Bridging Research Gaps Between Academic Research and Legal Investigations of Algorithmic Discrimination

Legal cases reveal a critical gap between theory and practice: The paper analyzes 15 real U.S. civil enforcement actions and finds that while algorithmic fairness research is "theoretically rigorous," it often fails to address what lawyers and regulators actually need to prove discrimination in court, highlighting a disconnect between academic work and real-world enforcement.

Companies can't easily swap discriminatory algorithms for fair ones: One of the five key research gaps identified is the challenge of "finding an equally accurate and less discriminatory algorithm"—meaning current research doesn't provide practical tools for companies (or regulators) to identify alternatives that maintain performance while reducing bias, a critical need in legal cases.

Discrimination can compound through multiple algorithms: The paper identifies "cascading algorithmic bias" as a major gap, suggesting that discrimination isn't always caused by a single algorithm but can accumulate as decisions flow through multiple systems—a complexity that legal investigations struggle to detect and prove.

Measuring discrimination in algorithms remains unsolved: The research gap around "quantifying disparate impact" indicates that there's no agreed-upon standard for measuring when an algorithm causes illegal discrimination, making it difficult for regulators and courts to establish clear legal violations.

Missing data about protected groups hampers investigations: The paper highlights that algorithms often lack information about race, ethnicity, or other protected characteristics, creating a practical barrier to investigating discrimination—a problem that academic research has not adequately addressed with solutions.