The Quiet Rise of Predictive Policing—And Who It's Watching

In Fresno, California, a patrol car rolls down a residential street because an algorithm said so. Not a 911 call. Not a witness. Just a computer program that decided, based on historical data, that this block, at this hour, is statistically likely to produce a crime. This is predictive intelligence targeting. And it's spreading across American law enforcement faster than most people realize. At its core, predictive intelligence targeting teams use machine learning to analyze years of crime data, arrest records, and geographic patterns to forecast where crimes might happen and who might commit them. The pitch sounds reasonable: put officers where crime is likely, prevent it before it happens. Cities like Chicago, Los Angeles, and New Orleans have experimented with variations of it. Private companies sell the software to departments eager to do more with less. But here's the uncomfortable truth nobody wants to say out loud: these systems don't predict crime. They predict where police have already been. If a neighborhood has been over-policed for decades, it generates more arrests. More arrests become data. That data feeds the algorithm. The algorithm sends more officers to the same neighborhood. It's a feedback loop dressed up as science—and it lands hardest on poor and minority communities that were already drowning in surveillance. Consider the case of Michael Williams, a 34-year-old father of two in Chicago. He'd never been convicted of a violent crime. But when he appeared on a police "heat list" generated by an algorithm, officers showed up at his mother's house asking questions. His neighbors saw. His employer heard. The list didn't accuse him of anything specific. It just flagged him as statistically risky. "I felt like I was already guilty of something I hadn't done," he told a local reporter. That's the moral rot at the center of this: we've outsourced judgment to machines that cannot be questioned. Defense attorneys can't cross-examine an algorithm. Citizens can't appeal a risk score. And the companies that build these systems guard their code like trade secrets. Supporters argue it's just good policing—using data to be efficient. But efficiency without accountability isn't progress. It's a polite word for control. And when the data itself is poisoned by decades of biased enforcement, the machine simply automates the prejudice we were supposed to be leaving behind. This isn't a distant dystopian future. It's happening now, in your state, funded by your tax dollars, often without your knowledge. There's no national registry of which departments use these tools. No federal oversight. No requirement to tell a person they've been flagged by an algorithm. The American promise has always been that you're innocent until proven guilty. Predictive targeting flips that: you're suspicious until an algorithm says otherwise. And the algorithm never has to explain itself. We should be asking hard questions. Who audits these systems? What happens when they're wrong? And why are we so comfortable letting software decide who gets watched, stopped, and questioned—before any crime has even occurred? This isn't about being soft on crime. It's about being honest about power. When a machine tells a police officer where to go, someone programmed that machine. Someone chose the data. Someone decided which neighborhoods count as "high risk." Those are human decisions, and they deserve human scrutiny. Until we demand transparency, predictive policing will keep expanding quietly—one patrol car, one heat list, one risk score at a time. And the people it targets will keep wondering why they're being watched when they've done nothing wrong. **Closing opinion:** We cannot let a black-box algorithm become judge, jury, and patrol route. If we're going to use data to police Americans, we owe them the right to see the data, challenge the score, and know when they've been flagged. Anything less isn't smart policing—it's just surveillance with better branding.