Many people see artificial intelligence (AI) and facial recognition software as the future of crime fighting worldwide. But British police say it should be used alongside traditional detection methods – including the use of so-called super-detectors.
Police forces nationwide use officers who have an exceptional ability to remember faces and have an above average ability to identify these people. However, only about 1% of the population has this “superpower”, says Tina Wallace, a surveillance expert at the Thames Valley Police.
Her team started recruiting specialist officers in 2017 and now has around 20 of these experts. This includes Alex Thorburn, an officer with 17 years of experience.
“I’ve always been good with faces, so when they sent out a notice about the tests, I did it!” Thorburn said.
“I was shown photographs, between 10 and 30 years old, of ten people. I had to find them in a crowd in a mall.
“I tracked them all down, but they looked very different from the photos. It was quite interesting.”
The team works on screens with footage from security cameras, but is also sent out into the field.
For the coronation of King Charles III, Thorburn was sent into the crowd at Windsor Castle, west of London. “We have been deployed to determine if there are any people with an obsession with the royal family present who might cause problems. Fortunately there wasn’t,” says Thorburn.
Mike Neville, who set up the first team of super-recognizers at London’s Metropolitan Police, says it is a cheap and effective way to tackle crime. He has already retired and nowadays runs the association Super Recognizers International which describes itself as the “world leaders in human recognition”.
Success
One of the first major breakthroughs of 2011 was during large-scale looting in London caused by the death of a man who was shot by the police. Police sifted through about 200,000 hours of security camera footage.
“Twenty officers identified 600 of the London protesters,” says Josh Davis, a professor of applied psychology at the University of Greenwich.
One officer, who was an expert on gangs, identified an astonishing 180 offenders by analyzing the images. Some of them he had never even seen alive. Anders’ faces were partially hidden, says Davis, who is also a member of Neville’s association.
The Thames Valley Police serves around 5,700 km² west and north-west of London with a population of 2.34 million people.
They deploy “super-recognizers” at specific times outside bars and nightclubs to point out those known to the police as sexual assaulters.
“We use plainclothes police officers and they are on the lookout for specific behaviour,” explains Wallace, an officer with 26 years of experience at the police training center near Reading.
“When we see offensive behaviour, we call in the help of an officer in uniform to stop it.
“Two out of every five men we stop have previous convictions for rape or serious sexual assault. We have stopped 520 of them in the last three years.”
No competition
As AI and facial recognition technology advances, human capabilities should not be discounted, says Neville.
“It’s not competition with facial recognition. These can be used together. AI is good with high-quality photos taken from before (like passport photos).
“People are better with lower quality photos, when the face is seen at an angle or partially covered by sunglasses or a mask.
“Also be aware that in terms of British and EU legislation, AI identification must be confirmed by a human before an arrest is made. Most people in a democracy would be happier if the decision to detain someone was made by a human, rather than a computer.”
Neville says there is a growing demand for the services of super-recognizers, especially from police forces in Germany and Australia.
Davis has uploaded a basic 14-point online test for anyone curious to determine if they have the abilities required of a super-recognizer.
“If you have less than 10 or 12, you are not going to be a super recognizer.
“But if you get 14 points, please call me…”
