Researchers have made an AI (Artificial Intelligence) understand the subjective notions of human beings about what makes a certain face attractive to us.
In this way, the device is able to create, completely independently, a personalized portrait of a face that has a very special characteristic: we love it. Or, what is the same, subjectively we would like it very much.
The results obtained could be used, for example, to model preferences and decision-making , in addition to potentially identifying unconscious attitudes.
Showing an attractive face depending on the brain we have
Researchers from the University of Helsinki and Copenhagen studied whether a computer was capable of identifying facial features that, subjectively, we could identify or label as “attractive”. And, on this basis, generate new images that correspond to our tastes.
The researchers used AI to interpret signals from the brain , combining a brain-computer interface with a generative model of artificial faces .
According to Michiel Spapé , the study’s principal investigator, previous research had designed models that were able to control and identify simple features of a portrait, such as expressed emotions and hair color.
However, as the scientist warned, while it is true that people largely agree on who smiles or who is blonde, whether an individual may or may not be considered attractive is, on the contrary, an issue. even more demanding study.
Not surprisingly, when judging what is attractive or not for us depends mainly on psychological and cultural factors , which are believed to play an unconscious role in our individual preferences, which is why on many occasions it is very difficult for us to explain what It is exactly what makes someone attractive to us: beauty is in the eye of the beholder (and, of course, in the beholder’s own brain).
Preferences manifested by the brain
The researchers initially assigned an aversive generative neural network (GAN) the task of generating hundreds of artificial portraits. Then, the images obtained were shown, one at a time, to a total of 30 volunteers who were asked to pay attention to those faces they found attractive, while their brain responses were recorded by electroencephalography (EEG).
Then, they analyzed the data obtained by EEG with machine learning techniques , linking us through a brain-computer interface to a generative neural network.
It would be an interface capable of interpreting the opinions of users about the attractiveness of a series of images. Then, based on that, it produces a completely new image of a face, with the characteristics of what a particular individual might consider attractive.
According to the scientists, the experiment worked in a similar way to the dating app Tinder: participants swipe to the right whenever they saw an attractive face. Therefore, they measured their brain’s response to those images.
The test results
By testing the double-blind procedure, the scientists found that the new images precisely matched the individual’s preferences by more than 80 percent.
Given that we are faced with a very personal and subjective psychological property , being able to measure an individual’s attractiveness is particularly important, since the possibility of detecting and generating images based on psychological properties , such as personal taste, would have been discovered. .
What’s more, its possibilities would go much further, since if AI is able to learn and understand more and more about subjective preferences, other important cognitive functions, such as decision-making or perception, could also be analyzed.
