Statistical Data Analysis

Predictive policing: Will the UK follow the US?

John Anderton: There hasn't been a murder in 6 years. There's nothing wrong with the system, it is perfect.

Danny Witwer: [simultaneously]

Perfect. I agree. But if there's a flaw, it's human. It always is.

(Scene from Minority Report)


In the early 2000s, a couple of my friends and I were at the movies one evening staring at the screen in awe as we watched Tom Cruise control a computer system just with a few swift hand gestures. It became one of the most iconic scenes in the movie as it demonstrated how awesome technology could be in the future. Fast-forward 13 years and we are using similar technology already. Although perhaps not in the same flashy way as portrayed in the movie, hand-gesture technology is being used in hospitals, and automated cars are set to become huge in the near future. But it was the storyline that got everyone thinking. The movie depicts a world where police officers work in a specialised department called “PreCrime” and arrest “murderers” before they commit the crime. This is based on foreknowledge provided by three psychics called “precogs”. The movie did showcase a highly futuristic concept, but are police officers using software to help them predict crime before it happens? It turns out that they are - just not to the extent it is portrayed in the movie.

“The information about who committed the crime is not sent to us. All we get is the event type, the location, and the time and date,” Larry Samuels, CEO of PredPol tells me over the phone from Santa Cruz, California.

PredPol is an American company that predicts crime using cloud software technology and wants to utilise the power of big data to make predictions. At the moment, Samuels tells me that PredPol is using up to 10 years’ worth of data to make these crime predictions. PredPol is not the only predictive policing tool provider out there but it has been gaining a lot of traction recently and is even being trialled in the UK.

How does it work?

“We process [10 years’ worth of data] through our algorithm and we then deliver predictions based upon that back to the law enforcement agencies in the form of map based data with tightly triangulated locations for crime predictions,” Samuels says.  “So 500 feet by 500 feet, or a better way to conceive that is half a city block, for example. The accuracy of those predictions ranges somewhere from 20-100% more accurate than other boxes in the city depending on where they are located.”

Samuels is keen to emphasise that this is not done in a “Minority Report” style way – PredPol only gets the what, the where, and the when. He says the technology is very useful in working with crimes involving gun crimes, burglaries, assaults, and grand theft auto. Some crimes are more difficult:

“Something like kidnapping, which is extraordinarily random and doesn’t really conform to any patterns because it is so random and so unusual, the number of kidnappings in any given year is so small that it is hard to derive a pattern from it. So that’s the sort of crime we are not in a position to predict.”

He tells me that police officers go about their regular day-to-day duties, where an officer could be responding to a call about a burglary or a car accident. But in between those calls of service, police officers drop by these flagged up areas - which according to Samuels makes a significant difference in helping to prevent crime.

“That episodic presence in the box at a random basis throughout the shift is literally all that is required to reduce the crime rate. It might seem surprising that that little of a presence could make a difference yet it does and makes a significant difference and the reason for this is because the police presence has caused that crime not to occur.”

But is an “episodic presence” of a police officer all it takes to prevent crime? I caught up with Dr Spencer Chainey, a specialist in geographical crime analysis and crime mapping at UCL, to get his opinion.

“My experience of working with police officers in the States, in the UK and in other countries is that, if you don’t tell them why they need to be there and what it is they need to do when they get there, chances are they are unlikely to do it,” Chainey tells me over the phone. “That’s not police officers being lazy - it’s just the nature of day-to-day policing. If you provide police officers with very clear prescription effectively on what it is they are required to do then they will do it.”

Chainey also questions how well PredPol’s software actually performs: “Nothing has been published to date which is an independent evaluation of how PredPol’s software performs in relation to some of the other predictive policing software and some of the general techniques that have been used in policing for years and years, such as good old hotspot analysis.”

PredPol’s predictive policing tool is not cheap.  Samuels says it can cost anywhere in the range from “10,000 dollars to several hundred thousand dollars”. So the question is: Is it value for money?

“The technology works very well and it works year over year so you get continued declines of crime,” says Samuel. “Atlanta has six districts and in two of the six districts they deployed PredPol and not the other four. 90 days into the test, the crime rate was down 8-9% in the districts that had deployed PredPol and it was flat or up in the other four districts. We then made the decision to deploy city-wide and crime went down 19%.”

The UK police force has taken note of these figures and now a number of cities in the UK, including Manchester, Kent, and West Midlands have been trialling PredPol and so far have seen some encouraging results. North Kent has been trialling PredPol’s software and has seen street violence slashed by 6%.

But Chainey is wary of the UK police force being seduced by PredPol’s “glamourous solution”:

“We should look at the success rate of how many solutions PredPol has sold in this country. They introduced this software to Kent Police two and a half years ago. How much of the police force has since been interested and enamoured by predictive policing software but have committed to buying PredPol? Zero.”

Chainey believes that part of the reason for this is because the policing styles of the US and the UK are very different. In the US police mostly drive around in cars, whereas in the UK the police are far more engaged with the wider community. He says the UK already has all the predictive policing software it needs. It’s just not utilising it to its full promise.

“There is the danger that the police force in this country can be easily seduced by nicely packaged software that claims to be the new thing since sliced bread but actually is unlikely to offer anything additional in terms of value for money on board of the existing technology that they have.”

Chainey says the danger with adopting these policing tools is that it puts more pressure on the police to be the ones who are the sole agency responsible for preventing crime when realistically they are not. He believes that the responsibility for preventing crime should be shared with other local agencies and the local public.

It is clear that countering criminal behaviour in a smarter way by using predictive technology offers plenty of promise. But the way crime is tackled in the UK is very different to the way it is handled in the US. As the trials continue in the UK - perhaps that will be the clincher.


« SurveyMonkey CEO answers a multiple-choice survey


GE sees robotics, data shaping a new industrial Internet of Things »
Ayesha Salim

Ayesha Salim is Staff Writer at IDG Connect

  • twt
  • Mail


Do you think your smartphone is making you a workaholic?