Rant: My city's new crime prediction software vs the old community board

I live in a neighborhood in Phoenix that got picked for a pilot program using AI to guess where crime might happen. They sent patrols based on the map it made. For six months, I compared it to our old neighborhood watch Facebook group where people just posted about actual weird stuff they saw. The AI kept sending cops to the same three blocks, mostly just hassling kids, while the Facebook group flagged a real pattern of car break-ins three streets over that the AI totally missed. The difference was the AI only looked at old police reports, but real people noticed new things happening right now. Has anyone else seen their local government trust a computer map over what people are actually saying?

3 comments

3 Comments

jade_singh20d ago

My buddy in Oakland had the same thing happen with their predictive policing software. It kept circling his quiet block because of a single old noise complaint, while the actual porch pirates were hitting a new condo building down the road. The algorithm just can't catch stuff that hasn't been logged before, which is wild when you think about it.

-1

zara_garcia20d ago

Ugh, it's like my streaming service only suggesting old shows I've already watched.

uma_johnson12d ago

Read an article about how these systems get stuck in feedback loops. They keep pushing what's already popular in the data, so nothing new breaks through. It's the same reason you see the same ten movies on every service's front page.