When AI Redlines Your Neighborhood Without a Map

Episode 51 explores a new frontier of algorithmic bias: predictive policing systems that assign risk scores to neighborhoods based on historical arrest data, creating feedback loops that reinforce over-policing in communities of color. Lucas and Luna break down a 2025 study from the AI Now Institute analyzing Palantir's Gotham platform in three mid-sized cities. They uncover how 'hot spot' predictions trained on biased reporting data lead to more patrols, more arrests, and more biased data. The episode delves into the technical mechanism of label bias, the difficulty of auditing opaque contracts, and the growing movement for community oversight of public-sector AI. Specific numbers include a 34% higher false-positive rate in low-income census tracts and a 22% increase in misdemeanor arrests after system deployment. No prior episode covered predictive policing specifically, making this a timely and urgent angle for listeners concerned about fairness in AI systems. #PredictivePolicing #AlgorithmicBias #AIEthics #CriminalJustice #Palantir #GothamPlatform #RacialBias #FeedbackLoop #LabelBias #PublicSectorAI #AIAccountability #CommunityOversight #DataJustice #PoliceReform #AI2025 #Technology #FexingoBusiness #BusinessPodcast Keep every episode free: buymeacoffee.com/fexingo