Justice by the Numbers: Meet the Statistician Trying to Fix Bias in Criminal Justice Algorithms

Abstraction: Kristian Lum using statistics to expose and fix bias in criminal justice algorithms

Key points:

• Kristian Lum, lead statistician at Human Rights Data Analysis Group (HRDAG), applies statistical methods to algorithmic bias in US criminal justice
• PredPol analysis: when fed Oakland drug crime data, algorithm directed police almost exclusively to poor minority neighborhoods, reproducing historical policing biases
• Police killings study: estimated ~10,000 US police killings 2003-2011 (~1,500/year), roughly half unreported in national data — government estimate of 7,400 was an underestimate
• Cash bail study (with Stanford's Baiocchi and NY Legal Aid Society): setting bail increased conviction probability by more than a third, primarily via guilty pleas
• Raises question whether improving biased algorithms perpetuates structural injustice vs. addressing root causes

Connections: Kristian Lum · Hrdag · Predpol · Algorithmic Fairness · Predictive Policing · Bias In AI

Source: https://psmag.com/social-justice/justice-by-the-numbers-meet-the-statistician-trying-to-fix-bias-in-criminal-justice-algorithms