16/06/2026
What if your hiring process was discriminating against candidates but your reporting showed everything was fine?
That's essentially what researchers at Stanford HAI found when they analysed 4 million job applications across 150 employers using the same AI hiring tool.
The headline finding is striking:
26% of Black applicants and 15% of Asian applicants applied to a role where the tool was discriminating against their group. But perhaps the most important finding is this: The bias was largely invisible in aggregate reporting.
It only became apparent when researchers analysed outcomes at the individual position level.
The study also highlights the risks of what researchers call "algorithmic monoculture".
When large numbers of employers rely on the same opaque scoring system, any underlying bias can be replicated across multiple organisations.
This isn't an argument against AI.
It's a reminder that hiring systems should be transparent, measurable and regularly validated.
At ThriveMap, we've long argued that assessment should focus on the work itself. The more closely an assessment reflects the actual job, the easier it is to understand what is being measured, explain decisions to candidates, and identify potential adverse impact.
In our own State of the Assessment Market research, 82% of candidates said they felt more confident about a role after completing a realistic job assessment.
Two takeaways for hiring teams:
① Measure adverse impact at the role and stage level. Don't rely solely on aggregate reporting.
② Be cautious of opaque scores that are difficult to explain, challenge or validate.
The goal isn't simply to automate hiring.
It's to make hiring fairer, more transparent and more effective.
https://hubs.la/Q04lbQtX0