Facial Recognition Systems in Policing and Racial Disparities in Arrests

• In Event: POLWG Panel 10. Automating Suspicion: Facial Recognition, Welfare Control, and Predictive Policing.

Abstract

This study examines the association between police facial recognition technology (FRT) deployment and racial differences in a1Tests across 1136 U.S. cities in 2016. We estimated doubly robust propensity score models using data from the Bureau of Justice Statistics, FBI Uniform Crime Report, and the U.S. Census Bureau. Results show that agency FRT use contributes to greater racial disparity in arrests. This relationship was underpinned by statistically meaningful and positive FRT effects on Black arrest rates and negative effects on White rates. We also observed more sizeable and significant impacts for adult a1Tests, indicating that FRT's association with adult rates primarily drives the overall disparity finding. Results suggest a need for civic leaders to scrutinize the relative contributions of structural factors, agency policies, and government directives to officer decision-making before widely deploying FRT in jurisdictions. For agencies currently using this technology, it would imply the need for policies and supervision that guide, and in some cases resist, officer discretion in FRT-assisted contexts.

Authors

• Thaddeus Lateef Johnson, Georgia State University/Council on Criminal Justice

• Natasha N. Johnson, Augusta University

• Denise McCurdy, Georgia State University

• Michael Stephen Olajide, Georgia State University