Implementing this approach has required navigating an array of challenges. The first involves data collection. To meet RIPA’s requirements, LAPD had to gather data across its sizeable force and on an array of variables for each stop (including the setting, the officer’s perception of the person stopped, the reason for the stop, actions taken, property seizure, contraband evidence, and the result). In addition, LAPD had to ensure that it was collecting the right data. This proved difficult because an officer would sometimes stop someone for one violation (e.g., speeding), identify another problem (e.g., the person did not have a license), and erroneously write down the second infraction as the reason for the stop. LAPD recognized this and has begun training officers on the correct approach. Nonetheless, this embodies a tricky dynamic. “We need to make a system,” Rhodes explained, “that we are building while we are flying it and while everybody else can see how we are building it. And that’s one of the big challenges that we are having. Nobody likes to make a mistake, but in order to get a better product, we may make some in the process.”

Even when data is collected successfully, the process of interpreting and analyzing it can be complicated. For instance, LAPD discovered that among people who were stopped, officers were more likely to have taken action against a white person, whereas people of color were more likely to receive a warning or have no action taken against them. To many officers, the data obscured how there were many situations in which they were exercising discretion appropriately. For example, an officer may stop a vehicle that made a turn without signaling. But, during the stop the officer may form the belief that the driver has been made aware of their violation, learned that they should signal in the future, and can ill-afford a costly ticket. The officer may then use discretion in not issuing a citation.  However, with the numbers alone, the public may have and has had a different interpretation. They assumed officers were pulling over white people when they were doing something wrong (and therefore would receive a citation) but only pulling over people of color to harass them. “So numbers might mean one thing to you,” Rhodes noted, “and something [different] to someone else.” This reflects how data “can be a doubled-edged sword.”  

While these conflicting interpretations can be difficult, they can also lead to positive change and learning. After acknowledging the dual views of the data, LAPD implemented a new policy on pretextual stops that balanced the different perspectives and represented a critical first step in building greater community trust. The policy affirmed that stopping vehicles is important for public safety, defined a pretextual stop, and limited the conditions under which a pretextual stop is permitted to situations where officers “are acting upon articulable information in addition to the traffic violation.” Furthermore, the department asked officers to use body-worn cameras while articulating the safety reason for the stop and emphasized the importance of procedural justice. Finally, LAPD integrated additional data on pretextual stops into a RIPA dashboard, which is part of LAPD’s COMPSTAT process and provides an opportunity for Rhodes’ team to engage in dialogue with captains about what they think is happening. “We have moved from the visual,” Rhodes said, “to the visceral.”

This illuminates how LAPD—which is in the process of training officers on these tools and techniques—may be building the plane while flying but has also started to leverage data to stimulate a dialogue about how to change officer behavior and improve outcomes. LAPD, as Patton suggested, is drawing on the ingenuity of its people—both within and outside of the department—to learn, grow, and build trust.

Be Data Curious: Leveraging Learning and Analytics at the Seattle Police Department

The Seattle Police Department (SPD) has similarly been on a journey to leverage data to foster discussion and learning that can lead to improvement in services and officer behavior and to better outcomes. Becca Boatright, SPD’s Executive Director of Risk Management and Legal Affairs, thinks of this as a continuous circular process in which the department asks insightful questions, collects data to test its hypotheses, and develops new queries after analyzing the information. “In a perfect world,” Boatright explained, “the research, the data governance, the data warehousing, and then the analytics that flow from there are not designed to answer a question and move on. They’re designed to drive curiosity and identify what additional questions need to be asked in that area.”

In an effort to create this virtuous cycle, SPD has invested heavily in its analytical and data capabilities and developed systems to integrate these tools across the department. This includes establishing a performance analytics and research group that Boatright manages and features six full-time research scientists and data analysts who respond to officer inquiries and oversee research agreements with institutions around the country. SPD also partnered with Accenture to build a data analytics platform and records management system to situate its data in one place and create a user-friendly interface that allows department personnel to engage with new information. Brian Maxey, SPD’s COO, explained, “The point of the data analytics platform is to take all of our disparate systems and pull them all into one place and relate events, people, and locations such that we can draw meaning from it.” “We tell our commanders,” Maxey added, “‘Be data curious. Here’s a dashboard, play with it, and see what you can see.’”

In parallel to fostering this overarching growth process, SPD has focused its data-driven work on three core outcomes – equity, accountability, and quality – that are often difficult to measure but illuminate the benefits of becoming a learning organization that skillfully uses data. For instance, law enforcement organizations have often relied on analyses of disproportionality to gauge equity; this refers to the deviation between an activity in a demographic group and that group’s representation in the population (e.g., one way to gauge disparate policing is look at disproportional treatment of specific demographics, such as Black citizens). The problem is that this fails to inform what should change if disproportional policing is occurring. In an effort to tease out more helpful insights, SPD has employed propensity score measuring, a technique to create quasi-experimental conditions to control for an array of factors (e.g., officer age, location, time of day, or weather) and isolate one independent variable (e.g., race). Then, if SPD identifies a disparity, it shares that data with officers and asks them for help analyzing it. “The question we’re posing,” Boatright explained, “is not, ‘Okay, well, here’s what we’ve controlled for, therefore it’s bias.’ Instead, SPD now engages captains in dialogue to ask, ‘This is what we’re seeing in your precinct. What do you think is driving this fluctuation and disparity?’” This has led to fruitful discussions where officers share what they were focusing on in a situation, and Boatright asks the department’s data scientists to draw on data and integrate that information into the model. This embodies the collaborative process that is necessary to leverage analytics to its greatest effect; it is also helping SPD isolate variation that can only be explained by a subject’s race, which in turn allows the department to conclude, “This is what we really need to focus on to ensure equitable treatment across the board.”

SPD has similarly employed sophisticated analytical techniques and collaborative dialogue to evaluate and augment accountability. Specifically, the department used a kernel density estimate to explore residuals of police presence. This means that SPD aggregated data on vehicle location and overlaid it on a map with service calls, which enabled them to identify hot and cold spots for policing and seek feedback from officers to try to explain the variation. What’s more, the department paired the analysis of hot and cold spots with an examination of metrics for community sentiment, including surveys of residents affected by Micro-Community Policing Plans that SPD established with Seattle University. This helped the department identify goldilocks zones where communities are receiving just the right amount of law enforcement support. It also reinforces how an interactive dialogue, paired with data analysis, can lead to learning that facilitates changes in services and behavior and ultimately improves outcomes.

Finally, SPD is experimenting with cutting-edge technology to analyze the quality of policing. The department is partnering with Truleo, a technology company, to examine audio recordings from body-worn cameras and use machine learning, natural language processing, and network systems analysis to evaluate whether the language and tone are positive or negative. The objective, Boatright explained, “is to balance around an equilibrium. What is the middle level you would expect to see in the course of an intervention?” Boatright emphasized that, although SPD will have the capacity to break this data down by individual beats and discuss it with commanders, this is not a tool to fish for disciplinary cases. Rather, it is a way to get at the higher-level “organizational health of the department” and continue to foster data curiosity and improve services and outcomes. “For far too long,” Maxey observed, “departments have relied almost solely upon crime data and on response times as the metrics [for quality]. We’re trying to go far beyond that, and really understand what our officers are doing, why, and what the impact is on our community.”

Conclusion

The experiences of LAPD and SPD with data illuminate that for police departments, becoming a learning organization is about not only leveraging innovative technology and statistical techniques but also engaging personnel across the department and helping them develop new capabilities. In other words, to bolster the capacity of the organization as a whole, leaders must help individuals develop their analytical and critical thinking skills, as LAPD did through leveraging officers’ ingenuity and SPD did through the cultivation of data curiosity. This will inevitably involve challenges, ranging from difficulties with data collection to conflicting interpretations of the findings to resistance from some personnel who do not want to change. But if leaders are persistent with introducing and encouraging these shifts, they will eventually create a “growth mindset” among their officers who will recognize that they have the potential to learn and contribute to positive change. This leads to a collaborative analytical process through which police departments unearth interesting findings and identify changes that will improve service, produce better outcomes, and foster stronger, more trust-filled relationships with the communities they serve.