America’s new policing tech isn’t cutting crime

ON MAY 31ST 2020 the life of Michael Williams, a 66-year-old from the south side of Chicago, fell apart. That evening Mr Williams picked up a young hitchhiker in his neighbourhood. A few blocks later, the young man was shot, apparently through the passenger window of the car. Mr Williams rushed him to hospital; two days later the man died, and a few weeks after that Mr Williams was arrested. He spent the next 11 months in the Cook County jail, accused of murder. But before the case could come to trial, it was thrown out, when the public prosecutor in the Chicago area decided to withdraw its main evidence. Since last year Mr Williams has been suing the City of Chicago, alleging that the city’s police department deliberately relied on a case it knew was flimsy.

That evidence was from Shotspotter, an “acoustic gunfire-detection” system supplied by SoundThinking, a firm based in California. Shotspotter automatically recognises and analyses the sound of gunshots from a network of microphones spread across cities—Chicago has by far the largest network in America. By triangulating the recordings it can, in theory, pinpoint where a gun has been fired. The idea is that this will help police officers respond more quickly to shootings, and find out about shootings that go unreported.

Chicago spends over $10m a year for the service. But it is controversial. Cases like that of Mr Williams are partly why. Before his election in May, Brandon Johnson, Chicago’s left-wing mayor, promised to end the city’s contract (as mayor he has extended it, seemingly by accident).

Shotspotter is just one of the most prominent technologies promising to transform policing for the better. These include facial-recognition technology, licence-plate readers, AI-assisted data analytics, web-connected CCTV cameras and technology designed to locate people from their mobile-phone signals. SoundThinking’s accounts hint at the scale of spending on such things: its revenues are expected to reach about $93m this year, double what they were in 2020. Other companies, such as ClearView AI, which sells facial-recognition technology to police forces, or L3Harris, which sells “Stingrays” to gather information from cell phones, are also bringing in large sums. These firms promise police a virtual panopticon—with the ability to gather and use more data than ever to respond to and investigate crime. But will this actually help reduce crime?

In theory, police can trace crimes in ways unimaginable even a generation ago. And sometimes they do. In July police in New York arrested and charged an architect with killing three women in an investigation that took more than a decade and involved not only DNA but also the whittling down of vast amounts of mobile-phone data to isolate the unregistered “burner” phones that the accused is alleged to have used.

Yet overall, clearance rates are at all-time lows. In 2019, the latest year with national data, barely half of murders resulted in an arrest; in the 1960s the figure was closer to 90%, according to Jeff Asher, an analyst of crime data. In Chicago, barely a tenth of non-fatal shootings result in an arrest. The figures for other violent crimes, such as robbery or carjacking, are worse. Why is the tech not helping?

Part of the problem is that criminals adapt too. Just as burglars learned a century ago to wear gloves, today’s criminals often wear masks, making cameras less useful than they were. Similarly, the ability to follow people by means of their phones is less effective when criminals can use burner phones, or encrypted apps.

Even so, criminals working out how to evade surveillance hardly explains the failure to capture so many of them: many carjackers post their exploits live on social media. Rather, says Eric Piza, of John Jay College of Criminal Justice in New York, part of the problem is that police often lack a clear sense of how technology will reduce crime, and end up overwhelmed by huge amounts of information that consumes resources. “Tech is only going to improve public safety if it puts into motion effective police practices that by themselves improve public safety,” says Mr Piza.

Shotspotter is a good example. In November Mr Piza published a paper which looked at its use in Kansas City, where the microphones cover an area of 3.5 square miles (7.8 square km). When he compared this area with a similar part of the city without microphones, Mr Piza found that the police did indeed find more evidence of gunfire—such as spent shell cases. “Our research found [the technology] did deliver on those promises,” he says. But the study also showed that no fewer people were shot in the area covered by Shotspotter, nor were more gun crimes solved. It is not obvious, says Mr Piza, that police know how the tech is meant to reduce crime.

And technology comes at a cost—not only that of SoundThinking’s fee. A large proportion of Shotspotter alerts lead to nothing, but they sap police time, as patrol officers have to chase down warnings. Over 90% of Shotspotter alerts do not result in any evidence of a crime being committed, according to a 2021 study by Chicago’s inspector-general. Another study published in November, by Michael Topper and Toshio Ferrazares, PhD students at the University of California, Santa Barbara, looking at Chicago, found that after the roll-out of Shotspotter the police responded to 9/11 calls two minutes slower than before, and in the case of domestic-violence calls made fewer arrests.

Ralph Clark, the CEO of SoundThinking, says people who think the tool is directly able to prevent violent crime are “singularly misinformed”. He argues that having extra information about gunshots, especially those that are not reported by a 9/11 call, means that police are able to build more community trust, and that the technology can help police find victims of gunshots more quickly. But even he admits that gadgetry on its own cannot drive down crime. “We are seeing a fairly consistent diminution of on-the-ground policing resources,” he points out. His firm’s technology (SoundThinking also provides several other police tools) is best when it is used to free up police resources, he says, rather than to add to the demands on officers.

Some fear, however, that police departments adopt shiny new gizmos precisely because this allows them to stick to old tactics. “We have 50 years or so of evidence about what works,” says Mr Piza. “Proactive strategies work better than reactive ones.” Officers who work a beat can prevent crimes before they happen. But for decades American policing has tended to prioritise patrol over detection, and drag nets over precision. Attend bond court, where newly arrested suspects are booked, and it is clear that many more people are arrested as a result of speculative traffic stops than as a result of diligent detective work. Such styles of policing do not build trust.

Daniel Webster, a specialist in gun crime at Johns Hopkins University in Baltimore, says that there, juries increasingly acquit in cases where there is no video footage, because they simply do not trust the police any more. “The police are the public, and the public are the police,” said Sir Robert Peel, the founder of modern policing. All the cameras and audio recorders in the world are not a substitute for that. ■