We're looking for people he considers "suspicious". These include lone, predatory males; youths cruising around on bicycles; and anyone who happens to be sporting gangland tattoos. "I'm trying to be highly visible," he explains. "Just because we don't see them doesn't mean they can't see us."
Eventually, after smiling at a few shoppers, waving to a casual acquaintance, and making a final lap of the car-park, Gomperz looks at his watch, winds up the window, and drives off. Job done, he announces.
To a casual observer, it might seem a waste of a quarter of an hour of valuable police time. But the way the sergeant sees things, the long, slow tour of a seemingly normal parking lot is one of the most important things he's done all day.
That's how things roll in the brave new world of computerised "predictive policing", a hi-tech crime-busting technique which Gomperz is helping to pioneer and which ,has the potential to revolutionise modern law enforcement.
The technique revolves around a single mathematical algorithm, developed by the University of California, Los Angeles. This complex equation can in theory predict, with pinpoint accuracy, where criminal offences are most likely to happen on any given day. Though that sounds like the stuff of science-fiction novels the principle is relatively straightforward. It mines several years' crime statistics to tease out hidden patterns. Once identified, they are projected into the future, to predict where cars might be stolen, or houses burgled, on any given day.
The forecasts are startlingly specific, and that's where the policing comes in. Each morning, Gomperz and his colleagues in the Foothill area of Los Angeles, where the technology is being tested, are given maps identifying a handful of "boxes" measuring 500ft x 500ft – the supposed crime hot-spots for that day. The officers are instructed to visit them as often as time allows. Once inside the "box", they make themselves as visible as possible. The shopping centre car-park which Gomperz has just slowly criss-crossed is bang in the middle of one of that day's boxes.
Computerised "predictive policing" is still in its infancy, but all the early signs are that it has a startling effect. In the city of Santa Cruz, California, where it was first trialled this summer, a 25 per cent drop in crime was recorded. Meanwhile in Foothill, a 50sq mile portion of northern Los Angeles, the number of non-violent crimes has dropped from around 50 a week to nearer 40, in the eight weeks since trials began. So striking have the results been that UCLA's system will this year be rolled out across Los Angeles, one of America's largest police forces. And it could soon be crossing the Atlantic. On 23 January, Captain Sean Malinowski, a LAPD officer who helped to develop the technology, will showcase it to delegates at the Defence Geospatial Intelligence conference in London.
"Anecdotally it seems to work. From the data we're seeing, it stops crime happening," Malinowski says, adding that an era of declining resources (the LAPD currently has a paid overtime ban), predictive policing can be a particularly valuable tool, since it isn't labour intensive. "If a suspect turns up, say, to steal a car, and he sees a police officer, then maybe that's enough to stop him committing a crime that day.
"Making arrests is still important. It keeps officers motivated. And in this trial, it has certainly been happening, when officers are in their boxes. But arresting people also takes up a huge amount of time. Booking one guy can take up most of a shift. So if we can reduce crime without doing that, so much the better." Malinowski struck on the idea for computerised predictive policing several years ago, when he was asked to supply UCLA with crime data for a research project. A bookish man, with an interest in academia, he began attending lectures there, and soon learnt that it was theoretically possible to use equations to model naturally occurring events such as animal population growths, or migration patterns.
In conjunction with several UCLA academics, he decided to create an equation which could model crime patterns. The algorithm they came up with uses three pieces of information about each crime – the time, date and co-ordinates – before teasing patterns out of the data.
Jeff Brantingham, a UCLA anthropologist who helped to develop it, says the ongoing trial in Foothill is using a randomised control. On some days, the co-ordinates of the boxes officers are asked to focus on are randomly generated; on others, they use the algorithm. "Comparing crime on the days we use the algorithm and the days we don't will give us a true gold-standard test as to whether it really works," he says. "We are also trying to measure appropriate dosage: how long an officer needs to spend in a box for it to be effective."
The concept of predictive policing is not without critics. Civil libertarians are concerned it might lead to life mirroring the film Minority Report, which starred Tom Cruise, in which police target people for crimes they might commit at some undetermined point in the future.
It also has legal complications. Under the Fourth Amendment of the US Constitution, police officers are forbidden from stopping a suspect without "reasonable suspicion" that they are committing a crime. No one yet knows whether simply being in a geographic box identified by a computer programme represents reasonable suspicion.
"What happens if officers turn up at a 500ft x 500ft area, on the look out for burglaries, see a guy with a black bag, search him and find it contains stolen goods?" wonders Andrew Ferguson, an assistant professor of Law at the University of the District of Columbia. "If that case goes to court, there will be the question of whether it was reasonable for the officer to suspect he was committing a crime."
If computerised predictive policing catches on, Ferguson expects a test case eventually to work its way up to the US Supreme Court. In the meantime, he expects noisy kickback from civil rights groups. "That a computer can effectively curtail the Fourth Amendment rights of individuals in certain areas would be particularly troubling to the civil liberties lobby," he says.
"There will also be concerns that if police wanted to target certain areas, or demographics, then they could simply tweak the algorithm to ensure officers visit certain neighbourhoods."
Back on the streets of Foothill, Sergeant Gomperz has no such concerns. He drives to another 500ft square box – this time at a park named after Ritchie Valens, the Mexican-American songwriter who wrote "La Bamba" and grew up locally. As we pull up next to a children's playground, a young man, with tattoos, cropped hair and a menacing pit-bull terrier catches sight of the patrol car. Seconds later, he turns around, and hastily disappears from the scene.
"Maybe he's going home to his mother. Maybe he doesn't have a licence for that dog. Maybe something way more sinister is going on," says Gomperz. "All I know is that he was here. He's seen us, and he's moving on. When I see that, I think maybe that's stopped him committing a crime. And that makes me think that this predictive policing thing really works."
Algorithms: Formulae for success
So what exactly is an algorithm? You could define it as a step-by-step problem-solving procedure, especially an established, recursive computational procedure for solving a problem in a finite number of steps. Or, to put it more simply, it's like a mathematical recipe, a set formula for finding your way through data.
Humble long division is an example of an algorithm that almost everybody would recognise. At the other end of the spectrum are the thousand character-long codes used for automated trades in the financial markets or the series of complex instructions that make up computer programs.
The man believed to have coined the term is the ninth-century Persian mathematician, Mohamed ibn-Musa al-Khwarizmi, who is also credited with giving us the word algebra, which like algorithm, derives from the Latin version of his name, Algoritmi.
Algorithms are put to good use in the 2011 film Moneyball, which stars Brad Pitt as a baseball coach who uses computer-generated analysis to pick a series of undervalued players who are gifted but affordable.
Portia Walker
Comments