RAHM EMANUEL is an expensive date for Ken Griffin. Encouraged by Chicago’s forceful mayor, after he complained about the overcrowded lakefront trail, the billionaire hedge-fund manager donated $12m for a separate bicycle path in 2016. He gave $3m for soccer fields in poor neighbourhoods in December. Mr Emanuel, a Democrat, even persuaded Mr Griffin, a Republican, to pony up $1m for his re-election campaign. And at a recent tête-à-tête, he persuaded Mr Griffin to part with $10m to bankroll the joint effort by the Chicago Police Department (CPD) and the University of Chicago’s Crime Lab, a research centre, to use data-analytics programs to predict and prevent violence in the crime-plagued city.

Mr Griffin’s latest gift to his hometown will mostly go to the CPD’s Strategic Decision Support Centres (SDSC), where civilian analysts and cops crunch data from gunshot detection-systems, surveillance cameras and computer programs with the aim of identifying the places where violence is likely to break out. Starting with six last year, the city has set up such centres in 13 of its 22 police districts. Some of Mr Griffin’s money will also finance mental-health care for officers; some will go towards evaluating complaints against them.

Policing software such as Predpol or HunchLab, their makers claim, is able to forecast where crime is likely to be committed. Certainly the numbers are intriguing. After 2016 turned out to be the deadliest year for two decades, with 762 murders and 3,550 shootings, the following year, which coincided with the establishment of the first SDSC, was less bloody, with 650 murders and 2,785 shootings. The decline in crime in police districts with the new data centres was steeper than in those without. This could just have been reversion to the mean. But the Chicago police department thinks that HunchLab, the particular program it bought, has something to do with it.

To see why this might be the case, consider Englewood. A hard-up, predominantly black neighbourhood on the South Side, Englewood saw a decline in murders of 44% in 2017 compared with 2016. Shootings fell by 43%. A byword for concentrated poverty, rampant crime, drugs, guns and gangs, Englewood seems to have taken everyone by surprise with its progress.

Laura West, an officer working at the district’s SDSC, which is staffed by two officers at all times, spends her days surrounded by screens. One shows a program called ShotSpotter, which uses the sound of gunfire to pinpoint shootings; another shows where surveillance cameras are (the city has more than 40,000); and a third displays HunchLab software. This blends data on crime statistics, population density and weather patterns with fixed points such as liquor stores and highway exit-ramps, to identify patterns of crime that may repeat themselves. (Predictive policing software also takes into account the phases of the moon and the schedules of sports games.) At-risk sites are marked with boxes colour-coded according to the type of crime. Patrol officers are encouraged to check them frequently.

The key to Englewood’s improvement has not been more aggressive policing, says Kenneth Johnson, the district commander. “We cannot arrest our way out of our problems,” he says. Instead, as he tells it, the change is the result of targeted interventions, combined with improved relations with the local community. The CPD’s relationship with black Chicagoans in particular has long been fraught. Its recent nadir was a white officer’s seemingly wanton firing of 16 bullets into Laquan McDonald, a black teenager, as he was walking away. The officer, Jason Van Dyke, who is about to be tried for first-degree murder, had been the subject of numerous complaints. Changing such a culture will take time. In Englewood, Mr Johnson tells his 350 officers to attend community meetings, to build relationships and to avoid behaving like an occupying force.

The risk with policing software is that it amplifies existing racial bias. “Technology is far from neutral,” says Kade Crockford of the American Civil Liberties Union. When police officers feed predictive policing algorithms with their data on past stops and arrests, they can reinforce the bias that police across the country stand accused of, says Ms Crockford. For example, whites and blacks consume and sell drugs at pretty much the same rates, but far more blacks are arrested for drugs than whites.

Used carefully, though, more data are better than fewer, says Andrew Papachristos of Northwestern University, and anyway HunchLab does not use arrest records. It is too early to say whether the new tools caused the decline in crime in Englewood and other districts, though the evidence suggests a correlation. This is good news for Mr Emanuel who is running for re-election next year and is already facing a crowded field of opponents. One of the contenders for the city’s top job is Garry McCarthy, whom Mr Emanuel sacked as boss of the CPD in the wake of the Laquan McDonald scandal. Mr McCarthy is likely to run mainly on crime—until now, one of Mr Emanuel’s biggest weak spots.