San Francisco is preparing to use a new artificial intelligence (AI) tool that combats racial bias in police reports.

The bias mitigation tool, which was developed by a team at the Stanford Computational Policy Lab, relies on AI algorithms to automatically redact information that identifies a suspect’s race in police reports, potentially eliminating racial bias among prosecutors.

Using computer vision and named-entity recognition, the tool recognizes words and phrases suggestive of race, eye and hair color and other descriptors and automatically redacts them. The bias mitigation tool will also redact names, neighborhoods and locations mentioned in the reports, so as not to suggest that a criminal suspect is of a specific racial background.

According to San Francisco District Attorney George Gascón, seeing the name Hernandez mentioned in a report could signal to prosecutors that the suspect is of Latino descent, potentially impacting an investigation. Additionally, the tool will also redact identifying information about police officers mentioned in the report, including removing badge numbers, so that information does not influence prosecutors either.

Identifying data is automatically replaced with generic words and phrases such as “location” or “officer #1” as stand-ins for that information.

The bias mitigation tool was developed by the same group who helped create the Patternizr software that enables New York Police Department officers to comb through thousands of database case files in search of possible crime patterns across the department’s 77 different precincts.

The launch of the bias mitigation tool, which is slated for July 1, follows closely on the heels of a recent move by city officials in San Francisco to ban the use of facial recognition technology by law enforcement in the city amid concerns that the tools are inaccurate and biased.