Mitigating bias in charging decisions with automated race redaction.
Prosecutors have nearly absolute discretion to charge or dismiss criminal cases. There is concern, however, that these high-stakes judgements may suffer from explicit or implicit racial bias, as with many other such actions in the criminal justice system.
To reduce potential bias in charging decisions, we designed a new algorithm that automatically redacts race-related information from free-text case narratives. In a first-of-its-kind initiative, we deployed this algorithm at the San Francisco District Attorney’s Office to help prosecutors make race-obscured charging decisions on incoming felony cases.
Our analysis shows that the redaction algorithm is able to obscure race-related information close to the theoretical limit. For example, the redacted case files leak roughly the same amount of information about one’s race compared to what one can infer from the alleged crime alone. We are working to expand to other jurisdictions in California, with an eye toward a wider open-source release of our tool.
A hypothetical example of our redaction algorithm obscuring information that could be used to infer an individual’s race.