AI study gives insights into why super-recognisers excel at identifying faces | Science

They have been used to track down Salisbury’s novichok poisoners, find murder suspects and even spot sexual predators. Now, research has revealed new insights into why super-recognizers are so effective at identifying faces.

Previous research has suggested that people with an extraordinary ability to recognize people look at more areas on a face than typical people.

Now researchers have used a type of AI to reveal how this approach contributes to their prowess.

“It’s not just about looking everywhere, it’s about looking smart,” said Dr James Dunn, first author of the study from UNSW Sydney.

Writing in the journal Proceedings of the Royal Society B: Biological Sciences, Dunn and colleagues report how they relied on eye-tracking data from one of their previous studies involving 37 super-recognizers and 68 typical recognizers.

In this work, participants saw both photos of full faces and others where the area of ​​the face they were looking at was partially visible.

In the new study, the team used this data to reconstruct the actual visual information seen by the participants’ eyes.

This “retinal information” was then fed into deep neural networks (DNN) – a type of AI system – which were trained to recognize faces. They also gave the AI ​​system a full image of either the same face the participant had seen or a different face.

In each case, the AI ​​system produced a score indicating how similar the retinal information was to the full facial image it had been given.

The team compared results from typical participants and super-recognizers as well as data based on randomly selected areas of the initial facial image.

The results reveal in all cases that the performance of the AI ​​system increased as the parts of the face looked at became more visible.

Additionally, at all visibility levels, the AI ​​system’s performance was highest when based on retinal information from super-recognizers.

“This shows that differences in face recognition ability arise in part from how we actively explore and sample visual information, not just subsequent processing by the brain,” Dunn said.

The team then investigated whether the results were simply due to the super-recognizers examining more areas of a face and therefore taking in more information.

However, they found that even when the amount of face captured in retinal information was the same, the AI ​​system performed better when fed data from super-recognizers.

“That means their advantage is not just about quantity, but also about quality,” Dunn said. “They choose regions that contain more identity cues, so each ‘pixel’ they choose has more value in recognizing a face.”

Dr Rachel Bennetts, a facial treatment expert at Brunel University in London who was not involved in the work, welcomed the study.

“For me, his main contribution to our understanding of super-recognition is the conclusion that superior facial recognition is not just about looking at a specific area, or looking longer or at more places on a face. Super-recognizers explore the face more broadly, but also allow more useful features to be sampled. information,” she said.

Dr Alejandro Estudillo from Bournemouth University said the study was based on showing still images to people under highly controlled conditions.

“It will be important to test whether the same pattern holds in more naturalistic and dynamic scenarios,” he said.

Although the study suggests that there are tactics that can make facial recognition easier, it seems unlikely that everyone can become a super-recognizer.

“At the moment, we don’t know whether these eye movement patterns could be trained effectively,” Bennetts said.

Dunn said studies suggest super-recognition is rooted in genetics and is hereditary.

“Super-recognizers seem to naturally select the most useful features, and that’s hard to train because it varies from face to face,” he said.

Researchers have developed a free test to help identify super-recognizers, available from UNSW Face Test.