Real robocops: how AI is changing policing

July 2016 marked a before and after in the annals of law. Under fire from a suspect who wanted to “kill white officers,” and with a dozen badges down, the Dallas police force sent a bomb-bearing robot in to deal with lone gunman, Micah Johnson.

The case is thought to have been the first time a U.S. police force had used a robot in a show of lethal force. More widely, though, police forces around the world are making more and more use of robots and artificial intelligence (AI) to combat crime.

The results are often less striking than the Dallas shootout, but no less vital in helping to maintain law and order.

See also: The future of machine learning is here

In the United Kingdom, for instance, an AI system called VALCRI (which stands for Visual Analytics for Sense-Making in Criminal Intelligence Analysis) is helping police spot patterns that could help solve crimes, acting as a kind of robot Miss Marple.

One other British AI system, Hart (for Harm Assessment Risk Tool), is set to help police in Durham, England, decide whether suspects should be released or kept in jail.

In trials, the system was 98 percent right in telling which people would be ‘low risk’, and could detect high-risk suspects 88 percent of the time.

Mick Neville, an ex-cop who works with Calipsa, a British firm making video surveillance systems built on deep learning models, said AI could help with mundane but key tasks such as going over footage.

“Most police forces still rely on humans to view hours and hours of video footage, when it comes to solving crimes,” he says.

But viewing huge amounts of video leads to fatigue, loss of attention, and errors, and is often done by people who may not have the right skills “or simply don’t want to be there,” Neville notes.

Using AI as part of this process is changing things in a big way. “Technology is helping the police to do their jobs better and more efficiently,” Neville says.

A classic instance of AI in police work might be the use of image-scanning systems to single out suspects seen near crime scenes.A classic instance of AI in police work might be the use of image-scanning systems to single out suspects seen near crime scenes. But there are plenty of other use cases.

See also: Cisco completes acquisition of AI company, MindMeld

“The technology can flag unusual situations, like someone running or individuals wearing masks, which is common in riot or public disorder situations,” Neville said.

And using AI to finger suspects need not be all about faces. In one case, Neville says, a thief was caught after AI matched the T-shirt filmed in three break-ins to one seen in a mug shot for an arrest.

A further way AI can help maintain law and order is in scanning many video sources in real time. In most control centers, says Neville, “it’s almost impossible for one individual to focus on several screens, however hard you try or however skilled you are.”

People get bored, look away, or focus on one thing rather than scanning all that is going on, he says. “AI, in this instance, can flag up incidents of note, such as a fire, fighting, someone running, or a car driving too fast,” Neville notes.

“It may not be something as dramatic as a terrorist attack, but anything unusual like a person entering an area they shouldn’t be in or running in an unusual environment is a red flag. Once it’s flagged, the human operator can then look at it and make a decision,” he says.

See also: Using AI to transform medicine

For all AI’s value, handing over to a human who can make the final call is crucial, says Richard Hartley, co-founder and CEO of Cytora, a startup that uses data and machine learning to gauge risk.

“AI should be used with precaution and decisions should not be solely based on its conclusions,” he warns. “AI can help our police system by freeing them of data analysis so that they can focus on more important tasks. But it should not be a replacement for police officers.”


The contents or opinions in this feature are independent and may not necessarily represent the views of Cisco. They are offered in an effort to encourage continuing conversations on a broad range of innovative technology subjects. We welcome your comments and engagement.

We welcome the re-use, republication, and distribution of “The Network” content. Please credit us with the following information: Used with the permission of