Skip navigation
Disciplinary Self-Help Litigation Manual - Header
× You have 2 more free articles available this month. Subscribe today.

Study Finds Combination of Experts and Algorithms are Better at Recognizing Faces

by Dale Chappell

I have to admit, I would be the worst witness in court, because it seems I can’t recall a face I’ve seen even a hundred times. However, there are some people who can see a face just once and recognize it even in grainy security-camera footage. These people are called super-recognizers and law enforcement routinely uses them to identify criminal suspects.

But how reliable are these super-recognizers? Since 2018, a team at the National Institute of Standards and Technology (“NIST”) has been trying to figure out this answer. At a forensics conference in fall, NIST researcher Jonathan Phillips presented what they’ve found so far. In short, their research shows no statistical difference between trained face-recognition examiners and super-recognizers, forensicmag.com reports.

The study compared four groups of test subjects: trained face-recognition examiners, super-recognizers, fingerprint examiners, and students with no training or experience (to represent the abilities of the general public). Trained examiners and super-recognizers performed about the same as a group, with both having some subjects who had perfect scores. Fingerprint examiners didn’t score as well as a group, but also had some subjects with perfect scores. The takeaway was that training in matching patterns was helpful in face recognition.

Next, the researchers combined groups in a process called “fusion.” Phillips said this reduced “overgeneralization” and “assumptions.” It also identified what process might be more reliable for face recognition.

“The results have been of interest in the community in terms of how to integrate algorithms and examiners in the process, and then have examiners write the report and testify in court that this is how they combined the two,” Phillips said.

The combination that researchers found worked best was fusing human face examiners with a facial-recognition algorithm, or at least two face-recognition experts working together.

Now the researchers are looking into other factors, such as whether looking at a face quickly or studying it longer affects remembering it, and how race plays a part. Cross-racial face recognition in court has been questioned in numerous studies over the years.

 

Source: forensicmag.com

As a digital subscriber to Criminal Legal News, you can access full text and downloads for this and other premium content.

Subscribe today

Already a subscriber? Login

 

 

The Habeas Citebook: Prosecutorial Misconduct Side
CLN Subscribe Now Ad
The Habeas Citebook: Prosecutorial Misconduct Side