In courtrooms across the United States, judges frequently use algorithmically generated risk assessment reports to gauge the likelihood of defendants reoffending and assist in determining sentence severity. A study by ProPublica analysing COMPAS scores assigned to 7,000 people arrested in Florida between 2013 and 2014 found that the algorithm contained an internalised bias that incorrectly predicted Black people were more likely to become repeat offenders. After running tests that isolated race from criminal history, age, gender, and recidivism, Black defendants were still 77 per cent more likely to be classified as high risk for future violent crime and 45 per cent more likely for any future crime. Northpointe, the firm responsible for the algorithm's 100-question assessment system, defended its methodology and stated that it disagreed with ProPublica's analysis and claims.

Background: Predictive Policing and Risk Scoring

← Back to the news feed