Neighborhood Knowledge in the Digital Age: Racial Patterns in AI Housing Recommendations

Megan Evans , Max Planck Institute for Demographic Research
Noli Brazil, University of California, Davis

More urban residents rely on online tools to search for housing, creating potential to either interrupt or reinforce racially influenced assumptions linking neighborhood racial composition to assumed neighborhood quality. Artificial intelligence (AI) now represents a new intermediary in how neighborhood knowledge is created and distributed through housing platforms and search engines. This study examines how AI systems characterize and recommend neighborhoods by analyzing responses to housing queries across multiple commonly used AI models in several large U.S. cities. Merging AI-generated recommendations with census data, we investigate whether AI systems incorporate user characteristics and neighborhood demographics into their housing guidance. Preliminary findings suggest that AI recommendations reflect and may amplify existing patterns of residential segregation through systematic differences in how neighborhoods are described and recommended to different users. Results have important implications for informing both fair housing policy and the responsible development and regulation of AI systems in the housing market.

See extended abstract

 Presented in Session 94. Flash Session Data and Methods in Internal Migration and Urbanization