UK police increasingly using AI to predict crime just like Minority Report

UK police chiefs are increasing their use artificial intelligence (AI) at such a rate that they could soon be able to predict and prevent crime in a mirror of sci-fi smash Minority Report.

The futuristic 2002 Steven Spielberg blockbuster starring Tom Cruise and Samantha Morton, saw a crack police unit use psychics to enable them to stop murders before they happened.

And such has been the advancement of technology that forces are now increasingly using AI systems to process vast amounts of data in seconds and produce risk assessments.

A team at the University of Chicago last year claimed to have developed an algorithm they said would be able to predict crimes a week in advance, with 90% accuracy.

And it is similar technology that the Home Office is currently exploring alongside police chiefs to create a new artificial system they believe will “transform policing” and represent the biggest revolution in policing since the Home Office Large Major Enquiry System ‘Holmes’ was introduced in 1985.

This saw the replacement of paper files with a computer database that immediately transformed the effectiveness and productivity of criminal investigations.

Serial killers such as Yorkshire Ripper Peter Sutcliffe had previously evaded capture because crucial details were not joined-up by investigators despite all the evidence being present in their files.

Now IT experts are developing the machine-learning tool, known as Project CDAP, which has undergone successful testing over the past three years and is expected to be rolled out to all UK forces within three years.

However, a new report by the Police Foundation states that one force is already “relatively advanced in its use of predictive policing software.

That force – Avon and Somerset – has been pioneering the use of AI technology since 2016 and they now use algorithms to weigh up the risk of allowing suspects out on bail.

Meanwhile an AI system trialled by Humberside Police links directly into UK police databases so that when a domestic abuse victim tells a 999 call-handler the name of her husband, AI instantly retrieves his details. This would confirm if the suspect is known to have a gun licence resulting in armed officers being deployed.

The report confirms that the biggest advancement has been in relation to AI “carrying out risk assessments of victims and perpetrators.

It reveals that programmes are used that with basic information such as name, address or car number plate, can predict the likelihood of an individual’s victimisation and vulnerability, of their being previously reported missing, of them being a victim of stalking and harassment, or of being a victim of a serious domestic or sexual violence.

It also confirms that these tools can also estimate the likelihood of a suspect re-offending and perpetrating burglary, stalking and harassment, serious domestic or sexual violence.

Rick Muir, director of the Police Foundation, states: “This overcomes a common problem facing many police forces, which is that they ‘do not know what they know’, because up until now it has been too time consuming and costly to run a manual check across multiple databases. That task now happens in seconds.”

A spokesperson for Avon and Somerset Police said: “We use risk models to help officers identify risk by scanning large volumes of data. These models do not conduct a risk assessment, but provide an indication of risk based on previously linked offences and crime types. No decisions are made by these models, they serve to support professional human judgement. Additionally, we need to be clear that the models are not predicting victimisation/offending, they are helping to quantify the potential risk – based on historical offending and victimisation. None of our models calculate the likelihood or probability to undergo an action or event. It’s also important to note that no demographic or geo-demographic variables are used in our models and we monitor for any bias routinely.”