|
Abstract:
|
Statistical and algorithmic methods are increasingly used throughout the criminal justice system, from setting police patrol schedules to guiding prison sentencing decisions. I'll present two recent applications of this approach: (1) real-time risk assessment in New York City's stop-and-frisk program; and (2) tests for discrimination in traffic stops. For the former, I'll argue that simple, statistically informed heuristics can improve the efficiency, equity and transparency of officer decisions. For the latter, I'll show that popular tests for racial bias suffer from a subtle flaw, and propose a new method for measuring discrimination.
|
