COMPAS errors fall unequally on Black and White defendants
Evidence
Looking at the defendants whom the model misjudged revealed a startling disparity: “Black defendants were twice as likely to be rated as higher risk but not re-offend. And white defendants were twice as likely to be charged with new crimes after being classed as lower risk.” — Brian Christian, The Alignment Problem, 279
Categorization
- corroborated: 2 independent passages support this claim — Brian Christian, The Alignment Problem, 279
- XREF: connects to: COMPAS misjudged Black and White defendants in systematically opposite ways. — Brian Christian, The Alignment Problem, 279; COMPAS assigned higher risk scores to Black defendants who did not reoffend — Brian Christian, The Alignment Problem, 36; Prior biased policing leaks into race-neutral variables like conviction history — Brian Christian, The Alignment Problem, 296 (+1 more)
Related
- COMPAS misjudged Black and White defendants in systematically opposite ways. (similarity 0.96)
- COMPAS assigned higher risk scores to Black defendants who did not reoffend (similarity 0.94)
- COMPAS scored Black defendants as falsely high-risk nearly twice as often (similarity 0.89)
- Prior biased policing leaks into race-neutral variables like conviction history (similarity 0.81)
- ProPublica tested COMPAS recidivism predictions against actual reoffense outcomes for bias (similarity 0.80)
- Racial sentencing disparities do not reveal which group is treated wrongly (similarity 0.78)