Promises and Limits of Inferring Protected-Class Data for Disparate Impact Testing of AI Systems

On September 8, 2023, DLA Piper and ORCAA jointly hosted Promises and Limits of Inferring Protected-Class Data for Disparate Impact Testing of AI Systems, a daylong, invite-only symposium focusing on race and ethnicity inference methodologies within the context of algorithmic fairness analyses. Experts from diverse fields, including academia, law and regulation, insurance, voting rights, fair lending, and statistics, gathered to share their work on the topic. The primary objectives were to facilitate a comprehensive exchange of insights concerning the implementation of these inference techniques across various sectors and to pinpoint common challenges and unanswered questions across the application areas. The conference culminated in a whitepaper which serves as a comprehensive summary and an official report of the conference proceedings, including a summary of technical discussions, lessons learned, and future directions for work.