Simon Montag

Do Algorithms Have a Place in Policing?

How a Pakistani-born retired pilot took on a controversial, data-driven policing program in Los Angeles—and won

Inside the bright-white walls of his office, in one of Los Angeles’s poorest neighborhoods, Hamid Khan is a calming presence. Khan, 62 years old, is the founding organizer of the Stop LAPD Spying Coalition, a police watchdog group headquartered in Skid Row. A tall, broad-chested man, with a pronounced forehead and grayish-black hair kept neatly trimmed, he is patient with the constant parade of people from the neighborhood who walk in asking for toiletries or seeking advice about dealing with prying officers. One afternoon, as we finished a meal at a Mexican restaurant, he got the leftover guacamole to go, and asked for more chips. “Sean will love this,” he said, then carried the paper sack to a homeless man whose tent was propped against a fence outside the coalition’s office.

At least once a week, though, Khan gives an almost theatrical performance of his role as one of the Los Angeles Police Department’s most vocal critics. At public meetings, he hurls startling obscenities: He has referred to the LAPD as “this motherfucking department” and called one official “Mr. Frankenstein himself.” During public comments, he’ll turn his back to police commissioners, occasionally urging supporters in the audience to drown them out with raucous chants. Khan’s methods are more radical than those of many other local activist groups, and yet, Stop LAPD Spying has arguably been more successful than any of them in challenging the powerful LAPD to end controversial practices.

Khan likes to say that he’s just the coalition’s coordinator, but in reality he’s the founder and de facto CEO. His organization has brought in a relatively small amount in donations and other outside funding, relying mostly on indefatigable volunteers. For years, Khan ran the coalition without pay while living on his pension as a retired UPS pilot. Now it pays him a small monthly stipend. Khan’s 32-year-old daughter, Nadia, volunteers as the head of the coalition’s youth-justice efforts.

Under Khan’s leadership, the coalition’s members have taught themselves how to file open-records requests to obtain information; organized dramatic, well-attended protests; and gained national respect. They have done all this while maintaining a core belief that has at times alienated even Khan’s closest allies: Law enforcement should be abolished altogether.

Shahid Buttar, the director of grassroots advocacy for the Electronic Frontier Foundation, a nonprofit focused on digital civil liberties, got to know Khan when they worked together to battle the expansion of police surveillance technologies. Because EFF aims to reform police use of technology, not abolish it, Khan won’t collaborate with the organization much. Still, Buttar told me he has an immense amount of respect for Khan, a man he said is “like the uncle I never had” and “a stalwart example of the resistance.” He said, “He gives no F’s and is always willing to speak very directly.”

Hamid Khan stands near the headquarters of Stop LAPD Spying in the Skid Row neighborhood of L.A. (Damon Casarez)

Steve Soboroff, the president of the Los Angeles Board of Police Commissioners, which oversees the police department, is often on the receiving end of Khan’s sharp language. He told me, “I think it’s to our credit that we give them credit, because it’s hard to listen to somebody who talks to us the way they talk to us.”

Khan already had a reputation for being well versed in police surveillance technology in October 2015, when he was invited to the Data & Civil Rights conference in Washington, D.C., to participate in a roundtable on “predictive policing”—the use of analytics and statistics to forecast crime—which was growing in popularity. While researching the LAPD’s use of various technologies, Khan had become particularly alarmed by a program called PredPol that had been pioneered inside the LAPD in the late 2000s, and had since spread to departments in Seattle, Atlanta, and elsewhere. Because the data that fed PredPol’s algorithm were based on reported crime, Khan and other critics argued that the program would lead to aggressive policing in communities of color, where reported crime tends to be high.

The LAPD and PredPol argue that, because the algorithm does not factor in demographic characteristics, its effect is proportional and fair. “It is math, not magic, and it is not racist,” an LAPD spokesman Josh Rubenstein wrote in an email. PredPol’s CEO, Brian MacDonald, cited his company’s “Stance on Privacy, Civil Rights & Transparency,” which makes the case that its algorithm helps to reduce the influence of officers’ biases in policing.

By the time of the conference, Khan had read everything he could by P. Jeffrey Brantingham, an anthropologist at UCLA who was emerging as PredPol’s frontman. At the roundtable, he was excited to see the academic seated across from him. “I told him how there was inherent racism in predictive policing, and he came back and said he wasn’t a racist,” Khan recalled. Later that evening, at a mixer, Brantingham asked whether he could get Khan a beer. Khan already had one, but the two men got to chatting. “He tried to offer his justifications, but I said, ‘Look man, however you approach it, the impact remains the same,’” Khan said. (Brantingham didn’t respond to requests for comment.)

About seven months later, Khan and Jamie Garcia, an oncology nurse who was active in the coalition, set up a call with Andrew Ferguson, a professor at the University of the District of Columbia’s law school who’d been researching the growing use of predictive policing around the country. Over the phone, Ferguson told them what he knew about PredPol. He also mentioned a second data-driven LAPD program called LASER that Khan and Garcia, for all their research into predictive policing, knew little about. LASER, Ferguson had learned, identified repeat offenders and distributed bulletins with their photos and physical descriptions so officers could attempt to stop them from committing crimes. Through his research, Ferguson said, he’d come to have privacy and civil-rights concerns about how these people were being identified and how the bulletins were being used.

After the phone call, Ferguson emailed Khan and Garcia an article that he planned to publish in the Washington University Law Review, called “Policing Predictive Policing.” In it, Ferguson wrote that technology in policing would only continue to advance: “Police have entered the age of actuarial justice.” Given that, he wrote, police departments should address issues with inaccuracies and biases, and be transparent with the public.

Khan and Garcia appreciated the information Ferguson had gathered, but rejected his conclusion that the use of predictive policing was inevitable and must simply be better managed. “His research was informative in understanding the structural elements of these things,” Khan said, “but he was looking for ways to reform, and we were on different sides.” A few months later, the coalition launched what would become its most ambitious campaign to date—trying to halt the LAPD’s use of predictive policing altogether.

The LAPD’s experimentation with predictive policing had begun quietly in 2008, under the police chief at the time, William Bratton. A former New York police commissioner, Bratton had built his career on finding patterns in criminal behavior. “I believed passionately that crime could be prevented,” he recently told me.

In New York, he developed CompStat, an approach to crime reduction that uses computerized crime maps, and became a champion of broken-windows policing, a strategy of pursuing crimes that affect quality of life, such as vandalism and public drinking, on the theory that they are precursors to bigger crimes. The results were mixed: Most research has found that New York’s approach contributed to lower crime, but some studies have suggested that it damaged relations between the police and the community.

In 2002, Bratton was named chief of the LAPD. The department was “a mess,” Bratton recalled, operating under the oversight of the U.S. Department of Justice because of a federal consent decree resulting from allegations of racism scandals and civil-rights violations, including the 1991 beating of Rodney King. The LAPD had its own version of CompStat, which wasn’t working, Bratton said, and needed to be overhauled. Sean Malinowski, then a sergeant with a doctorate in public administration, was named an aide to the chief in 2004. Over the years, the men spent hours talking about policing, technology, research, and crime, and together fixed the issues with CompStat, a program Malinowski would eventually oversee. But they had higher aspirations.

In 2007, Bratton invited researchers to experiment on the LAPD. “How can we use predictive methods to create even more timely and successful intervention and crime reduction initiatives?” he and Malinowski asked in a paper they co-wrote.

As Bratton had hoped, researchers began to contact his department. Brantingham, the son of respected criminologists, wanted access to the LAPD’s data, and Bratton asked Malinowski to play middleman. Brantingham’s research team had adapted a model that predicts earthquake aftershocks to try to forecast property theft, based on the notion that after one crime is committed in an area, more are likely to follow. Brantingham said that if he was given the department’s crime data, he would apply an algorithm that could spit out 500-square-foot boxes showing where property crimes would most likely occur. Eventually, the program became known as PredPol.

Around the same time, the LAPD approached Craig Uchida, the president of the consulting firm Justice & Security Strategies and a former Justice Department senior executive. He proposed a program that would use the past six years’ worth of LAPD data to find geographical areas of gun violence, so that the department could focus patrols where they would most likely be needed. Uchida and Malinowski applied for a Justice Department grant to fund it, won $499,958, and named their program Los Angeles Strategic Extraction and Restoration—LASER, for short.

Top: Stop LAPD Spying created this map, using data it received in open-records requests, showing areas in downtown L.A. where PredPol predicted crime was likeliest to occur. Bottom: Hamid Khan addresses the audience, as he often does, at a weekly Los Angeles Board of Police Commissioners meeting. (Damon Casarez)

In October 2009, Bratton retired, but Malinowski and his data-driven experiments—anchored by PredPol and LASER—remained. The next month, Bratton’s successor, Charlie Beck, co-wrote an article for Police Chief magazine echoing Bratton’s faith in the tactics: “Predictive policing already has been shown to enable doing more with less, while significantly improving policing outcomes.” Three months earlier, a judge had provisionally lifted the federal consent decree on the department, ending Justice Department oversight. It was a new era at the LAPD, complete with a 10-story, steel-and-glass headquarters next to city hall.

For three years, researchers and the LAPD designed LASER and PredPol experiments, and other cities including Chicago, Memphis, Minneapolis, and Dallas followed suit. The predictive-policing trend was becoming so much a part of modern-day life that in 2011, Time magazine called “pre-emptive policing”—another name for it—one of the 50 best inventions of the year.

The LAPD field-tested LASER in September 2011, in the nine-square-mile Newton Division south of downtown, home at the time to 44 documented gangs and a high concentration of gun-related crime. The team identified five corridors, called “LASER zones,” where gun violence would likely occur and where, accordingly, the LAPD would increase patrols. It also identified people who were likely to be involved in crime based on a point system. Known gang members, for example, got five points. The worst offenders—the ones with the most points—would be featured in something called a “chronic offender bulletin,” which looked like a wanted poster, circulated among officers.

PredPol field-testing began that November in the Foothill Division of the LAPD’s Valley Bureau, where property crime was high and Malinowski was captain. During roll call, patrol officers got printouts of mission maps with tiny red boxes generated by either PredPol or a human analyst; Brantingham’s research team wanted to compare their effectiveness. Between calls, officers were to visit the boxes. “Just make your presence known,” Malinowski instructed them.

In both programs, officers were told to measure their “dosage”—the amount of time they spent in a targeted area. The programs faced skepticism among officers, who didn’t like being micromanaged. Malinowski would ask officers to take a leap of faith, playing on the natural sense of competition among the department’s bureaus: “Don’t you want to be part of the winning team?”

In October 2012, Uchida and others wrote in a paper for the Justice Department that LASER was “analogous to laser surgery, where a trained medical doctor uses modern technology to remove tumors or improve eyesight.” During the first nine months of LASER, they wrote, violent crime in the Newton Division had declined enough to be statistically significant. (Uchida declined to comment on criticism of LASER.)

PredPol was tested through January 2013, in Foothill and then in two other areas. While it was ongoing, the researchers raised $1.3 million and founded PredPol Inc. Two years after the experiment, PredPol researchers concluded in a peer-reviewed study that the algorithm was about twice as successful at predicting crime as human analysts.

After the pilots, the department expanded its use of LASER and PredPol, funding the former in part through an additional $400,000 in Justice Department funds. The LAPD was gaining attention for its use of predictive policing, and Hamid Khan was taking notice.

Khan, born and raised in Lahore, Pakistan, has always been skeptical of authority. His father had an elementary-school education and ran away from home as a teenager, first joining the military and then becoming a police officer. When Khan was young, his father worked out of town during the week; on weekends, he’d tell his children about how policing was a dehumanizing experience.

In the late 1950s, General Ayub Khan engineered a coup and appointed himself president of Pakistan. Economic disparities widened. Khan was 11 when students, workers, and artists held strikes and engaged in sometimes-violent protests opposing military rule. One day, he watched in horror as a police officer shot a student protester. “I saw the body rising up, and then falling, with blood coming out of his mouth,” Khan said. After that, he’d occasionally join in, throwing bricks at police.

Still, he obeyed his parents and made good grades. He earned a bachelor’s degree in economics at Government College, Lahore, before leaving Pakistan for the U.S. in 1979, at the age of 22, on a tourist visa. He headed straight for his uncle’s home in Orange County, California, and worked at a hamburger stand, then as an airport-shuttle driver. Later, with financial help from his parents, he attended flight school—he’d always dreamed of flying—and got a job as a flight instructor and then a commercial pilot.

Orange County, at the time, had gone from a sleepy agricultural region to one bursting with residential growth. Immigration-law changes in the 1960s and a post–Vietnam War refugee crisis had added more than 500,000 residents in the 1970s, many of them immigrants. Yet as a dark-skinned man from a Muslim family, Khan constantly felt like an “other”: “It was always with me; I couldn’t shed it.”

Khan fell in love with a woman and married her. In 1985 they had a daughter, Sasha. She was born with an incurable virus called congenital cytomegalovirus, which left her epileptic and lacking motor control. Khan told himself, I’m not going to look at her with pity.

While starting a family, Khan, with his forceful, charismatic personality, was also becoming an informal leader among local South Asian immigrants. In 1985, a few Pakistani friends confided in him that, along with some black and Latino friends, they’d been turned away from the popular Santa Ana location of a disco chain called Red Onion. Khan helped plan protests and bring the chain’s actions to the press. Over time, more people voiced their concerns. Together, they filed civil lawsuits and complaints, which resulted in two settlements with Khan and others. In 1986, the chain paid $500 each to 39 individuals and, two years later, $15,000 each to 26 individuals.

In 1988, Khan got a lucrative job flying Boeing 727s for UPS. But as he crisscrossed the globe delivering packages, his mind was occupied with thoughts of home—Sasha’s condition and the troubles of his immigrant friends. By then, his wife had given birth to Nadia, and Khan had become a citizen. In 1989, Khan flew home from a training in Louisville, Kentucky, and Sasha died suddenly of pneumonia that night, at the age of 4. (He and his wife would divorce years later.) The following year, Khan founded a grassroots organization called the South Asian Network to unify South Asians and help them to better advocate for themselves. Some Muslim friends balked at this new venture, thinking Khan was associating too closely with Hindus. But Sasha had taught him to focus on people’s commonalities, not their differences.

For years, he balanced two lives—flying for UPS to make a living and running the South Asian Network in his time off, often with Nadia in tow. By the mid-Aughts, Khan’s double life was wearing on him. In 2006, he quit flying and joined the South Asian Network full-time.

In late 2008, Khan got an intriguing phone call from a Muslim friend asking whether he knew about a program the LAPD was developing called iWatchLA. The two got on the phone with the American Civil Liberties Union and learned that the program encouraged citizens to report behavior that appeared to be linked to terrorism. Although the LAPD says the program is strictly “about behaviors and activities, not individuals,” Khan worried that people who looked like him would be reported for no good reason, and subjected to uncalled-for investigations.

His fears were compounded by a program the department had established earlier that year called suspicious-activity reporting, which gave officers a formal channel to flag suspected terrorism-related behavior. Rubenstein, the LAPD spokesman, told me behaviors that trigger a suspicious-activity report (SAR) must fall into one of 16 categories, such as “express or implied threat,” and officers may not detain someone based on a report without reasonable suspicion of criminal activity or probable cause. In Khan’s eyes, though, these two programs were codifying speculative policing.

Hamid Khan (right) and Jamie Garcia (left) walk with another Stop LAPD Spying member in the Skid Row neighborhood of L.A., where the coalition has its headquarters. (Damon Casarez)

Unlike with Red Onion, this time it wasn’t just his own community that would be affected. Khan had been spending more time with a friend and fellow organizer, Pete White, who is African American and runs an organization called the Los Angeles Community Action Network (LA CAN), in Skid Row. White had helped educate Khan about historic relations between the LAPD and the city’s black community, and how homelessness had plagued African Americans.

As many as 5,000 homeless people lived in the area at the time. On the sidewalks outside LA CAN’s office, people slept in tents, flattened cardboard boxes, and lean-tos. Khan was appalled that the homeless in Skid Row lacked services such as trash cans and public bathrooms, yet the area was thick with officers. “I became very conscious and aware,” Khan said. While he’d seen extreme poverty as a child in Pakistan, he expected better in the richest country in the world.

Khan left the South Asian Network in 2010 to focus more on the LAPD. In March 2011, he won a Soros Justice Fellowship with a proposal to form a coalition to “challenge Los Angeles Police Department surveillance and profiling practices that criminalize benign and legal activity, normalize racial profiling, and render people in certain communities as criminal suspects.” (Rubenstein argued that predictive policing does not criminalize benign and legal activity, but instead seeks to prevent crime and reduce victimization through the placement of officers at specific locations.) He asked White for some space in LA CAN’s Skid Row office, and established Stop LAPD Spying.

As Khan learned more about LAPD technologies, he gave presentations to classes, community groups, neighborhood associations, and religious organizations. In October 2011, he joined hundreds of people at an Occupy Los Angeles protest outside city hall. There he met Jamie Garcia, the oncology nurse, who was volunteering at the encampment as a medic along with an Iraq War veteran. Garcia would point out weird-looking vehicles, and the veteran explained that they were police. “I thought, Why would there be so many police?” Garcia told me.

She attended a few of Khan’s presentations and learned how much money the city spent on law enforcement. At work, Garcia could only do so much for her patients. The public health providers’ budgets were limited, but Garcia now saw a police department that had enough cash to buy surveillance equipment and spy on protesters. Garcia started volunteering for Stop LAPD Spying, where she soon became a leader.

In April 2012, Stop LAPD Spying filed an open-records request to learn more about the police department’s iWatchLA and SAR programs. The coalition also established its mission to demystify police surveillance for the public, and build a long-term, grassroots movement to stop it, with the lofty goal of ending policing altogether.

Later that year, Michel Moore, then an assistant police chief and the director of the Office of Special Operations (he would later become the chief of police), asked to meet with Khan to hear more about his concerns regarding iWatchLA and SAR. Fine, Khan replied, but only on our terms: Moore couldn’t wear his uniform or bring a gun, and the meeting had to be open to anyone. About 70 people attended the June 2012 meeting, Khan said, and Moore arrived in a suit with no gun visible. “He asked what LAPD could do, and we said, ‘Dismantle the programs,’” Khan told me. “He said that wasn’t going to happen, and that was that.” Both iWatchLA and the SAR program remained in place.

In some corners of the policing world, the LAPD has been seen as an innovator ever since the Bratton years. In 2015, Malinowski was inducted into the Hall of Fame of the Center for Evidence-Based Crime Policy, at George Mason University. While the LAPD had spent the previous couple of years expanding its use of PredPol and LASER, the technologies were optional and used inconsistently throughout the department. “I once called a captain about starting to use it, and they said, ‘How long do we have to use this shit?’” Malinowski told me. “So I said, ‘Never mind, you’re not ready.’”

Still, inside the department, several officers told me they were rankled by the rollout, which felt mandated to them. They considered policing to be a craft, not a science, and had spent years perfecting it. Plus, they rarely had time between calls to visit the locations they were asked to patrol—let alone use their computers to mark their presence. “I don’t think anyone was super excited about it,” Hannu “T. J.” Tarjamo, a 22-year veteran officer, told me. “It came down as a directive, and takes the talent and creativity away from police officers.” Rubenstein disputed that critique, saying that officers understood why the programs were needed “and have embraced them.”

In 2016, when anti-predictive-policing sentiment was on the rise among the general public, a collection of 17 groups—including the ACLU, the NAACP, and the Electronic Frontier Foundation—signed a high-profile statement expressing their concerns that the approach had been implemented without transparency and reinforced racial biases. In February 2018, Brantingham co-wrote a paper in response to these concerns. Using data from early PredPol trials, he and his co-authors compared arrests by officers who relied on PredPol predictions with arrests by officers who relied on a data analyst’s predictions, and did not find significant differences in the proportion of arrests for people of color. He also wrote, though, that he could not be sure that the PredPol algorithm reduced crime any more than human analysts—a conclusion that cast doubt on earlier findings.

Much of the criticism of the programs centered on the fact that little was known about them. In May 2017, Khan’s coalition had filed a public-records request for LASER policy documents, correspondence, and chronic-offender bulletins. In February 2018, the coalition sued the LAPD, alleging its requests had not been answered. (The LAPD declined to comment on the lawsuit, which is still pending.) Three months later, in May 2018, it filed a request for PredPol records.

After a public meeting that spring, Garcia asked Mark Smith, the inspector general of the police department, to conduct an audit of the LAPD’s use of predictive policing. Coalition volunteers had spent dozens of hours on records requests to piece things together, but an audit—a formal process that the inspector general undertakes occasionally to evaluate police programs—would publicly expose the use of predictive policing. “Why are you making us do the work?” Garcia recalled asking Smith. “It shouldn’t just fall onto the community.” Khan made a similar request in writing, and Smith responded that he would consider auditing the programs in the future.

Garcia and others repeated their plea to the police commission. Commissioner Cynthia McClain-Hill, an African American who was known for urging transparency about police initiatives, particularly the use of drones, said Garcia’s aggressive outreach, along with the commission’s conversations with the police department and the inspector general, made it clear that the programs, “regardless of how well intended, had not been adequately reviewed.” She called for a hearing focused on predictive policing.

Scheduling conflicts forced the hearing date past June 2018, when Moore—the only LAPD executive who’d attended a coalition meeting—was sworn in as chief, and as McClain-Hill prepared to resign from the commission. Khan and Garcia had always felt the board didn’t question the LAPD enough. (The Los Angeles Times recently reported that Police Commissioner Sandra Figueroa-Villa failed to disclose that PredPol made a $7,500 donation, in 2014, to a nonprofit she leads. Figueroa-Villa told the Times that she didn’t realize this might represent a conflict of interest, and will “take responsibility” for the donation.) When we spoke, McClain-Hill challenged Khan and Garcia’s characterization, pointing to commission-championed changes to policies regarding de-escalation, use of force, and the release of body-camera footage. Soboroff, the president of the board, went further, calling the criticism “categorically incorrect.” At McClain-Hill’s last meeting, the board, including Figueroa-Villa, unanimously approved a motion directing Smith to probe the department’s use of PredPol and LASER, conduct a cost-benefit analysis of the programs, and evaluate their effectiveness.

The commissioner who made the motion was Shane Murphy Goldsmith, the president of a social-justice organization called the Liberty Hill Foundation, which had once recognized Khan as a leader to watch. Goldsmith told me she had been concerned because “people from the community had been saying that these strategies made them feel less safe.”

The coalition’s mission might have also been aided by a new police chief who was more sympathetic to the group’s point of view than it had realized. Moore later told me that he had his own reservations about some aspects of the LAPD’s predictive policing and had encouraged an audit.

In retrospect, Moore characterized the prolonged lack of oversight as unintentional. The programs were not meant to be secretive but grew incrementally at a time when public mistrust of law enforcement, especially when conducted out of view, had ballooned. Malinowski conceded that this was problematic. “At some point, had I been more aware, I would have said, ‘Let’s do a town hall or something,’” he told me. “But the environment at the time wasn’t the same.”

That spring and summer, Khan had been receiving emails from the city notifying him that it was releasing records on its online portal in response to the coalition’s requests for LASER and PredPol information. A series of requested crime maps showed that Skid Row was surrounded by PredPol boxes. But the records included very few details regarding internal policies or contracts with vendors. In general, the information was startlingly limited, Khan thought. His initial theory that the LAPD’s predictive policing constituted a sophisticated surveillance operation began to give way to a different concern: It was so poorly coordinated that no one could hope to unravel its inner workings. “All these things coming together were major red flags,” he said.

In March, when Inspector General Mark Smith released the requested audit, Khan’s suspicions were largely confirmed. The data sets were inconsistent, according to the auditors, which made it impossible to use them to draw useful conclusions about the programs’ effectiveness. Even after nearly a decade of use, it was not clear whether PredPol, in place citywide, and LASER, implemented in all but five divisions, had respectively caused property crime and violent crime to decline.

The audit also revealed that the LAPD had quietly stopped publishing chronic-offender bulletins—a centerpiece of LASER—the previous August, after Moore developed misgivings about them. The program had been filled with inconsistencies: Many so-called chronic offenders appeared to have no history of violent crime at all, and the intricate point system wasn’t always used. Of the 637 individuals who were on the list at the time the bulletins were suspended, 112 had zero points. Some chronic offenders had received letters telling them they were on the watch list; others didn’t.

In addition, officers’ time in target zones had been inconsistently monitored. In 2018, 524 hours were recorded in the Pacific Division LASER zones, while 53,841 hours were recorded in the LASER zones of the Hollenbeck Division. Both LASER and PredPol seemed to have incorrectly counted time spent parked at LAPD facilities as “dosage,” often rendering officers’ activity logs unreliable and making it “difficult to draw conclusions about the effectiveness of the system in reducing vehicle or other crime,” the audit stated.

Malinowski said the auditors couldn’t measure the programs’ effects, because “they’re not researchers,” but added that he takes “credit and blame for what happened here.” Bratton, who hadn’t read the audit when we spoke in May, said crime decreases in Los Angeles and New York, without corresponding increases in incarceration, indicate that his life’s work of looking for crime patterns had been fruitful. “The common linkage is me,” he said. Both broken-windows policing and predictive policing were based on the same concept—it was just the technology that had evolved. “They’re entitled to their opinion,” Bratton said about the auditors. “It’s an opinion that I don’t share.”

Hamid Khan leads members of the Stop LAPD Spying Coalition in a discussion, while Jamie Garcia takes notes. (Damon Casarez)

It was gratifying for Khan and Garcia to have their suspicions confirmed, but discomfiting to know that the programs were a mess. Their haphazard implementation made it impossible to know, for certain, what their impact had been. The audit’s findings of significant inconsistencies aligned with what Sarah Brayne, a sociologist at the University of Texas at Austin who shadowed the LAPD from 2013 to 2015, saw while riding along with officers using PredPol and LASER and observing analysts using the programs in substations. She expressed particular concern about LASER. “The fact that we don’t know if this thing works is indefensible,” Brayne told me. “This is $900,000 worth of federal grant money for literally, specifically, evidence-based policing, and our evidence is so bad at this point in the game.”

Moore promised to respond publicly to the audit, and the commission set up an email address to collect public comments on data-driven policing for two weeks. Khan and the coalition encouraged citizens to submit comments, flooding the commissioners with negative correspondence. Through email, Khan sought comments from professors and students he’d come to know in UCLA’s anthropology department, on the top floor of a red-orange brick, Romanesque-style building off the main quad, where Brantingham still worked as a professor. One anthropology professor who’d met Khan through his work on drones emailed Brantingham to see whether they could talk about concerns she and others had with his research and PredPol work. He said he was too busy to meet. She and 67 other UCLA academics submitted a letter to the commission questioning Brantingham’s research.

That professor requested that she not be named, because she did not want to be identifiable to her colleagues in the anthropology department. Joshua Mayer, a 28-year-old doctoral-degree student in the department who also signed the letter, told me Brantingham’s work is troublesome from an anthropological perspective because of its outsize impact on immigrants and people of color. “Anthropologists have a terrible track record of engaging in racist and colonial efforts,” he said. Of the 819 who emailed, none expressed support. Goldsmith, well versed in community organizing in Los Angeles, told me that more than 800 people weighing in on an issue was “very significant and very, very rare.”

On April 9, Moore publicly addressed the commission about the audit. Seated on a dais, he sounded humble and almost apologetic: “This is an evolution of policing that has existed during my entire time with LAPD. And LAPD is seen not just here but across the country as an agency willing to learn.” Changes to the LAPD’s predictive-policing techniques, he said, were already under way.

The department was preparing a usage manual for officers, and from now on, its Office of Special Operations would oversee data-driven technologies. The LAPD would tweak its use of PredPol, using a new GPS system to better record dosage, and would rely more heavily on old-fashioned methods of tracking people who might be at risk of committing crimes—using physical descriptions of suspects and those released from parole or on probation. In general, the department would still try to reduce crime through geographical targeting, but it was transitioning to something called precision policing, a newly popular concept that incorporates both data-driven and neighborhood-policing techniques. Moore told commissioners the program was still “developing its attributes, if you will, as we speak.”

Afterward, Khan and Garcia stood in a circle with other activists to discuss what happened. No one was quite sure which portions of predictive policing were still intact, and which had been canceled. At a press briefing, Moore answered a couple of questions about predictive policing, in which he clarified that the entire LASER program had been canceled. (Two months later, the inspector general released an audit of the SAR program, which found that about half of the reports made in 2016 and 2017 were ultimately deemed unfounded. In addition, the audit raised concerns about how suspects were racially identified. Shortly after its release, Moore promised to address these concerns, and said the department needed to be more transparent about the program’s results.)

In an interview, Moore spoke passionately about the value of data in modern policing. Research has validated the relationship between crime and geography, and the fact that people convicted of crimes are more likely to reoffend, he said, but he also acknowledged that data must be used in a way that doesn’t erode the public’s trust in policing. Moore gave Khan’s coalition some credit for spurring a deeper review of how the LAPD was employing predictive policing, but added that he and other officials had concerns about the programs, too. While he conceded that some of the coalition’s criticisms about predictive policing were fair, he told me he disagrees with allegations that it is racist and biased. Also, he said, the coalition’s belief in abolishing police is out of touch and disrespectful.

Moore downplayed the importance of the predictive-policing programs: “These are small and only a part of a much more complex system of how we go about policing in L.A. They do not define policing in Los Angeles.” The comment was remarkable, coming from the chief of a department that once proudly positioned itself as an early, and aggressive, adopter of predictive policing.

As predictive policing has proliferated in recent years, it may be falling out of fashion, Malinowski told me. Milpitas, a city of about 75,000 near PredPol Inc.’s headquarters in Santa Cruz, signed a contract with the company in 2013, but canceled it after one year because the city wasn’t benefiting enough, the police chief at the time said. In 2015, after groups including the ACLU voiced concern, the Arizona governor vetoed a bill that would have allocated more than $1 million to predictive-policing pilot programs. In 2018, then–New Orleans Mayor Mitch Landrieu decided against renewing a six-year agreement with the data-analysis company Palantir for a pro-bono predictive-policing program.

Researchers continue to critique predictive policing, most recently at New York University’s AI Now Institute, which studies the social impacts of artificial intelligence. A paper published in May concluded that predictive policing is potentially discriminatory if it relies on “dirty data” from policing practices that disproportionately affect minority populations.

Ferguson, the professor who informed Khan about LASER three years ago, told me he hopes the LAPD audit draws attention to the pitfalls of implementing these programs incorrectly. Ferguson feels that some of the coalition’s concerns have been unfounded, and some of its tactics extreme. But, he said, “the wins in the fight over data policing have been in that vein—where outrage and a lack of nuance have driven the conversation. And it’s been effective.”

Conceding that the programs went awry is difficult for Malinowski, but he gives the activists credit for bringing issues with predictive policing to the inspector general’s attention. “I try not to be defensive about it,” he said. “When it got expanded and expanded, the execution was not consistent, and that’s where it fell apart.” In July, he formally retired from the department after 25 years to run a company he founded, Strategic Focus, which had recently won a $635,000 contract extension to run data-driven policing programs in Baltimore.

After Moore’s announcement in April about reining in some aspects of predictive policing, Khan returned to Stop LAPD Spying’s office in Skid Row, where he admired newly hung pictures and news stories on the walls commemorating the coalition’s milestones over the years. He sent out a celebratory press release about the diminished role of predictive policing in the department. Then, he Googled precision policing.