In cities across the United States, police departments are deploying algorithmic tools that predict where crime will happen before it does. They're using these predictions to allocate patrols, justify stops, and generate what amount to pre-crime watch lists — not of individuals who have committed crimes, but of individuals the algorithm has flagged as statistically likely to be involved in future criminal activity.
This is not science fiction. Chicago had a Strategic Subject List. New Orleans contracted with Palantir to build a predictive surveillance network. Los Angeles used PredPol — a commercial predictive policing software — for years. The tools have different names. The logic is the same: train an algorithm on historical arrest data, let it identify patterns, and deploy resources accordingly.
Here's the fundamental problem that no jurisdiction using these tools has honestly answered: historical arrest data is not neutral. Decades of over-policing in Black and brown neighborhoods means those neighborhoods generated disproportionate arrest records — not because those communities had more crime, but because they had more police. When you feed biased data into an algorithm, the algorithm doesn't correct for the bias. It amplifies it. It sends more police to the same neighborhoods. Those police generate more arrests. Those arrests feed back into the algorithm as confirmation.
The ACLU's documentation of predictive policing programs shows a consistent pattern: the communities flagged as high-risk are consistently the communities that were already over-policed. The algorithm didn't find new crime. It found the old injustice and called it math.
And because it's math — because there's a model, a score, a probability — the decisions feel objective. Officers aren't making prejudiced judgments; they're following the data. Except the data was shaped by prejudice from the start.
The most dangerous part isn't even the policing. It's that these systems are beginning to inform decisions outside law enforcement — employment screenings, housing applications, insurance risk assessments — in ways that are not disclosed and not consented to.
The algorithm already decided what kind of person you are. You weren't there for the trial.
SOURCE LINK: https://www.aclu.org/report/dawn-robot-surveillance
Comments
Post a Comment