Humans better at recognising faces than algorithmns

Humans better at recognising faces than algorithmns

Humans are better at recognising faces than algorithms, according to research at a British University.

According to research undertaken by the Applied Psychology Unit at the University of Greenwich, so-called ‘Super Recognisers’ are humans with extraordinary abilities in recollecting facial features. And for every one recognition made by one of the best computer systems in the world, 1000 identifications are made by Super Recognisers.

The research also says that humans are better at recognising low quality pictures of people than algorithms and super recognisers go even further by recognising faces they have not seen in years, in crowds, in disguise and degraded photos and images.

Facial recognition technology fails in real life scenarios, where people are viewed in low light and resolution, from diverse angles argues Dr Josh Davis of the Applied Psychology Unit, who commissioned the survey. He adds: ‘most criminals don’t smile for the camera.’

Dr Davis uses the survey commissioned by the University from Qualtrics to recruit Super Recognisers from the public and the police, and to fine tune his understanding of the cognitive processes that enable a certain part of brain to lend itself to face processing.

The research was designed to test the suitability of 2000 police candidates to work in specialist front line units, but its results needed to be validated with a control group of 300 members of the public.  Instead, the test went viral and in just over a year more than two million people had taken the test, and the number keeps rising, giving Dr Davis and his team more data to analyse than they had thought possible.

From his research, Dr Davis suggests that approximately one percent of the population are super recognisers.

Of the five hundred images gathered by Metropolitan Police every week, 250 offenders are recognised, of whom 125 are identified by super recognisers, adding up to 6500 a year.

The research has so far focused on short term memory, but in its next phase Dr Davis will set long range tests, in which participants view a certain number of photos or videos. These are then followed up weeks, months and if possible years later, to test how well they recognise faces over long periods of time. Initial tests reveal uncanny abilities to remember faces even from degraded photographs, after years have passed, in photos never seen before.

Are you a super recogniser? Take the test.

Register for Intersec, 17-19 January 2023 here