Society dictated your face, according to biometrics research and a patent

Software apparently can guess a person’s surname with some certainty by just looking at their face with enough accuracy to have been awarded a U.S. patent. If accurate, the premise could be used in facial recognition algorithms.

After checking to see the patent was not awarded April 1 (it was not: Mar. 30, 2021), it seems that algorithms can identify someone’s name solely by examining a photo. Research reportedly indicates that that is because people grow into the expectations people have of his or her name.

A 2017 research paper by researchers at The Hebrew University of Jerusalem, made the case that stereotypes associated with names can, over time, change a person’s appearance enough that algorithms (and people) can pick it up.

They can match the right name with the right face with sometimes surprising accuracy.

Mitre researchers write in their patent documents that their matching accuracy has been from 72 percent to 80.5 percent.

A Forbes article about the patent found a doubter in Lex Gill, a Canada research fellow with Citizen Lab, which monitors surveillance technologies. Gill said much of what can be picked up in a photo are social constructs and thus unreliable indicators of the person’s name.

Edited 04/14/21 to update the date the patent was issued. 

Article Topics

accuracy  |  algorithms  |  biometric identification  |  biometrics  |  biometrics research  |  facial recognition  |  Mitre Corp.  |  patents