How Bad Are Humans at Facial Recognition?

“This doesn’t mean that contemporary facial recognition technology has made us redundant…”

It’s no secret that facial recognition technology has become so advanced as to surpass human ability in a number of ways, but University of York researchers have just announced new findings about just how bad humans are in comparison.

To be fair, our capabilities are surprisingly impressive: The researchers found that on average, a human can remember about 5,000 faces. That includes people we know personally and famous people.

The researchers took an ingenious approach to their methodology, asking study participants to write down the names of people whose faces they could remember over the course of an hour. By assessing how participants’ ability to come up with names slowed over time, the researchers extrapolated their performance to guess at how many faces each subject would ultimately come up with.

In the end, there was a pretty wide range in performance, with subjects able to remember anywhere between 1,000 and 10,000 faces. Still, even those at the lower end of the spectrum would seem to be more than capable of recognizing acquaintances in everyday life. But that capability pales in comparison to the automated facial recognition systems now available, many of which can store databases of hundreds of thousands, and sometimes millions, of registered individuals.

This doesn’t mean that contemporary facial recognition technology has made us redundant, of course. A study published earlier this year in the Proceedings of the National Academy of Sciences, for example, found that when it comes to matching faces between two different photos, the best performance is produced when a human forensic expert is paired with a biometric algorithm. But when it comes to sheer volume, it does appear that machines are more up to the job. One security guard at a given company, to take a hypothetical example, probably won’t be able to recognize a staff of more than 10,000 passing through the gates each day, and might not even be able to recognize more than a thousand familiar faces.

Source: University of York

—

October 10, 2018 – by Alex Perala