‘I called another chief constable’  –  how best practice in use of AI has spread

Police Foundation report says chiefs lack ‘data literacy’ and adoption should be driven from the centre.

Most of the innovation within policing in the use of artificial intelligence has been achieved at a local level and forces will continue to ‘do their own thing’ if a new national centre for policing is too slow to deliver, the NPCC lead for AI has warned.

Speaking at the launch of a Police Foundation report into the police use of AI in London today, Alex Murray said that some of the best examples of using the technology to improve productivity had been achieved at a local level with other forces copying those initiatives.

The Police Foundation Report says that too many chief constables lack the ‘data literacy’ to exploit the potential of AI and are also risk averse around data. It calls on the proposed new national centre for policing to include a full-time national lead for digital technology with powers to set mandatory standards.

Mr Murray, who is also Director of Threat Leadership at the NCA, told the event: “We are, in policing, governance heavy and engineering light. Whenever we think about AI we think around how can we put some governance around this.

“The amount of times I have heard we can’t do this 43 ways whereas, in actual fact, you can. When you innovate sometimes you have to fail fast.

“But if you look at the leaders in this area  –  Avon and Somerset, Surrey, Sussex, Humberside and Bedfordshire  –  what did I do when I was at West Mercia, I phoned Trevor [Rodenhurst] and said can I copy what you did? That’s how it spreads. “

He said longstanding complaints about the quality of police data and the effect that has on AI  –  the so called garbage in, garbage out issue –  shouldn’t stop forces progressing with AI projects.

“If you look at the forces who are pioneering this they are on a twin track of improving data quality whilst embracing some of the benefits of AI. Don’t think it is a before and after [equation] because we will never do AI if that’s the case.”

With some forces having to make ‘eye watering’ savings in their budgets and AI having the potential to deliver that he was asked how quickly can the new national centre for policing help them to on board the “best in class” systems.

Mr Murray said the Police Productivity Centre within the College of Policing was already looking to ‘blueprint’ a number of in-use police AI systems including Amazon Alexa used by West Midlands Police to answer all their 101 calls.

He said: “The route map is there for this to happen but forces will not wait around if the centre is too slow. They will just go off and do it.”

He said using AI to improve productivity within UK policing was an area where forces were “excelling” moving at pace to use it effectively and copying best practice.

But failing to exploit AI to its full potential could cost the service dear, he warned. The Baumol Cost Disease is an economic theory which says that the tendency for wages in jobs, which have experienced little or no increase in productivity, to rise in line with other professions [and this could be applied to policing] means they become more expensive to run over time.

“So if our input costs go up and our productivity does not we will become inexorably more expensive as an agency delivering a worse service to the community,” Mr Murray said.

In need of further exploitation was the potential for use of AI for prevention and ‘predictive policing’ the report launch event was told.

“There is very little going on in that space,” Mr Murray said. “Yet for hundreds of years policing has said preventing crime is at the heart of what we do.”

Geoff Barnes is working with the Met on an AI technique called ‘random forest forecasting’ which helps it to decide where to focus its efforts by understanding who are the most dangerous VAWG and domestic violence perpetrators.

Mr Murray said this work is often not sufficiently understood as you would rather have too many false positives that false negatives because that means you are missing fewer potential suspects.

“You can tune models for what you want it to give you and we need to understand that more effectively” he added.

Work in the Met is also proving the effectiveness of Live Facial Recognition to prevent further offending by known suspects. The technology was used to upload a watchlist of active paedophiles. Deployed in the street at a specific location a ‘hit’ identified an offender walking down the street holding a young child by the hand. It turned out that the child belonged to a new partner he had just started a relationship with.

Mr Murray said: “No cop in a million years is going to remember all these faces going past.”

Bedfordshire is using AI to deal with its Clare’s Law inquiries the event heard. Chief Constable Trevor Rodenhurst told the event that until it adopted an AI solution it was using cancelled rest days and overtime to “do loads of searching” of its systems.

Using AI meant that “overnight we found 63 people who may have been a domestic violence perpetrator.” The time saved meant that those officers could be used to contact potential victims “which is what we wanted them to do in the first place,” he said.

“The risk that’s been taken out of that small use case is incredible.” He said the service tended to have a lot of people around whose job is to say “that’s really risky” but leadership needed to change that culture in order to get the best out of AI.

Category: InnovationLife facial recognition