HELSINKI, Finland — Beauty may be in the eye of the beholder, but a new study finds artificial intelligence can figure out which faces will make a person “swipe right.” Researchers from the Universities of Helsinki and Copenhagen say they have successfully taught an AI program to understand subjective notions which make certain faces more attractive to an individual.
Their study reveals the program could not only understand what sparked attraction in a person’s brain, but could also use this information to create new faces which appealed to the participants. The team says this breakthrough may pave the way for computers which model behavior based on the user’s preferences. It may also reveal unconscious attitudes people have regarding subjective topics.
The European team wanted to see if a computer could actually learn what facial features a person finds beautiful. From there, could the program then create new images — or artificial faces — to the participant’s liking?
To do this, researchers used AI to study brain signals of 30 volunteers. They then combined that brain-computer link with a generative model of artificial faces. This allowed the computer to model its own faces which it determined to be attractive to each user.
“In our previous studies, we designed models that could identify and control simple portrait features, such as hair color and emotion. However, people largely agree on who is blond and who smiles. Attractiveness is a more challenging subject of study, as it is associated with cultural and psychological factors that likely play unconscious roles in our individual preferences. Indeed, we often find it very hard to explain what it is exactly that makes something, or someone, beautiful: Beauty is in the eye of the beholder,” says Michiel Spapé from Helsinki’s Department of Psychology and Logopedics in a university release.
Beauty is in the brain of the beholder?
Study authors started their experiment by having a generative adversarial neural network (GAN) create hundreds of artificial faces. Each of the volunteers then viewed these portraits one at a time. Researchers asked the group to pay attention to the faces they found attractive while electroencephalography (EEG) scans measured brain responses.
“It worked a bit like the dating app Tinder: the participants ‘swiped right’ when coming across an attractive face. Here, however, they did not have to do anything but look at the images. We measured their immediate brain response to the images,” Spapé explains.
From there, the team analyzed the readings using machine learning techniques, connecting each person’s EEG results back to the GAN.
“A brain-computer interface such as this is able to interpret users’ opinions on the attractiveness of a range of images. By interpreting their views, the AI model interpreting brain responses and the generative neural network modelling the face images can together produce an entirely new face image by combining what a particular person finds attractive,” says Academy Research Fellow and Associate Professor Tuukka Ruotsalo.
AI gets really good at guessing your type
To test just how well a computer can really guess which faces get someone all hot and bothered, researchers had the AI create a new set of faces based on each person’s individual brain scans. The results reveal AI could accurately create a face the user found appealing over 80 percent of the time.
“The study demonstrates that we are capable of generating images that match personal preference by connecting an artificial neural network to brain responses. Succeeding in assessing attractiveness is especially significant, as this is such a poignant, psychological property of the stimuli. Computer vision has thus far been very successful at categorizing images based on objective patterns. By bringing in brain responses to the mix, we show it is possible to detect and generate images based on psychological properties, like personal taste,” adds Spapé.
“If this is possible in something that is as personal and subjective as attractiveness, we may also be able to look into other cognitive functions such as perception and decision-making. Potentially, we might gear the device towards identifying stereotypes or implicit bias and better understand individual differences,” the Finnish researcher concludes.
The study appears in the journal IEEE Transactions on Affective Computing.