Computational Policy Lab
Guarding against bias in charging decisions using AI-powered redaction.
Prosecutors have nearly absolute discretion to charge or dismiss criminal cases. There is concern, however, that these high-stakes judgments may suffer from explicit or implicit racial bias, as with many other such actions in the criminal justice system.
To prevent race from being used in charging decisions, we designed an algorithm that automatically redacts race-related information from free-text case narratives. We piloted our algorithm at the San Francisco and Yolo County District Attorney offices to help prosecutors make race-obscured charging decisions on thousands of incoming cases. Our analysis shows that the redaction algorithm is able to obscure race-related information close to the theoretical limit. Blind charging also helps prosecutors demonstrate that they have taken concrete steps to ensure fairness in the charging process, potentially improving perceptions of procedural justice for their constituents.
In 2022, legislators in California recognized the success of our pilot deployments and passed AB-2778, which required prosecutors across the state to adopt blind charging by 2025. We are now using large language models (e.g., GPT-4o) to automatically redact case records for prosecutors across the country as part of a large scale experiment evaluating its impact on charging outcomes.
Read more about our project at blindcharging.org.
A hypothetical example of our redaction algorithm obscuring information that could be used to infer an individual’s race.
