Brain Scans May Explain Why Deepfakes Look Hyperreal

University of Florida

AI-generated faces — known as deepfakes — increasingly flood the internet, often used to amplify common scams with fake romantic partners or supposedly trustworthy investment partners.

New research out of the University of Florida shows that deepfake faces can appear even more authentic than photos of real people. The study also links this "deepfake hyperrealism" to a crucial brain region known as the anterior insula, which quiets down when analyzing deepfakes, potentially helping explain why these images seem so convincing.

What's more, older adults — already more susceptible to scams for a number of reasons — showed an even stronger bias toward rating deepfake faces as more real than genuine photos.

"Fake faces are being perceived as more real than real faces," said Natalie Ebner , Ph.D., a professor of psychology at the University of Florida who led the study. "You're actually worse than just guessing, because you're actively being drawn in the wrong direction."

"All adults have hyperrealism bias, not just older adults. But older adults have it more," she added.

Ebner and her colleagues published their findings in October in the journal Social Cognitive and Affective Neuroscience.

The team assembled 80 images of people's faces, half real and half generated by AI. They then showed these faces to 126 participants ranging from 18 to 89 years old who rated the reality of the faces while undergoing brain scans to track the activity of the anterior insula. This brain region, located deep inside the brain, approximately at the level of the temples, acts as a central hub that connects to other brain regions and helps the brain recognize important or potentially suspicious information.

Both the younger and the older adult groups rated AI-generated faces as more real than the real photographs. But older adults rated the deepfakes as significantly more realistic, demonstrating a greater hyperrealism bias.

Looking at the brain scans, the researchers saw that the anterior insula was less active when participants viewed deepfakes, potentially explaining how our brains not only fail to perceive suspicious signals in the fake images but actually perceive them as even more genuine.

"The muted response in the anterior insula directly tracks with people thinking the fake faces are more real," Ebner said.

However, activity in this region was not significantly different between younger and older adults. So it remains an open question precisely why older adults have a stronger hyperrealism bias.

AI-generated faces likely appear more real than photos because they are generated as an average of many different faces, triggering our brain's preference for familiarity.

"We know that when you average across faces you end up with something that has less distinct features, is more attractive and represents more of the prototype," Ebner said. "That makes these faces easier to process and more pleasing."

Ebner said that AI-detection tools can help people assess the genuineness of photos online, but that the real advantage may simply be in awareness about deepfake hyperrealism bias. That's especially true for older adults, who are often targeted for scams involving fake people using AI-generated faces with many of them not even knowing that deepfakes exist, Ebner said.

"Identifying that this potential deception scheme is the first step to an intervention," she said. "We can now work on awareness that if something looks too good to be true, think twice before you go along with it."

/Public Release. This material from the originating organization/author(s) might be of the point-in-time nature, and edited for clarity, style and length. Mirage.News does not take institutional positions or sides, and all views, positions, and conclusions expressed herein are solely those of the author(s).View in full here.