In legal contexts, particularly within the United States, discrimination in AI hiring is often assessed through the lens of disparate impact. This occurs when a seemingly neutral hiring process disproportionately excludes members of a protected group, such as a specific race, gender, or age bracket.
The most common technical benchmark used is the four fifths rule. This principle suggests that if the selection rate for a protected group is less than eighty percent of the rate for the group with the highest selection rate, it may be evidence of adverse impact. Regulators and legal experts look for these statistical deviations to determine if a tool is biased.
Beyond simple selection rates, companies should monitor False Negative rates across different demographic groups. If an AI tool incorrectly rejects qualified candidates from one demographic more often than others, it indicates algorithmic bias. Regular audits of training data and model outcomes are essential to ensure compliance with labor laws and to mitigate legal risks. For official guidance on employment law and non-discrimination practices, you should consult the resources provided by the Equal Employment Opportunity Commission (EEOC).
