They say beauty lies in the eye of the beholder, but it actually goes deeper than that.
The concept of physical beauty lies in the mind, defined by the qualities we find attractive in others’ faces. These subtle preferences represent some of our most private inner thoughts – but that does not mean that they can not be monitored, and even predicted.
In a new study, researchers used electroencephalography (EEC) measurements to determine what kind of facial features people found attractive, and then fed the results to an artificial intelligence (AI) program.
The machine learning system – called a generative adverse neural network (GAN) – could first become familiar with the faces that individuals find desirable, and then produce completely new ones specifically designed to please: customized visions of synthesized beauty as unattainable because they were perfect.
The experiment, conducted by a team of psychologists and computer scientists at the University of Helsinki in Finland, was like a massive Tinder session for the 30 volunteers who participated.
Except for some big differences.
While the participants sat in front of a computer screen and showed them a series of faces, none of the faces were real people, but they were artificial portraits that looked realistic, from a data set of about 200,000 images of celebrities.
Unlike the usual use of Tinder, participants also wore elastic caps with electrodes designed to measure their brain activity while looking at the faces. They also do not have to wipe properly when they see someone they like the look – it’s all taken care of.
“They have to do nothing but look at the images,” explains cognitive neuroscientist Michiel Spapé. “We measured their immediate brain response to the images.”
Those individual measurements of neural activity were then assessed by the GAN, who could interpret the brain responses in terms of how attractive each artificial face was deemed by the viewer.
Using the data, the GAN was then able to generate new faces that were informed by people’s EEG attractiveness identifiers.
In a second experiment, these newly invented faces are again displayed to the volunteers, who have judged them in terms of attractiveness, along with other images of randomly generated faces.
Eventually, the results validated the test of the researchers, and in about 80 percent of the cases, participants rated the custom images as attractive, while the other faces were selected only 20 percent of the time.
Although only a small study, it’s just another example of how sophisticated AI systems become in their understanding of what makes us type – even in intimate and often unspoken concepts, such as the domain of personal attraction.
“Being successful in judging attractiveness is especially important because it’s such a gripping, psychological trait of the stimuli,” says Spapé.
“If possible in something as personal and subjective as attractiveness, we may also be able to look at other cognitive functions, such as perception and decision-making. Possibly we can focus the device on identifying stereotypes or implicit prejudices and better understand individual differences. ‘
The findings are presented in IEEE transactions on affective computers.