A bias audit is a structured assessment of whether an employment process or automated tool produces materially different results across demographic groups. In recruiting, it can examine candidate selection rates, scoring patterns, data, criteria, accessibility, and the way people use a system. A legal bias audit has a defined jurisdiction and method; a voluntary fairness review may cover a broader set of questions.<\/p>\n
The term has no single universal method. Its meaning depends on the tool, employment decision, demographic data, and governing law. New York City Local Law 144 sets a specific requirement for covered automated employment decision tools used in hiring or promotion. Other reviews may follow employment-discrimination law, a risk framework, or an internal standard.<\/p>\n
A completed audit does not prove that a process is fair in every setting. Results apply to the tested version, data, population, time period, and use case. Recruiters still need job-related criteria, accessible processes, clear records, and a way for people to question or correct an automated result.<\/p>\n
A useful audit begins with a system map. The reviewer identifies where automation affects sourcing, ranking, screening, assessment, recommendation, or progression. The map should show the inputs, outputs, decision owner, downstream action, and point where a person can intervene.<\/p>\n
The audit may examine:<\/p>\n
The auditor documents the method, assumptions, exclusions, and limits. Corrective action should name an owner and a retest date. A ratio without the tested decision or data offers little operational guidance.<\/p>\n