The Algorithm Knows Where, Not Why

townhall.com

America's crime conversation has a data problem — not a shortage of data, but an overconfidence in what data can actually tell us. Homicides fell 17.7 percent in Q1 2026 across 67 major law enforcement agencies tracked by the Major Cities Chiefs Association, with violent crime broadly continuing its post-pandemic retreat. Homicides dropped nearly 20 percent for all of 2025, the largest single-year decline on record. Politicians on both sides are already claiming credit. That's the tell. When everyone claims credit for good numbers, it's a safe bet that no one fully understands what's driving them.

I graduated from Northeastern University with a criminal justice degree in 1990 and have spent 35 years watching crime patterns evolve — from the 1992 Los Angeles riots, where I drove in the presidential motorcade for George H.W. Bush, to the fentanyl-soaked neighborhoods of present-day California. Criminal justice is an exercise in pattern recognition, but the pattern is never the whole story. The distinction matters as law enforcement agencies rush to adopt AI-driven predictive tools and policymakers cite crime statistics the way Gordon Gekko cited earnings reports — selectively, with supreme confidence that they've spotted the signal.

Consider the most clarifying example in recent memory. In December 2025, the House Oversight Committee released a report finding that DC Metropolitan Police Chief Pamela Smith pressured and at times directed commanders to manipulate crime classifications — downgrading serious offenses to lesser categories so they'd vanish from public reporting. The DOJ accused Smith's department of "placing a higher priority on suppressing public reporting of crime statistics than stopping crime itself." Smith resigned in December 2025. By May 2026, 13 MPD officials faced termination. The city's celebrated crime-drop numbers had a significant asterisk. This is what happens when the people running the engine start steering.

Recommended

Here's what honest data does well: it confirms patterns social science has documented for decades. Family structure matters — cities with higher shares of single-parent households consistently show elevated crime rates, even after researchers control for income. A 2025 peer-reviewed study confirmed that the correlation between non-intact family structures and juvenile offending is "one of the most consistent findings in social science research." Thomas Sowell has been making this argument for 40 years. The data keeps agreeing with him.

The geographic concentration of crime follows the same pattern. Memphis, St. Louis, Baltimore, and Detroit lead large-city violent-crime rankings not because of population size but because of decades of institutional decay and policy failure. Of the 25 largest U.S. cities, roughly 20 to 22 are governed by Democrat mayors. That's a structural fact, not a partisan talking point. The data makes the connection; voters should draw the conclusion.

The most instructive example of data working with judgment is broken windows policing. Introduced in a 1982 Atlantic essay by Wilson and Kelling, applied by Bratton in New York beginning in 1990, it produced measurable results. NBER research confirmed that a 10 percent increase in misdemeanor arrests drove a 2.5 to 3.2 percent drop in robberies. New York's twenty-year crime decline followed. A theory, applied by disciplined leadership, produced it.

Then there's the border. From 2021 to 2024, the Biden administration's policies produced what congressional testimony described as the worst border crisis in American history — more than 10 million known illegal crossings, over 240,000 "gotaways" in 2024 alone, and cartel smuggling revenues that grew from a $500 million industry in 2018 to an estimated $13 billion by 2022, per Homeland Security Investigations. Crime data during that period was compiled in an environment where the foundational inputs — who is in the country, where they are, what offenses are being committed — were systematically obscured by catch-and-release policies and sanctuary city reporting gaps. You cannot run a reliable crime model on deliberately incomplete data.

As an expert witness in investment disputes, I make this same argument to juries: the model is only as good as the inputs, and the inputs reflect human choices. A risk model trained on biased data doesn't reduce risk — it launders it with a veneer of objectivity. The best framework for criminal justice is identical: pattern recognition is the engine; human judgment is the steering wheel. The more consequential the decision — charging a suspect, sentencing a defendant, deploying lethal force — the more it demands adversarial testing, constitutional restraint, and a human being who can be held accountable.

The answer isn't to abandon data-driven approaches. It's to build accountability into how the tools are used. Require transparency in the models. Mandate human review before any enforcement action triggered by algorithmic output. Treat the tools as what they are — decision-support instruments, not decision-makers. Scalia's textualism would demand the same discipline of any legal framework: the rule governs; the tool advises.

The best news is that crime is genuinely dropping. Homicides fell 20 percent in 2025, with Q1 2026 continuing the trend. That progress didn't happen because an algorithm willed it. It happened because enforcement decisions were made, prosecutorial standards were enforced, and in some cities, proven theories like broken windows were re-embraced after years of being dismissed as retrograde. Data described the improvement. Leadership produced it.

Criminal justice has always been part science, part art, and part philosophy. The data tells us where the fire is. The law tells us what we're allowed to do about it. And wisdom — the kind that can't be trained on a dataset, and that can't be faked by a police chief manipulating a spreadsheet— tells us what we should do. Any system that confuses the engine for the driver is going to end up somewhere it didn't intend to go.

Jay Rogers is a financial professional with more than 30 years of experience in private equity, private credit, hedge funds, and wealth management. He has a BS from Northeastern University and has completed postgraduate studies at UCLA, UPENN, and Harvard. He writes about issues in finance, constitutional law, national security, human nature, and public policy.

Editor’s Note: Do you enjoy Townhall’s conservative reporting that takes on the radical Left and woke media? Support our work so that we can continue to bring you the truth.

Join Townhall VIP and use promo code FIGHT to receive 60% off your membership.