A Stanford report conducted in spring shows that AI hiring tools show racial bias. The esteemed university’s report, “AI Hiring Tools Can Yield Racial Bias and Systemic Rejection,” released May 26, surfaces racial bias at scale and illustrates algorithmic monoculture during the hiring process. The researchers and contributors of this project are Rishi Bommasani, Sarah H. Bana, Kathleen A. Creel, Dan Jurafski, and Percy Liang.
Summer 2026 is currently one of the most difficult labor markets since the COVID-19 pandemic era, 2008’s Great Depression, and, before then, the year 2003. The class of 2026 is on the job hunt, right along with everyone else, as entry-level hiring is hard to come by. Most companies are hiring internally while still legally posting their positions online, as it is law.
A current example of this is a lawsuit involving Derek Mobley, who is represented by civil rights attorney Ben Crump, against Workday, a hiring company that corporations such as Fox use in their hiring process. Mobeley, a 40-year-old Black man, has recently filed the lawsuit in California and said he was “rejected for more than 100 jobs before a person ever reviewed his application,” according to a statement on Crump’s social media. Workday asked a federal judge to toss the case, but the judge is moving forward with the lawsuit as of June 22.
A statement from Crump’s social media said, “An algorithm does not erase accountability. If a hiring tool filters people out by race, by age, or by disability, that is still discrimination. Building it into software does not make it legal. Bias does not get a pass because a machine made the call.”
A key finding in the study is as follows: “We discovered that 26 percent of Black applicants and 15 percent 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 of their applications would have advanced to the next stage of hiring,” said the research page of their website.
Also, the Stanford study focused on the one AI tool that employers use for filtering out applicants, Pymetrics.
“We’ve speculated in past work that if many firms relied on the same AI vendor to screen job applicants, that could prevent some applicants from getting any interviews. But this study was the first time we were able to show this effect in real hiring data,” said Creel.”
Bommansani said the following: “The biases we see arise from a different mechanism than most prior work in hiring. Over the last 20 to 30 years, researchers of the labor economy have shown that human decisionmakers exhibit bias when screening resumes due to implicit biases rooted in demography: names like Jamal or accolades like being the captain of the college softball team are associated with certain demographic groups.
Sarah Bana, who is also an assistant professor at Chapman University, spoke to Our Weekly and said the following about the study.
“In layman’s terms, what we’re finding is that there are certain positions — certain models — that Black applicants are applying to, and their outcomes on those models fall below the standard the EEOC would use to decide whether to investigate a company for adverse impact.”
Bana also spoke about what she thinks employers can do if they are using AI tools to assist them with the hiring process.
“I think it’s very important that employers understand who their specific tool is screening out right now. They can rely on vendors’ claims about whether the tool is biased or unbiased, but they also have their own data that can show whether the screening is inequitable in their particular case. To me, that’s a very natural step for employers to take, because they should want to understand who they are screening out.”
Bana added further context, suggesting that AI chatbots are not the sole issue but automated rejection potentially is.
“Some AI chatbots are actually better at surfacing important information from candidates. In some cases, it’s not something to fight against — it helps recruiters speed up the process and look at candidates they might not have considered.”
She continued, “The case we need to worry about is automated rejections. I don’t want people to walk away thinking AI is bad and there are no jobs. These tools help employers manage a huge number of applications. I really hope people focus on how to show what they can do uniquely, because if you can do something uniquely, there’s probably a job out there for that.”
Creel also said the following about algorithmic monoculture. “Algorithmic monoculture occurs when the same algorithm dominates a sector, or in its weaker but more typical form, algorithms made in similar ways using similar data such that they make similar decisions.”
On the tail end, Bana informed that they’ve seen that Black and Asian applicants are impacted by the Pymetrics AI hiring tools, yet there is no causal evidence yet, according to her response during the Q&A press conference.
A growing number of job applicants online are frustrated and speculate that the job search process is posing difficulty for applicants to actually have an interview due to AI, yet employers are utilizing AI hiring tools for convenience and a smoother hiring process. To read the full report visit https://hai.stanford.edu/news/ai-hiring-tools-can-yield-racial-bias-and-systemic-rejection.
