AI used to uncover racial covenants in Santa Clara County property records
A Stanford University project is shedding light about patterns of housing discrimination in the South Bay by using the power of artificial intelligence.
Genevieve Singh works in the Santa Clara County Recorder's Office. Lawmakers in 2021 mandated every county in California to cull their property records to identify and redact racially restrictive covenants.
"No person of African, Asiatic or Mongolian descent shall be allowed to occupy said property," stated a property deed in the Santa Clara County Recorder's Office archives. It's that kind of racist language county officials are required to identify.
"We would actually review either books or the digital reel to locate restrictive words," said Singh.
It's a daunting task. The county has 24 million deed documents totaling 84 million pages, dating back to 1850.
"Once we locate them, we work with the county council to send that up to get it approved. Once it was approved, it was then re-recorded," said Singh.
But Singh was just one of two county employees reading through these deeds. To comply with the law, some California counties contracted with commercial vendors.
Los Angeles County, for instance, hired a company with an $8 million to conduct the scan over seven years.
Stanford Professor Daniel Ho began to wonder if there was a more efficient way to document these racial covenants. He bought a home in Palo Alto and was somewhat surprised to come across the racially explicit language.
"It was obviously disturbing to read the language and realize this about the history of the very home you're about to move into," said Ho.
Ho, the director of the university's Regulation Evaluation Governance Lab, which aims to modernize government, began working on an AI language model to speed up the process.
Researchers trained the model using thousands of racial covenants from across the country.
"AI really opens up this opportunity to study this, map this, and create a historical registry," said Ho.
The lab put it to work, scanning 5.2 million records between 1902 and 1980. It took about a week. They estimate that one in four homes in the county by 1950 were subject to racial covenants.
"The presence of these racial covenants can have long lasting effects due to the kind of community that they signal," said Ho.
Researchers determined only 10 developers were responsible for a third of the identified covenants, suggesting that developers had a lot of say in how Santa Clara County was constructed.
The REGLab says the AI model will save more than 86,000 person hours for the county.
"It's really hastened the process. Instead of us going through everything with a fine-toothed comb and looking for these words, they've identified it so now we can just do quality control and move ahead with that package to county counsel to get the process complete," said Singh.
Researchers also found racial covenants excluded African Americans at the same rate as Asian Americans, even when African American residents were less than a tenth the size of the Asian American population.
The team also found a striking instance of a San Jose-owned cemetery that included burial deeds only for "Caucasians."
The REGlab says this goes against the conventional historical account that racial covenants were used only between private parties after the Supreme Court banned state-based racial zoning.
The team has released the paper online and is making the model available to enable all jurisdictions faced with this task to identify, redact, and develop historical registers of racial covenants more effectively.