Police Data Challenge: Automation and AI

THE POLICE DATA CHALLENGE Copyright Policing Insight/Cognyte 2023 In a society where the volume and complexity of digital data are ever-expanding, it should be no surprise that policing – as with many other public and private sector organisations – is looking increasingly to technology to address some of the key challenges. What sets policing apart from many others is that not only does the data it sources and interrogates have to be processed and stored securely, often with the need to maintain evidential integrity, but it may also need to be analysed at speed to help the decision-making processes and guide responses for operational officers on the ground. Given those considerations, law enforcement is looking increasingly to automation, artificial intelligence (AI) and machine learning (ML) for solutions to those challenges. And early indications are that these technologies hold huge potential for policing. Operational extraction and analytics Operation Venetic – UK policing’s role in the global campaign to target organised crime through cracking the EncroChat encrypted phone service – was the country’s largest ever law enforcement operation. The police data challenge: Maximising the potential of automation and AI The difficulties policing faces around the scale and complexity of digital data are widely acknowledged, but what about the potential for new technology to overcome some of those challenges? In the second in a series of five features looking at the police use of data, Policing Insight’s Keith Potter looks at some of the ways new tech – including automation, artificial intelligence and machine learning – are being employed in both investigative and preventative approaches. The ‘Police Data Challenge’ is an independent series researched and written by Policing Insight and commissioned by Cognyte. Policing Insight is the leading platform to keep up with the latest in progressive policing. It is where the global police and criminal justice community both consume and share knowledge, opinion and analysis. Cognyte is a market leader in investigative analytics software that empowers a variety of government and other organizations with Actionable Intelligence for a Safer World™. Find out more at www.cognyte.com/nexyte. Keith Potter is the editor of Policing Insight

THE POLICE DATA CHALLENGE Copyright Policing Insight/Cognyte 2023Led by the National Crime Agency (NCA), the operation was launched in July 2020, and has so far led to more than 3,000 arrests, over 1,200 convictions, and almost 8,000 years in combined prison sentences. It also generated huge amounts of data that needed to be collated, analysed, and intelligence and evidence extracted – something which Paul Williamson, at the time the NCA’s regional Head of Investigations for the Midlands, Wales and the West, believes would have been impossible without the use of the latest technology. “It was like standing at the end of a huge hosepipe and getting drenched with a massive amount of data in one go, ” explained Paul, now Programme Leader for Policing at the University of the West of England. “When that happens, if you haven’t got the capabilities immediately available, you’ve got to build them very, very quickly. “On Venetic we used AI and machine learning very much bespoke to that operation to actually enable us to extract and analyse the data. You might not have an off-the-shelf solution to deal with the issues you hit on in an operation of that nature. In the case of Venetic it was the volume and nature of the data that required us to adapt our approach so that we had the capability to receive the data, and effectively analyse it to ensure we were targeting the most harmful in accordance with our prioritisation strategy. “The use of link analysis has significantly increased over recent years, and is really, really helpful in serious organised crime investigations to help investigators understand the social connections between suspects or associates. It’s so important, given the fact that serious organised crime groups are much more fluid. We saw that on Venetic. ” Completing the data jigsaw Using data analytics to target that ‘fluid’ and expansive organised crime activity is also very much at the heart of the work undertaken by the Tackling Organised Exploitation (TOEX) team, led by Detective Chief Superintendent Kate Thacker. By providing intelligence and analytical support to forces undertaking complex investigations, and with staff based in regional organised crime units (ROCUs) across the UK, the team uses data science techniques to deliver evidence-based actionable packages for investigating forces. “Criminal groups exhibit poly-criminality, so we basically said ‘look, stop looking through a threat thematic lens, and start putting these resources together to develop organised exploitation, intelligence and investigations, regardless of the threat type’ , ” Det Ch Supt Thacker told Policing Insight.Paul Williamson, former regional Head of Investigations, NCA “The use of link analysis has significantly increased over recent years, and is really, really helpful in serious organised crime investigations to help investigators understand the social connections between suspects or associates. ” Paul Williams Former regional Head of Investigations, NCA

THE POLICE DATA CHALLENGE Copyright Policing Insight/Cognyte 2023“Because if somebody, for example, is a victim of county lines drugs exploitation, then if they’re not already a victim of child sexual abuse or exploitation, there’s a good chance they’re soon going to be, or are at elevated risk of becoming one. “The resources in ROCUs weren’t talking to each other. They weren’t sharing data, they weren’t sharing intelligence. They were often looking at the same offenders, and victims were represented across lots of different threat types. ” Whatever the nature and scale of the investigation – from carrying out a ‘data wash’ or analytical support on local inquiries, up to complex international investigations – having as many parts of the data jigsaw as possible helps the TOEX team to complete the puzzle for investigators. “We don’t have to have a magic marker in the sand before we get interested; we want to know about anything to do with organised exploitation, ” added Det Ch Supt Thacker. “And we want to help investigators, wherever they sit across the tiers of law enforcement, with the jobs that are beyond their capacity and capability that they currently have. ” Recognising the potential for AI and automation The potential for new digital technology to enhance the data analytics process has been recognised by forces in the UK and internationally, and across a range of crime types. For Former West Midlands Chief Constable Sir David Thompson, previously the UK lead for developing the National Common Intelligence Application (NCIA), it was the successful application of natural language processing that proved particularly effective in targeting modern slavery and human trafficking. “When you start applying natural language processing to key words, phrases, and things that might be associated with human trafficking, you pull out of the woodwork tons of intelligence that actually begins to show links, and a broader pattern, ” explained Sir David. “Some of that ‘proof of concept’ work – particularly on modern slavery and human trafficking – threw up tons more groups and clusters of individuals than we’d have ever seen on a traditional model. ” Many law enforcement agencies are still in the early stages of identifying how to maximise the greatest leverage from automation, AI and machine learning in terms of data analytics, but their potential is increasingly recognised. New Zealand Police has been something of a pioneer in the use of technology; James Foubister, who works as a Principal Adviser alongside the force’s Manager for Emerging Tech, Inspector Carla Gilmore, describes the forces use of AI and machine learning as “very, very nascent” . “There are bits and pieces, things in development, but we’re very early in that process, ” he explained. “There are huge opportunities for large language models to start ingesting and making sense of data, and then re-categorising using recursive things like machine learning and AI, or clever generative AI, to start actually generating answers instead of just generating results. ”Detective Chief Superintendent Kate Thacker

THE POLICE DATA CHALLENGE Copyright Policing Insight/Cognyte 2023In most cases the current uses of these new techniques are focusing on crimes that have already taken place, and identifying the criminals and their networks who have committed them. That holds true whether the technology is employed on large datasets to identify targets – such as in NCA’s Operation Venetic, or in the work of TOEX – or to help guide response action, such as the increasingly advanced decision intelligence platforms now in place in some forces, which have both data fusion and analytics capabilities. Importantly, the end result of all of these processes is usable data, advisory packages and insights which can empower staff who themselves may have minimal analytics knowledge and skills. But aside from the reactive use of these forms of technology, what of the proactive potential? Prediction and prevention “Every single serious case review you pick up, you will find that data sharing is a problem, ” said Thames Valley Supt Lewis Prescott-Mayling, who played a lead role in establishing the Thames Valley Together multi-agency platform as part of his work with the force’s Violence Reduction Unit (VRU). “But there’s less in serious case reviews about what actually needs to be done so that agencies are not blinded to those problems until someone has already come to harm. “I think we do keep reverting to type in policing, in that when we say we want an analytical product, it’s quite often descriptive of what’s already happened around, say, a crime profile, or a journey of a person, or a social network. “Very little of that is predictive, about how that’s going to affect the future, and we never really get to the stage of diagnostic, so that we can diagnose our tactical options, what we should do, and measure whether it has an effect.” Combining the “huge, huge, huge opportunities in fairly low-hanging fruit around robotic processing” , with “machine learning to identify text strings or keywords” could, he believes, change the policing approach. “We are developing what you could call artificial intelligence around a machine learning model. It’s using a random forest model at the moment to predict high-harm domestic abuse in the future.“When you start applying natural language processing to key words, phrases, and things that might be associated with human trafficking, you pull out of the woodwork tons of intelligence that actually begins to show links, and a broader pattern. ” Sir David Thompson former West Midlands Chief ConstableFormer West Midlands Chief Constable Sir David Thompson

THE POLICE DATA CHALLENGE Copyright Policing Insight/Cognyte 2023“We’re a Niche force, so we have our data coming from our source systems, and put that through a dimensional model, which is a way to structure the data so you can query it a lot easier; we then layer it with the data supplied from the other strategic partners. “We’re taking data from source systems – from social care, or a prison, or a probation service – which is the equivalent of their Niche system, and we’re layering it together in an automated way, so automated extraction, automated cleansing, and automated products, to produce insights. ” Using that predictive insight is one area where Chief Constable Lisa Winward, NPCC National Intelligence Portfolio Lead, and Vicky Hough, the Intelligence Portfolio Staff Officer and a former analyst, argue that “intelligence hasn’t fully reached its potential yet” . “At the moment it’s very reactive, in terms of intelligence being used in investigations, ” said CC Winward. “I do think that’s where we’re going, that we will be able to risk assess somebody’s potential to commit an offence and wrap around an element of prevention. ” However, they are both aware of the potential pitfalls of failing to fully explain such approaches. “The wording and the commentary around this is absolutely critical, ” explained Vicky. “Preventing something happening in the first place [through] early intervention is what the public need to understand – and what that ambition is – as opposed to machine learning and the power of AI in terms of trying to predict patterns. “Because people read it as ‘predicting X is going to do Y at this time in this location’ . It’s about how do we embed a whole systems approach to actually stop something happening in the first place?” “And wouldn’t it be great if we could intervene early?” continued CC Winward. “I’m talking about utopia here, but unless you think of utopia, you’ll never get there, will you? “Wales, as a collective, did a lot of work on adverse childhood experiences, using them as a predictor to which children would be at risk, or enter the criminal justice system, or be vulnerable of sexual exploitation. “So there are a huge number of predictors there. What’s missing at the minute is us intervening at the right time – and not necessarily policing, but the public service system intervening, knowing that all those predictors are there to stop that child ending up in that situation in the first place. ”“Intelligence hasn’t fully reached its potential yet. At the moment it’s very reactive, in terms of intelligence being used in investigations. I do think we will be able to risk assess somebody’s potential to commit an offence and wrap around an element of prevention. ” CC Lisa Winward NPCC National Intelligence LeadNorth Yorkshire Chief Constable Lisa Winward, the National Police Chiefs’ Council (NPCC) Intelligence Portfolio Lead