By 6 a.m., before most Americans have poured their first cup of coffee, software has already ranked them. Not by credit score. Not by zip code. By the probability that they will do something—or become something—worth watching. Predictive intelligence targeting teams, the bland corporate name for this machinery, are now embedded in police departments, insurance conglomerates, employers, and even school districts. And almost nobody agreed to it.
Here's how it works. These teams ingest everything: arrest records, 911 call logs, social media posts, utility payments, shopping histories, even the metadata from your phone. Then machine-learning models spit out lists. Who's likely to commit a crime. Who's likely to file a costly claim. Who's likely to quit, or sue, or blow the whistle. The output isn't evidence. It's a suspicion, dressed up in a spreadsheet and treated like fact.
The most notorious example remains PredPol, later renamed Geolitica, which directed patrols to "high-risk" blocks. Studies in cities like Chicago and Los Angeles found the system mostly sent cops back to the same poor, Black and brown neighborhoods they already over-policed. The data feeding the model came from arrests, and arrests came from where cops already were. The algorithm learned our worst habits and called it science. Meanwhile, ShotSpotter sensors have triggered thousands of stop-and-frisks in neighborhoods where the devices were placed—not where gun violence was statistically highest.
Corporate America is worse, because there's no badge to sue. A 2023 investigation by The Markup found that UnitedHealth's algorithm flagged Black patients as needing extra care less often than equally sick white patients—not out of malice, but because it used health care spending as a proxy for illness. Employers now use "turnover prediction" scores to decide who gets promoted and who gets quietly managed out. Amazon scrapped a hiring algorithm in 2018 after it taught itself to penalize résumés containing the word "women's." These aren't bugs. They're the product.
What makes predictive intelligence targeting teams different from old-fashioned profiling is scale and secrecy. A human bigot can only hurt so many people in a day. A model can score millions before lunch, and the company that built it will call it a trade secret. Defendants have a constitutional right to confront their accusers, but try cross-examining a neural network. Courts in Wisconsin and New Jersey have already split on whether defendants can even see the source code that put them in prison.
And the targets are not abstract. They are the father in Cleveland denied a job because a wellness app sold his sleep data. They are the teenager in Tampa put on a "chronic offender" list at fourteen for a fight in a parking lot. They are the woman in Phoenix whose insurer dropped her after a predictive model decided her pregnancy was "high-cost." This is not the future. This is a Tuesday.
We built a country on the promise that you are innocent until proven guilty, that you get to face your accuser, that you can start over. Predictive targeting doesn't break those promises. It makes them quaint. The score follows you like a second shadow, and you will never see it, never challenge it, never know why the door closed.
The fix isn't tweaking the algorithm. It's demanding disclosure, audits, and an outright ban on using opaque risk scores for decisions about liberty, employment, and health care. Until then, we are all suspects.
**Our take:** A society that pre-punishes people for crimes they haven't committed isn't safer—it's just more efficient at cruelty. If you don't know your score, someone is already deciding your future with it. That's not intelligence. That's a caste system with better branding.