What is Algorithmic Discrimination?

In practice, algorithmic discrimination can appear in model outputs, thresholds, ranking, recommendations, automated decisions, or downstream business rules that produce disparate impacts for protected classes. Compliance teams need to assess data quality, feature selection, training and test outcomes, proxy variables, and human oversight to identify whether an AI system creates unequal treatment even when sensitive attributes are not explicitly used.

This concept is especially important in employment, lending, housing, insurance, healthcare, and other high-impact settings where discriminatory effects can trigger legal and regulatory exposure. It is directly relevant to the EU AI Act’s risk management and data governance requirements, U.S. employment and consumer protection enforcement, and broader governance frameworks such as NIST AI RMF and ISO/IEC 42001, which emphasize fairness, accountability, and monitoring for harmful bias.