Attractiveness is considered a challenging subject of study. This is associated with cultural and psychological factors that likely play unconscious roles in our individual preferences.
What makes a face attractive? You would have wondered over this question for sure. Guess what! Researchers have completed an Artificial Intelligence system to understand the subjective notions of what makes faces attractive. The device demonstrated this by creating new portraits that were tailored to be found personally attractive.
A researchers’ group from the University of Helsinki and the University of Copenhagen worked to determine whether a computer would recognize the facial features that are considered attractive. Based on this, the creation of new images was expected that match the criteria. The team used AI to interpret brain signals and combine the resulting brain-computer interface with a generative artificial face model.
According to Docent Michiel Spapé from the Department of Psychology and Logopedics, the University of Helsinki, models were designed to recognize and control simple portrait features in the previous studies.
Initially, the generative adversarial neural network (GAN) created hundreds of artificial portraits. The images were shown one by one to about 30 volunteers. The volunteers paid attention to faces that they found attractive, and the responses of the brain were recorded via electroencephalography (EEG). The EEG data were analyzed using ML techniques. The above connected the individual EEG data through a brain-computer interface to GAN.
A new portrait was generated for each participant to test the validity of their modeling. When Tested in a double-blind procedure against matched controls, the team found that the new images matched the subjects’ preferences with an accuracy of over 80%.
According to Spapé, the study demonstrates that the team could generate images that match personal preference, which was achieved by connecting an artificial neural network to brain responses. He also stated that by bringing in brain responses to the mix, the team shows it is possible to recognize and create images based on psychological properties.
The above study may benefit society by advancing the capacity for computers to learn. It is also believed to increasingly understand subjective preferences with the interaction between AI solutions and brain-computer interfaces.