Platforms such as SmilePass’s SaaS solution can be crucial for those seeking to verify digital identity accurately, quickly and safely. However, a new study from the National Institute of Standards and Technology (NIST) will make interesting reading for many users of such services.
The multidisciplinary study tested the ability of experts versus artificial intelligence (AI) on the matter of facial recognition. Such experts frequently play a vital role in criminal cases, assessing security-camera images of defendants. Scientists from NIST and three universities sought to determine whether AI could assist these forensic facial examiners in their work.
What did the study discover?
The Proceedings of the National Academy of Sciences published the results of the research in late May. They showed, surprisingly, that trained human beings fared better when a computer, rather than another person, was their partner.
- Jonathon Phillips, electronic engineer at NIST, commented: “This is the first study to measure face identification accuracy for professional forensic facial examiners, working under circumstances that apply in real-world casework. Our deeper goal was to find better ways to increase the accuracy of forensic facial comparisons.”
The study found that neither human nor machine achieved the best results in isolation. Instead, collaboration between the two was necessary to make maximum accuracy possible.
What form did the research take?
184 people took part in the study, partly comprising 87 trained professional facial examiners and 13 ‘super recognisers’, signifying exceptional natural ability. The remaining 84 were the control-group participants. 53 of these were fingerprint examiners, alongside 31 undergraduate students with no facial comparison training.
The test took the form of participants rating 20 pairs of face images on the likelihood of each one being the same person on a seven-point scale. The researchers then tested four computerised facial recognition algorithms, developed between 2015 and 2017, with the same facial image pairs.
One finding of the study was unsurprising: the trained professionals delivering far superior results to the untrained control groups. The algorithms also did well. However, it was a greater shock to the team to find that the combined opinions of multiple examiners did not give the best outcome.
Instead, Phillips stated: “Our data show that the best results come from a single facial examiner working with a single top-performing algorithm. While combining two human examiners does improve accuracy, it’s not as good as combining one examiner and the best algorithm.”
Trusted software to verify digital identity
By creating biometric profiles for customers and verifying future risky transactions against these profiles, your organisation can significantly reduce fraud. Contact SmilePass today about how we can provide the platform that enables your firm to consistently verify digital identity.