Can you imagine a world in which a machine will find out if you are gay, lesbian or heterosexual just by analyzing the peculiarities of your face? This could happen if a computer algorithm, that has just been designed by a team of researchers from Stanford University (United States), is as accurate as its creators proclaim.
According to these scientists, their system identifies the sexual orientation of the subject whose face studies with a higher success rate than humans. Find out if a man is gay in 91% of the cases (people stay in 61%) and if a woman is a lesbian in 83%, compared to 54% of the times that an intelligence of flesh and blood achieves , according to the tests performed by showing the same portraits to a group of volunteers.
The work will be published in the Journal of Personality and Social Psychology, but its authors have disclosed a draft and conclusions, which have created controversy and are being refuted by specialists and members of the LGBT community.
Criticisms of this research are based on the way in which it was carried out and the budgets from which it starts. The authors, Michal Kosinski and Yilun Wang, used 130,741 photos of 36,630 men and 170,360 of 38,953 women from an American dating website.
These images went through a filter of facial detection technology that was left with the 35,326 clearest photos of 14,776 individuals, which were the ones analyzed by the software. It took into account physical features – such as the width of the nose or the shape of the mouth, for example – and other aspects such as ornaments, makeup and hairstyle. From these parameters, it indicates the sexual orientation of the subject in question.
LGBT groups have protested that the images used in the study show only white people and do not include elderly, bisexual or transgender individuals, for example. But they have also complained that they have been extracted from a website to find a partner, where most people show similar photos, aimed at seducing and not showing reality.
Gay by nature?
But the aspect of research that has received the most criticism has to do with the theory that supports it. The authors start from the hypothesis that the exposure of the fetus to certain hormones during pregnancy influences their sexual inclinations. For example, if the baby is a child and the mother has low testosterone during pregnancy, certain regions of the brain of the fetus may be less masculinized and increase the chances of it being homosexual.
Since the androgens (the male sex hormones, among which testosterone is found) influence the formation of the face, Kosinski and Wang assume that gays will have effeminate features and masculinized lesbians, an assumption with which they have developed their algorithm.
However, there is no conclusive evidence that validates this prenatal hormonal theory as the main cause of homosexuality. In fact, science still does not know if being gay, lesbian or transsexual is something that comes marked only in genes, or if it is the result of a complex interaction between DNA, hormones and the environment.
In addition, critics of this investigation warn of the danger of developing algorithms and artificial intelligences that identify the sexual orientation of people, an interference in privacy that could lead to discrimination