There are people who just need a glance to remember someone, and others who wouldn’t even recognize a relative on the street (so to speak): the first in English are called super recognisersliterally super-recognizers, and have uncommon abilities in recognizing and distinguishing other people’s faces.
A study published in Proceedings of the Royal Society B: Biological Sciences he investigated their abilities to understand whether theirs is an innate or learnable gift, and what the secrets of their recognition technique are.
The study. The researchers used eye-tracking technology to measure where and for how long 37 super-recognizers looked at photos of people’s faces on a screen. They then compared the results obtained with those of 68 people with normal facial recognition skills, and then fed all the information collected to nine artificial neural networks already trained to recognize human faces.
Then the researchers asked the AI to complete the same task as human participants: deciding whether or not two faces belonged to the same person. The tests showed that artificial neural networks trained on data collected by super-recognizers were much more accurate at recognizing faces than those trained on data collected from people with normal abilities.
The right points. But how can super-recognizers be so precise in distinguishing the details of the faces of unknown people? «They don’t simply look more carefully, but more intelligently. They choose the most useful parts of a face to observe,” explains James Dunn, coordinator of the study. It is therefore not a question of capturing the greatest number of details, but of selecting the most useful ones to remember.
Caricaturists. And us mere mortals could we learn this technique? No, explains Dunn, comparing super-recognizers to caricaturists: «When you exaggerate the distinctive features of a face, it is easier to recognize it: super-recognizers do it by eye, assimilating the most characteristic features of a face».
Super recognizers VS AI. In a hypothetical race between super-recognizers and artificial intelligence, who would win? In situations of ideal conditions (stable light, fixed distances, high quality images − such as at passport control at the airport) AI is superior to humans; but in conditions that are more real than ideal, super-recognizers can do better, especially when it comes to identifying known people, because we humans can add context to the visual data (a mole on the mother’s face, a cut eyebrow typical of the brother).
«Our study», concludes Dunn, «shows that the ability to recognize faces is not just about what happens in the brain afterwards: it all starts with how we look. How we explore a face determines what we learn about it.”
