, and more accurately than humans, whether a person is heterosexual or homosexual. Thus, revealed a study carried out in this institution.

The researchers used about 35,000 facial photographs of men and women, posted on an American online dating site. The algorithm correctly distinguished the sexual orientation of 81% of males and 74% of females, according to the orientation mentioned by them on the site, while humans correctly identified 61% of males and 54% of females. The accuracy of the algorithm increased to 91% and 83%, respectively, when 5 images were analyzed per person.

That is, “faces have much more information about sexual orientation than can be perceived and interpreted by the human brain”, said study authors Michal Kosinski and Yilun Wang, quoted by The Guardian.

This study also argued that homosexual women and men tend to reveal characteristics and expressions typical of the fixed, such as the jaw, nose or forehead size; to conjunctures, such as hairstyle.

Although there are clear limits to the research, once black, transgender or bisexual people are not included, the implications on artificial intelligence are vast, raising questions about the biological origins of sexual orientation, the ethics of facial detection technology and the potential of such software to violate people’s privacy or be used for anti-LGBT purposes.

“It’s certainly haunting. Like any new tool, if it falls into the wrong hands, it could be used for unlawful purposes”, said Nick Rule, a professor of psychology at the University of Toronto, told The Guardian.

“Artificial intelligence can say anything about anyone with enough data”, said Brian Brackeen, executive director of facial recognition firm Kairos, in a statement to the same newspaper.

The authors of the study further advance that artificial intelligence can be used to explore the relationships between facial features and other phenomena, such as political views, psychological conditions or even personality.