“Just Right” Policing: a Job for Science

Can police develop the concept of “signal failures” to build a science of “just right” policing, learning from the mistakes of both “under-policing” and “over-policing”? Qualitative documentation of “signal failure” events across the history of Anglo-American policing that have generated widespread public disapproval of police actions or inactions. This communication analyzes and illustrates the potential value of applying scientific methods to signal failures, as a potential source of learning from past failures to prevent future disapproval. There are at least four dimensions for pre-mortems against signal failures: quantity of officers present at an incident, quantity of force used by police, quality of choices police make in deciding what to do, and quality of how well police do what is decided to be done. Signal failures can be collected historically and contemporaneously in large police forces, or across multiple smaller forces at a state or provincial level. They can be compared to samples of similar events that did not become signal failures. Even a simple case-control comparison can be a method for predicting which dimension of an event might be most likely to cause a signal failure. Yet building such data bases requires that police records systems become easier to access, so that both internal and external research can be done to enhance “just right” policing. Like many rare events, it is convenient to dismiss signal failures as “flukes” with little potential for prevention. But like airplane crashes and house fires, rare events can be made even rarer. The value of the scientific method cannot be dismissed without testing its application to these major threats to police legitimacy. “Just right” policing may depend on it.


Introduction
From the assassination of former Prime Minister Shinzo Abe in Nara, Japan to the slaughter of school children in Uvalde, Texas, police around the world were widely blamed in 2022 for under-policing, defined as failures to prevent harm.
From the 2020 murder of George Floyd in Minnesota to the 2021 Vigil for Sarah Everard in London, police lost legitimacy for over-policing, defined as causing harm disproportionate to the harm they were trying to prevent.
As the pace of both kinds of "signal failures" 1 seems to increase across the digital world, many good citizens shake their heads and ask of policing, "Why can't they just get it right?".
The truth, of course, is that police do get it right, most of the time, around the democratic world, millions of times an hour, on routine business (Sherman, 2022). That fact is not newsworthy. It has always been the rare event that gets police in trouble. The same is true for airplane crashes, fires, automobile collisions, and other examples. Yet each of these rare events becomes predictable-and preventable-with a long enough time frame (Perrow, 1999). Rather than blaming "bad luck" or an "act of God," scientists in recent decades have massively reduced deaths from commercial air travel, household fires, and car crashes.
Their success derives in part from abandoning the attractive idea of a "good enough" decision in favor of a "just right" decision. Both approaches may have their place, in the right context. As a new parent, I was comforted by the concept of a "good enough parent," as developed by English pediatrician Donald Winnicott (1971). This idea assured me that I did not need to be a perfect parent who gets it right all the time. Mercifully for me, his standard was to be doing enough parenting to make my children feel secure and loved. That logic is still appealing, but it does not work for every parent, or every child. Nor can it be generalized to such complex institutions as the police.
As the reforming Metropolitan Police Commissioner, Sir Robert Mark (1978: 162) wrote a half-century ago, "Of all the guardians of freedom the police are now in practice the most important." His claim referred to the endless need for finding just the right balance (Neyroud, 2017) between freedom of the people from interference by the state and freedom from crime and violence by fellow citizens.
Commissioner Mark's claim is arguably supported by rising public expectations for police performance, which will not allow for "good enough policing," despite the strong case of that idea made by criminologist Ben Bowling (2007). As Peter Neyroud (2017) has observed, the growth of science in policing has now made it possible to move from "satisficing" to "optimizing," not just in rare events, but in everything police do. That is arguably how science has helped to tackle COVID, as well as fatal car crashes and fire deaths (Perrow, 1999).
Yet policing remains far behind other fields in its application of science. That is not for lack of science to apply. It is, rather, the result of scientific knowledge about policing growing at a rate that is faster than its application in policing. The lack of application, in turn, results from insufficient communication in police training, as well as in the evidence-free, unscientific design of policing systems. The latter point can be illustrated by this fact: some major police organizations have long employed a Chief Technology Officer, but none to my knowledge have created a position for a Chief Scientific Officer. Police analysts, when they work as in-house scientists, suffer from subordination to technology, which determines what kinds of science may or may not be possible. Where stop and searches are done, for example, is unknowable in a technical system that does not currently ask for a street address where each stop occurs. What (or whether anything) is found during a strip search of a suspect in custody, for example, cannot be analyzed scientifically without a software system that includes a requirement for entering those data.
While science may be necessary (if not sufficient) for police find their way to "just right" policing, a major culture change in policing may be needed to assign technology to the aid of science. It is only when police leaders accept science as a way of knowing that they can accept the role of science, rather than technology, in a driver's seat. Yet accepting science as a requirement of "just right" policing is arguably their duty.
As the most important guardians of freedom, the police duty is to apply the best science to every decision, so that police can strive to make "just right" decisions as a matter of routine, including rare events.

What Is "Just Right Policing?"
As the theme of the 14 th Cambridge International Conference on Evidence-Based Policing, "just right" policing is a broad concept that applies to both quantities and qualities of policing. Some dimensions of this idea were developed in my August Vollmer Award Lecture for the American Society of Criminology, which was published this year in the Criminology and Public Policy journal (Sherman, 2022) and is available on video at https:// www. youtu be. com/ watch?v= NXcLy LFAhTg. The main emphasis in that Lecture is on the role of the three "Ts" of evidence-based policing: targeting, testing, and tracking in a continuous effort to keep patterns of policing "just right." Other dimensions can be added to those concepts, for a more systematic approach to tracking whether policing is "just right." That kind of tracking, in turn, can be discussed in any community, with any police agency, as a checklist of enduring questions about any particular event, in four categories: • Quantity of officers, • Quantity of force, • Quality of choices police make in deciding what to do, and • Quality of how well police do what has been decided to be done.
These categories can be simplified as four questions about any incident that did (or a matched incident that did not) develop into a signal failure to provide "just right" policing: 1. Did police assign just the right numbers of officers to be present at the places where they were needed most, to prevent the most harm, and to detect or arrest the most harmful people? 2. Did police use just the right amount of force to prevent even greater harm in every situation: no more, and no less, than was necessary to prevent harm as great or greater than police force itself? 3. Did police use just the right selection of tactics or actions, based (if possible) on prior testing, to reduce maximum harm in the short run as well as over many years to come? 4. Did police deploy the best tactics in the best way, or were there errors of omission or commission in the course of deployment?
The importance of asking these questions whenever a "signal failure" occurs is to learn from failure (Syed, 2015). Whenever something goes badly wrong, those four questions can be a starting point for what the diagnosis or post-mortem must be. Signal failures-those events which attract a measure of negative publicity and protest above a measurable threshold-can be assembled and categorized as to which of the four questions were most relevant. A further analysis of the sources of error can then be launched, with priority given to the most common errors-or even those errors with the highest harm.
All of these questions about the right practice might be separable from questions about the right people, whose misconduct may be totally extraneous to the issue of under or over-policing, as in posting police photos with murder victims online or officers using misogynistic language in private texts. The questions about specific officers may also be part of the core four questions, as in explaining why Patrol Officer Derek Chauvin was allowed to continue in a job where he applied criminally excessive force on George Floyd, after nearly annual citizen complaints against Chauvin for 18 years, including shooting people.
The key point is that there can be a science of signal failures bringing disrepute upon policing. That science, like the science of automobile crashes, can provide a useful platform for systematic post-mortems (Perrow, 1999). Such a data base can, in turn, produce hypotheses about risk factors that can be controlled through better design of policing systems for recruitment, training, supervision, and digital tracking of officer behavior.
So, there can be such a science. Whether there will be such as science depends on the kind of people who are reading this essay. It also depends on whether police organizations can reform their overall strategy for collecting and retrieving the kind of data needed to support that science.

Data Collection: a Central Source Strategy
One challenge to scientific analysis of police decisions is the development of "archipelagoes" of police agency data sets. Unlike the era of mainframe computers, in which all data came from single machine operated by a single data processing team, the current era of individual data processing gives rise to diverse data "ownership:" they who collect the data, own the data-and often say "you can't have it!" The refusal to share data sets within large police agencies is supported by data protection laws and complex duties of care that may prevent police executives from getting their questions answered. Any question that requires data owned by two different units may not be answered easily in the short run. What would seem to be routine can become a battle for control, not unlike combat among inhabitants of warring islands in an ancient Greek archipelago.
If, for example, police in Minneapolis had wanted to predict the highest risk officers for killing a citizen illegally (Sherman, 2018(Sherman, , 2020, they might have faced the following challenge. Predicting police officer risks prospectively requires data on every officer ever hired. It also requires information on arrests made by each officer, complaints made against each officer, use of force reports, absenteeism records, responses to computer-aided dispatch (CAD) calls, injuries on duty, and even annual performance reviews. All of these categories of data are likely to be found in different data systems-if they are digitized at all. Internal Affairs (or Professional Standards in the UK) would have some data, Human Resources would have others, the police control room may have still more. Each system may use different commercial software, and data systems may be very hard to link. Separate machines may not "talk" to each other.
Even if every "data-owning" source gladly yielded the data for every officer ever hired, how much time would it take to bring together these diverse data sets?
A lot. Which explains, in part, why there is, as yet, no science of signal failures. The facts behind each incident (or person) are hard enough to get. A comparison group of similar incidents (or people) who did not generate signal failures is even harder-but just as necessary as facts about the officers who attract infamy for a police agency.
The long-term solution to building data technology to serve science is to do what the UK Home Office (analogous to the US Department of Justice) has done. The Home Office has created a Chief Scientific Advisor (CSA) who supervises all of the ministry's technology projects. The exact title is "Director-General, Science, Technology, Analytics, Research and Strategy (STARS), Home Office." There are experts leading each of the STARS teams, with hundreds of people supporting this agency of some 35,000 employees. Yet they are all coordinated by the single Home Office Chief Science Advisor, who can promote integration of these resources to support the strategic objectives of a broad range of national security operations.
Reorganization of large police agencies alone is not enough, at least in the short run. But as digital systems develop greater capabilities, reorganization can become the solution in the long run.
The best time to plant a tree, it said, is 20 years ago. The best time to let science drive technology and analytics in policing is now. It is hard to see how else we can build a science of "just right" policing.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.