AI hiring tools can still have racial biases, study finds
About 90% of companies are now using AI in the hiring process, according to the World Economic Forum. A new study from the Stanford Institute for Human-Centered AI finds that one AI hiring tool used by lots of companies is often biased against Black and Asian applicants.
Anyone who’s been job searching knows it’s rough out there. Companies aren’t posting many jobs, and when they do, they get tons of applications.
“They almost have to use an algorithm in some way to be able to look through the number of applications that are coming in,” said Sarah Bana, an assistant professor at Chapman University and one of the co-authors on the Stanford study.
The AI tool that she and her study co-authors looked at is called Pymetrics; Marketplace was unable to reach them for comment by publication time. The tool screens applicants by having them play behavioral video games. She said the company behind it actively tries to eliminate bias, “but what we find is they still end up with racial bias in the algorithm.”
The AI discriminated against Black applicants in about 25% of jobs and against Asian applicants in about 15% of jobs. This is a tool that many companies are using.
If you apply to five different places, “it could be that the same system is actually screening your resume at all five of those,” said Kathleen Creel, an assistant professor at Northeastern University and one of the co-authors on the Stanford study. “And so these five supposedly independent chances are no longer independent.”
This is a real problem, per Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies.
“Unless companies take steps to really include human review and sort of manage how these systems are being deployed internally, they’re going to replicate patterns all across industry,” he said.
And that means that people who are already marginalized in hiring could be marginalized even further.
