EEOC AI Bias Audit

The term "selection rate" is used to denote the ratio of candidates who are hired, promoted, or selected in some other way. The calculation of the selection rate for a group of applicants involves dividing the number of individuals from the group who are hired, promoted, or selected by the total count of candidates in that group.

To illustrate, imagine a situation where 80 White applicants and 40 Black applicants complete a personality test that is assessed via an algorithm for a job application, with 48 White and 12 Black candidates moving forward in the selection procedure. In this context, the selection rate for White applicants would be calculated as 48 out of 80, equating to 60%, and the selection rate for Black candidates would be 12 out of 40, which is equivalent to 30%.

The four-fifths rule is a guiding principle introduced by the Equal Employment Opportunity Commission (EEOC) and featured in the Uniform Guidelines on Employee Selection Procedures to identify potential adverse impact or discrepancies in hiring rates. According to this rule, the recruitment rate for any demographic should not be less than 80% (four-fifths) of the rate of the group with the highest recruitment rate, to ensure fairness in selection processes.

However, it's important to note that the four-fifths rule isn't an absolute assurance of the absence of disparate impact. Even with smaller variances in recruitment rates, there may still be an indication of adverse impact, especially in cases where the selection tool is utilized significantly or when the employer inadvertently dissuades particular candidates from applying. Furthermore, even if a recruitment tool aligns with the four-fifths rule, it may still be found to possess an unlawful adverse impact if it results in a statistically significant disparity in selection rates.