In December 2013, Gillian Routledge, a police analyst in Durham, received a surprising phone call from her boss. “Do you fancy a master’s from Cambridge?” Mike Barton, then Durham’s police chief, asked. It had been almost two decades since Routledge had last studied at university and, now in her 40s, she packed her bags and set out on a four-hour journey to an institution which was unlike anything she had experienced, with black gowns and gilded dining halls.
Barton was on a mission to train his police to think scientifically and use rigorously gathered evidence to decide how best to limit crime. At Cambridge, Routledge would help to design one of the UK’s most radical policing experiments as part of her criminology master’s thesis. The experiment would see hundreds of people arrested for low-level offenses ranging from assault and burglary to fraud and child neglect have their futures decided by chance.
Half of them would face traditional prosecution, while the other half would be offered a four-month deferred prosecution programme called Checkpoint. Instead of prosecution, this group would be given personalised treatment plans and agree to conditions such as meeting their victims, working with psychologists or undertaking community work. Routledge and colleagues wanted to see if an alternative to criminal prosecution could reduce reoffending.
In spring 2015, Durham police began testing Checkpoint, and in summer 2016 they launched a 20-month trial. This was a randomised controlled experiment, the kind of scientific trial typically used in medicine to test new drugs, where groups are randomly given different treatments to assess their efficacy.
The trial was targeted specifically at people who were predicted to be at “medium” risk of reoffending. Officers identified potential participants with the help of a machine-learning algorithm called the Harm Assessment Risk Tool (HART), designed by Cambridge professor Geoffrey Barnes, Routledge’s thesis supervisor, which forecasts how likely someone is to reoffend in the next two years based on 34 variables ranging from age to criminal history. When they found someone eligible for the trial, they decided whether they should be offered Checkpoint, or face conventional prosecution using a computer randomiser – akin to electronically flipping a coin.
There were initial challenges. Some police officers struggled with the arbitrariness of the randomiser. When they knocked on Barton’s door complaining it was unfair to shuffle people’s fates for an experiment, he would patiently tell them they needed to think long term – to seek evidence about better ways of policing.
For the police running it, Checkpoint is a clear sign that not prosecuting certain people can in fact reduce cycles of crime. Final results were still under peer review at the time of publication, but according to Durham police data, 15 per cent of people referred to the Checkpoint experiment reoffended during the four-month rehabilitation programme and around 37 per cent in the following two years – compared, respectively, with 26 per cent and 47 per cent of people in the control group. More than 2,600 people have now completed Checkpoint, and for every 1,000 people who pass through it, Durham estimates it saves £2 million in reduced crime. Several other UK police forces are now rolling out similar initiatives.
Checkpoint is among dozens of policing experiments spawned globally in the past decade in a small department of the University of Cambridge called the Jerry Lee Centre of Experimental Criminology. The centre is run by Lawrence (“Larry”) Sherman, an indefatigable American professor who believes that unleashing trials live on our streets can radically transform policing. Aged 71, Sherman begins his days with a very specific morning exercise routine which could challenge many people half his age; he claims to do 300 push-ups daily (“in batches of 100”), followed by a 6km run.
Over the last 50 years, Sherman has worked on trials around the US, UK, Australia and beyond. His field, called evidence-based policing, aims to bring a scientific approach to the foundation of policing – with rigorous experiments to test innovative ways to reduce crime and adapt policing tactics based on constantly-evolving evidence. Each year he trains more than 100 senior officers from around the world, who come to study his criminology master’s course. He also helps to design and analyse experiments run by students and colleagues. His department’s recent trials have included comparing patrols of police civilians (community support officers) with those of fully-fledged officers, and randomly allocating police armed with Tasers on different shifts in London.
In Sherman’s view, police too often act on instinct and hunches rather than on the basis of evidence. He describes policing like medicine of the past, before the field was reshaped by randomised trials. Sherman has accused British police of treating computers like “electronic filing cabinets” and feels officers over-rely on arresting people even when this is counterproductive. “The police are responding to too many cases where they’re not needed and therefore aren’t available to proactively identify who is at greatest risk of serious harm and to intervene with those victims, offenders and places,” he says. “So if we could stop doing a lot of the senseless errands [that officers run] and start investing much more time in planning where police can do the most good based on statistical analyses, we would have a much safer population as well as a healthier one.”
Sherman was born in New York in 1949 to a Baptist minister mother and a father who worked for the YMCA. His parents raised him with a commitment to social justice and community activism. He recalls them joining Martin Luther King’s 1963 march to the Lincoln Memorial, where King called for an end to racism and economic and civil equality. Sherman’s teenage years were a period of particular racial tension in America, including police brutality towards civil rights activists and the failure to protect Black communities who were lynched and attacked. The teenage Sherman was captivated by the police’s potential to defend society’s vulnerable, but also their capacity to maltreat and abuse people when their power is unchecked.
When he finished university in 1970, Sherman took up a national fellowship with the New York City police department. His first job focused on developing neighbourhood policing strategies, and he also joined sting operations to target corrupt officers. As violent crime rose around the US in the 1970s and 1980s, a view spread that police were powerless to stop crime – a philosophy known as “nothing works”. Sherman didn’t agree. Policing wasn’t condemned to fail, he believed; they were simply going about things the wrong way.
A few years after finishing a PhD in sociology in 1976, he got a lucky break. His former boss from New York, Tony Bouza, had taken charge of the Minneapolis police department and let Sherman start his own “laboratory”, running a series of trials across the city. (The same police department is now embroiled in charges of endemic racism; former Minneapolis officers are charged in relation to the death of George Floyd in 2020.)
In one of the first trials Sherman ran, launched in 1981, a group of officers was asked to use different approaches to misdemeanour domestic violence crimes: arrests, mediation or sending suspects away from home for a few hours. Finding that those who were arrested reoffended less over the next six months than the other groups, the experiment helped to drive a slew of new laws requiring mandatory arrest for domestic violence across the US.
Besides domestic violence, Sherman spent many of his early years investigating the mysteries of crime concentration. In an early 1990s trial targeting illegal guns in Kansas City, Missouri, he identified an 80-square-block area, called beat 144, with a homicide rate 20 times the national average and a high number of driveby shootings. In a bid to cut shootings, four officers spent around five hours a night over 200 nights patrolling, including stopping vehicles where there were legal grounds for doing so. Following the trial – which wasn’t a rigorous randomised experiment – there were 49 per cent fewer gun crimes in beat 144, even as crime remained at high levels across the rest of the city, while local community surveys suggested people were less fearful of crime. When the results were published, Sherman recalls receiving calls from police around the US interested in testing similar approaches. Writing up the trial, however, he noted that intensified patrols like those in beat 144 – whose residents were overwhelmingly from minority backgrounds – cannot be arbitrary and should be carefully considered.
In the late 1990s, on a flight to launch an experiment in Australia, Sherman had an epiphany. He began reading a book he’d stumbled across in a local shop, about the rise of evidence-based medicine: how running randomised experiments, ranking hospitals by death rates and pushing doctors to follow research had revolutionised healthcare. “That’s exactly what we need in policing,” he thought to himself. Shortly afterwards, in 1998, he delivered an impassioned speech to a Washington think tank, announcing the birth of evidence-based policing.
On a sunny Thursday in September 2020, I join Sherman in Cambridge. In a large lecture hall, a few dozen socially-distanced police officers – some dressed in hoodies, others in shirts – attentively take notes as Sherman and colleagues speak about their latest experiments and the principles of evidence-based policing. Stacey Rothwell, an officer from Kent, joins Sherman on stage to explain her master’s thesis: a randomised experiment with officers responding by phone, rather than in person, to certain non-urgent 999 calls. Such calls, she says, would be crimes that would not necessarily elicit an immediate physical response: they have included reported thefts, sexting offenses and a case of someone threatening to burn down a victim’s home and kill them (victims are given a choice over participating in the experiment).
After Sherman speaks, he walks around the room offering a microphone for students to question and comment; all seem to agree that policing needs to be more intelligence-led, more proactive and founded in evidence. “I just feel that our general approach is pretty imprecise,” one student declares. In the coffee break, officers and professors elbow-bump one another for Covid-friendly greetings. One London officer I speak to, who has recently started Sherman’s course, says he is enjoying it but is disappointed to be missing a major raid by his team that morning; colleagues found guns, drugs and two suspected victims of slavery.
After lunch in Sherman’s Cambridge college dining room, I meet some of his students in the college garden, where we speak mainly about race and policing. Since I began reporting on Sherman in early 2020, protesters around the world took to the streets in anger at longstanding police killings of Black Americans including Breonna Taylor and George Floyd, but also at policing more broadly – at how officers too often compound, rather than alleviate, society’s racial inequities. In the UK, the House of Commons Home Affairs committee has launched an investigation into racial disparities in policing, including allegations of institutional racism.
Justice Tankebe, a Cambridge professor focusing on police and state legitimacy, says that calls in different countries to reform the police aren’t necessarily in conflict with evidence-based policing. (Some, however, are calling to abolish policing as we know it entirely.) Tankebe argues that many critics want policing to be less arbitrary, for officers to work with communities to gain legitimacy, and for them to protect minorities. “The fundamental issue is that the police have failed to respond to peoples’ legitimation expectations,” he says. “People are saying, ‘It appears the police are becoming part of the problem rather than the solution.’ So we need to reimagine the kind of police service we have.”
“If you look at it that way,” he continues, “evidence-based policing may even be part of the answer to critics who call for defunding police.” For Tankebe, this is a moral obligation: to show which policing methods are helpful and which are causing more damage to the people police are supposed to serve.
Sherman hopes that public outcry will lead to far-reaching changes for policing. As he sees it, officers spend too much time rushing out to crimes that have already happened, or attending minor offenses – like George Floyd’s alleged possession of a counterfeit banknote. Instead, they should focus on preventing high-harm crimes through prediction and targeting. Just as doctors have the tradition of the Hippocratic Oath, vowing to “first, do no harm” and to care for patients, he believes officers should be schooled in Hippocratic policing: only using interventions such as arrests or searches where there is rigorous evidence it will prevent harm. “We want to have as little policing as possible,” Sherman says. “Like we want to have as little surgery as possible. The question is: when will we kill people by not giving them surgery?”
In 2014, following the fatal shooting of Michael Brown, a Black teenager, by a white police officer in Ferguson, Missouri, Sherman was part of Barack Obama’s Task Force on police reform. The officer involved was eventually judged to have acted in self defence, but one of Sherman’s suggestions after the incident was that US states create independent inspectorates with the power to punish officers and departments that had been identified as abusive. His ideas were not adopted, but recently he has pushed for reforms including turning predictive policing on to the police themselves, using algorithms to identify rogue officers. He says he is deeply motivated to use science and technology to make policing fairer, “taking a long view that is steeped in the history of the racist societies in which all modern policing has [operated].”
Cynthia Lum, a former Baltimore police officer who now runs the Center for Evidence-Based Crime Policy at George Mason University, says that recent decisions by some US police departments to restrict chokeholds and neck restraints are “low-hanging fruit”. Many problems with policing are more difficult to address, such as how to create real accountability and how to prevent crime without fueling racial disproportionalities in the criminal justice system.
Sherman also points out that while some minority communities are overpoliced, they are simultaneously underserved by police and the criminal justice system – they are disproportionately victims of crime. In a paper published in November 2020, Sherman and colleagues analyse national data to show that the most recent (2018-19) homicide rate for 16-24-year olds in the UK is 24 times higher for Black than for white people. A fairer policing system, he argues, would need to help address such disparities.
As Tracie Keesee, senior vice president of the Center for Policing Equity, sees it, the goal is to find “the sweet spot where we’re not overburdening one community, yet we are providing [services] differently” in different places. Dismantling the police could hurt the poorest communities, including many minorities. “I’m not satisfied with just saying that’s the collateral damage,” she says. Keesee is calling to rethink the role of officers, acknowledge racial injustices, and spread new models of public safety informed by local people.
Experiments, evidence and technology alone are never going to solve the kinds of racial and economic inequities that persist across society – and arguably the police only have a limited role to play in this fight. Lum and others in this field are confident that, when carefully used, evidence-based policing’s toolbox – including experiments to understand which policing tactics work and which don’t and qualitative analysis of what communities want – is important for deciding a new model of policing. One thing she and colleagues are sure of: policing that focuses on arrests and force as its main tools is not the way forward.
Keesee emphasises that research needs to be coupled with action and political will. For several of the communities she works with, there is impatience and pain. “They’re not interested in yet another study,” she says. “We know what it’s going to take to do this,” she adds, referring to the need to reform policing, invest in deprived areas and focus on serious crimes. “Really the question is: are we going to?”
A core tenet of Sherman’s approach is detecting and focusing on crime “hotspots” – areas with high levels of violent crime that can be as small as a single address or intersection. Sherman’s colleague and former British police chief Peter Neyroud says that policing hotspots requires “keyhole surgery” and that most police are “still cutting the whole body open” in their approach: “They’re not scientific about it, quite often they’re wrong, and they frequently don’t know the real high-crime places.”
Sherman believes that over-patrolling wide areas hasn’t only cost the police legitimacy, but is also an ineffective way to stop crime. Many, including a senior minority officer in the US I spoke with, see the widespread use of tactics like stop and search as racist and illegal. By focusing instead on small hotspots with a variety of different approaches, Sherman believes police can limit indiscriminate profiling as well as reduce violent crime. Tankebe cautions that data for identifying hotspots mustn’t come only from police themselves, since this creates a “self-fulfilling prophecy”: some areas may appear to have higher levels of crime not because more offenses are necessarily being committed, but because they have been over-policed.
In the UK, Sherman has begun applying his hotspot approach to knife crime, which rose by seven per cent in 2019 in England and Wales to the highest levels recorded. In 2019, he supervised one of his Cambridge students, a London homicide detective called John Massey, who pinpointed the locations of thousands of non-fatal knife attacks around the capital in one year and of lethal stabbings the following year, in search of patterns. They discovered that the vast majority of knife murders happened in neighbourhoods that had seen at least one non-lethal stabbing the year before. Clusters of streets which recorded six or more injuries the first year were 15 times more likely to see a homicide in the second year than areas which recorded no knife assaults (the majority of the city).
National newspapers covered the work with dramatic headlines suggesting a new way to tackle Britain’s violent crime epidemic. Massey has since trained London police to better understand where knife assaults are concentrated, but for Sherman, this isn’t enough. To put the research into practice, he wants an experiment with randomly surging patrols in different hotspots over different time periods, or to test other approaches such as chaperoning children to schools or installing knife arches at school entries in the high-risk neighbourhoods, to see if this can reduce deaths.
Sherman is perhaps most enthusiastic about the counterintuitive prospect that policing less, rather than more, could reduce crime. He sees powerful potential for reducing hostility in over-policed neighbourhoods, as well as helping forces save money. Several of his department’s recent experiments have been centred around this idea. They include a trial across London train stations published in early 2020, in which a pair of officers were assigned to patrol half of London’s most high-crime Underground platforms in 15-minute bursts; reported crime dropped by 28 per cent on these platforms, with reductions occurring mostly when the police weren’t actually present. Sherman and his team call this a “phantom effect”, with a brief appearance by police officers resulting in a residual benefit after they leave.
One of the biggest tests of “minimalist policing” took place over 248 days in western Australia in an experiment published in summer 2020. It was run by Sherman’s colleague Geoffrey Barnes, with Australian police. The experiment gave officers GPS-enabled phones to track their movements. They were randomly assigned between three and seven neighbourhoods to patrol each day, from a list of 15 high-violence districts in Perth; the remaining neighbourhoods were left without extra officers for up to 20 days. After one mere 15-minute patrol (or “treatment day”, as Sherman calls them), there was a notable drop in crime for four days afterwards – suggesting that the threat of a police visit may be as much of a deterrent in some cases as another patrol. (The effect doesn’t last indefinitely: on the fifth day, crime would surge by an average of 66 per cent.)
The next stage of this research launched in October in Bedfordshire, UK. In a project managed by Michelle Leggetter, another alum of Sherman’s course, police are working with the Cambridge Centre for Evidence Based Policing – Sherman’s for-profit company, which he runs with Heather Strang, his colleague and wife, and which sells products including consultancy and online tutorials – to run a new experiment. For three months, Bedfordshire officers will patrol seven neighbourhoods on foot for around 20 minutes each day; the list will be shuffled randomly daily, from a group of 21 neighbourhoods with high rates of serious weapons crimes involving young people.
The team want to find the optimal gap between police patrols – an equivalent to Perth’s four-day “sweet spot” showing how frequently (or rarely) they should visit neighbourhoods to reduce violent crime. They will run the experiment’s second phase in 2021: an alarm system linked to police radios will alert the control room to dispatch officers to neighbourhoods when the “sweet spot” period is about to expire and officers haven’t visited.
If they get similar results in Bedfordshire and elsewhere, Sherman is hopeful that the notion of policing less often, in a more targeted way, will proliferate. One potential objection is that people seeking to break laws could learn policing patterns, but Sherman insists the patrols can be sufficiently randomised so that they can’t be guessed. Matthew Bland, a Cambridge researcher working on the Bedfordshire experiment, hypothesises that targeting crime hotspots in this way could reduce harm from serious assaults by up to 20 per cent across the county.
Sherman sees his work as a kind of religious mission; he likens himself to a priest delivering sermons, viewing evidence-based policing as a moral crusade. Some senior police colleagues see him in similar terms. Sara Thornton, who previously headed the UK’s National Police Chiefs Council and has sat with Sherman on an informal Downing Street group advising on the future of policing, calls him the “John the Baptist of evidence-based policing” because of his relentless messaging. She says his work has had far more impact on changing policing from within than senior officers’ own attempts to spread evidence and science. (She later questions her metaphor, remembering that John the Baptist ends up with his head served on a platter.)
John the Baptist may not be too far off, given Sherman’s potential to provoke. Some colleagues accuse him of proffering facile solutions to complicated questions. Mike Hough, emeritus professor in law at the University of London and founder of the Institute for Criminal Policy Research, has guest lectured Sherman’s students but cautions about the limits of evidence-based policing. He says that Sherman is “a force for good and a very clever guy”, but that he “has for reasons I just don’t understand oversold randomised controlled trials.” While such trials might be helpful for answering some straightforward policing questions, Hough argues, many solutions to crime reduction can’t be easily captured by experiments. Hough is also skeptical about the comparisons with medical trials, and suggests that the effectiveness of a drug is easier to measure than the complexities of human behaviour and getting people to comply with laws.
Sherman’s work on domestic violence remains among his most controversial, and shows the limits of individual experiments. His subsequent trials have suggested that while arrests may reduce repeat offending among employed suspects, they can fuel further offenses among unemployed suspects. Arrests also seem to harm victims. In 2014, Sherman ran a 23-year follow-up to an earlier experiment and found that partners of those arrested on domestic abuse charges were 64 per cent more likely to die early – from stress-induced conditions including heart disease – than people whose partners were warned and allowed to stay at home. African-American domestic abuse victims were 98 per cent more likely to die early when their partners were arrested.
The example of domestic violence also demonstrates a wider challenge for Sherman’s field: how to spread evidence-based conclusions when they clash with political and public opinion – when people feel that a particular approach is required even where it doesn’t necessarily reduce crime or lead to the best outcome for victims. Sherman has pushed for changes to mandatory arrest laws for domestic violence in the US and UK for decades with little success. “When you talk to police chiefs even today, with all the subtlety and complexity, they still feel vehemently that they don’t care about the research,” says Linda Mills, a domestic violence expert and public policy professor at New York University who has worked with Sherman on randomised experiments. “They are clear they are doing something [for victims] and they are doing the right thing in their community.”
Sherman will have to convince more police people as well. While some officers are convinced by his promise of replacing gut instinct with evidence, others feel he dismisses the human dimensions of policing and scorns officers’ experience. His colleague Barnes recalls that during a large talk Sherman once gave to Australian police, many officers were enthusiastic, but a substantial group seemed furious. Barnes heard one officer turn to another mid-lecture and joke about attacking Sherman, saying “when do we get our steak knives?”
When Sherman’s team have tried to run experiments, they’ve at times encountered challenges: British police pressing randomisers repeatedly to get the answers they want, for example, or officers in Trinidad reluctant to turn on their body cameras during experiments, seeing the request as an intrusion. Some of Sherman’s students are dismissed as elitist, or mocked when they try to push evidence in their forces. “We don’t want to hear any more about your Cambridge experiences,” one British officer recalls a colleague saying during a meeting.
Even police who support Sherman’s experiments recognise that they can carry risks. There’s the chance that people could be injured or killed in neighbourhoods with fewer patrols during randomised trials, as well as concerns over the justice of experimenting on certain areas and not others. Sherman is all too aware of the dangers, both physical and political, noting that police chiefs are sometimes scared by his suggestions and that explaining them publicly can be challenging. “Why are you allowing those people to die?” he says, imitating questions people might ask of the police. “Oh for this experiment, some crazy professor at Cambridge wants us to do it.”
There are also fears around the technological aspects of some criminal justice experiments. One is that predictive algorithms can be opaque, unaccountable and may perpetuate social and racial biases – especially when they are trained using historical police data. A 2016 ProPublica investigation showed that an algorithm used in courtrooms to help guide judges’ decisions in several US states discriminated against Black people. Some civil rights groups and academics were fearful that the HART algorithm in Durham could discriminate against people living in poorer areas, prompting postcodes to be removed from the data set.
Barnes, who designed HART and has helped build forecasting tools to predict fatal car crashes and serious domestic violence, among other things, acknowledges that algorithms use millions of data points and it may be hard to understand their decisions. But he believes police officers’ own biases can be harder to change, and that using carefully-designed algorithms can provide accurate and consistent predictions to guide human decisions. “One real frustration in the algorithmic sphere is everyone wants to compare the results to perfection,” Barnes says. “Wait a minute, we didn’t start with perfection, we started with human judgment.”
Sherman believes that medicine’s history has instructive tales for the future of policing. As Sherman has pointed out, Ignaz Semmelweis, the doctor who discovered that hand-washing could reduce deaths in childbirth, was ridiculed and lost his job. The American College of Surgeons felt so threatened by the first performance rankings of US hospitals in 1919 that they immediately scorched the report in the furnace of the New York hotel where they had convened. “Science is like that,” says Barnes, who recently started a new job as deputy director of the Metropolitan Police’s strategic insights unit. “It starts out with lots of petty jealousies. The police are maybe, at best, 20 years into this, and that is being generous.”
Sherman is playing the long game. He’s confident that his work chimes with the scientific currents of our age and remains the most effective path to a better, more just policing system. He plans to spend many more years on this quest, despite having recently celebrated his 71st birthday. “Eighty is the new 50,” he says. “I feel like I just got started. There is so much to do and more opportunity to do it than ever before.”
More great stories from WIRED
💉 A universal coronavirus vaccine could stop the next pandemic
🍩 Can The Simpsons replace characters with deepfake AI?
📱 Looking for a new phone? These are the best smartphones for any budget