Face recognition technology is impressive and may surpass humans’ performance in recognizing faces. The FBI last month said that its Next Generation Identification (NGI) system became fully operational last month; it will house more than 51 million photographs by next year, according to CNNMoney. NGI is expected to support 196 facial-recognition searches per day in 2015, according to the FBI’s NGI Facial Recognition Trade Study Plan.
Of course the jury is still out on how reliable and accurate face-recognition technology will be when deployed on a large scale. But there is already good evidence that people are relatively poor at the task. “Our sense of sight is famously unreliable,” writes Kevin Hartnett in the Boston Globe. That’s what lets magicians get away with their tricks and—in the case of facial recognition—how Hollywood can substitute body doubles for famous actors in movies.
Now, researchers at the University of California at Berkeley have shown that our brains strive to perceive sequential images of faces as consistent, even when the faces if viewed side-by-side would clearly be distinct.
“We found that how we see faces can actually be wrong—it can be biased by faces we’ve just recently seen,” wrote lead researcher Alina Liberman in an email to Hartnett. “Our brain seems to take advantage of the fact that we don’t expect people to change from one moment to the next,” she continued.
Psychologists say the preference for visual stability a “continuity field,” writes Hartnett, adding that “…it’s a good thing that helps us tame what otherwise would be a visually chaotic world.”
A paper describing Liberman and her colleagues work is published online in Current Biology.