A recent study by Stanford University Human-Centered Artificial Intelligence found substantial evidence of racial disparities in AI-based candidate screening. These disparities are exacerbated by the dominance of a few large vendors in the hiring and recruiting space.
The researchers measured for adverse impact using the "four-fifths rule", which flags a position when one group is recommended at less than 80% of the rate of the most-recommended group.
26% of Black applicants and 15% of Asian applicants applied to positions where the AI system discriminated against their racial group. To put this in perspective: if the AI had recommended Black and Asian candidates at the same rate as it recommended the most-favored group (typically white applicants), 40,000 more applications would have advanced to the next stage of hiring.
An organization cannot both say it practices DEI, or abandon DEI efforts, saying that it relies on meritocracy, when the very tools and systems it operates are not merit-based.
Hiring and firing have long been entrenched in implicit racial and ethnic biases that have very real consequences for marginalized talent.
- Black unemployment rate has been exactly double the white unemployment rate for over 70 years, including today. (Brookings)
- Résumés with Black-sounding names receive 50% fewer callbacks than identical résumés with white names. (Bertrand & Mullainathan, American Economic Review)
- Black employees are overrepresented at 18% in frontline jobs and underrepresented the higher up the rung you go (McKinsey)
This isn't just a negative systemic impact on a large population of workers, but it also means that organizations are not always hiring the best candidate for the job.
Let's talk referral culture.
It sounds like a reasonable solution to a saturated candidate pool and one where an organization wants good employees to recommend good candidates. Except that referral culture simply amplifies biases based on who is in your social network.
An analysis of over 2 million job candidates by consulting firm Paradigm found:
- Referred candidates are 4.5x more likely to be hired
- Women of color are 35% less likely to receive referrals
- White candidates are nearly twice as likely to be hired as candidates from other racial/ethnic backgrounds, even in organizations that "prioritize diversity."
Now, we are magnifying these biases and scaling it through the use of AI. As talent professionals, hiring managers, and organizational leaders, this should not just be concerning; it should be a foundational relook at your ATS.
What questions have you asked your vendor about how they ensure their AI isn't scaling bias or rejecting qualified candidates? How have you trained your HR professionals to validate the talent pool rejected by AI? Are you reviewing the candidates that weren't referred?