Artificial intelligence is gaining an increasingly central role in hiring, but it is important to consider what this means for the workforce. New data from one of the most comprehensive independent studies on the technology showed that the use of AI hiring tools can further racial biases that we have long vowed to eradicate. The study, led by Stanford University alongside Chapman University and Northeastern University, found substantial evidence of racial disparities in AI-based candidate screening.
Data suggests that nearly 90% of US employers use AI tools in their initial screening of candidates, and due to the prominence of some tools over others, most turn to the same set of platforms, which bring their hiring algorithms and biases along with them. Not only does this mean that candidates lose out on one job, but they also stand to face similar risks with every application they submit. Acknowledging this racial disparity is crucial for employers who have started using AI screening tools to identify top talent on the market.
One of the largest studies on AI hiring algorithms found substantial evidence of bias and an adverse impact on specific racial groups. (Image: Pexels)
Using AI Tools in Hiring Introduces a Risk of Racial Bias in Your Operations
The Stanford-led study on the use of AI in hiring followed 3.4 million job applicants who submitted 4 million applications to 1,700 job postings across 156 employers and 11 industry sectors. The applications were reportedly all screened by the same vendor, Pymetrics, which is a popular platform that uses “neuroscience and machine learning” to help employers screen candidates based on task performance across an elaborate set of mini-games. While this approach has been championed as one of the best ways to hire based on social, cognitive, and behavioral features with an unbiased approach, new data shows that the AI tool still furthers biases in hiring.
The paper, titled Algorithmic Monocultures in Hiring, is set to be presented in Montreal next month, but the data gives us a head start on diving into the findings. With the job market shrinking and roles fast disappearing, candidates are being forced to compete for fewer job opportunities. Add in the threat of unreported racial biases, and some sections of workers find themselves at a severe disadvantage.
How Did the Study Identify Racial Disparities in AI-Assisted Hiring?
Dialing in on algorithm bias in hiring is harder than it looks, as these tools operate at unprecedented scales and are used in differing ways across employers. Seen as one united screening vendor, the disparities may not be as obvious, but when split among employers and roles, the impact becomes more obvious. Still, the expansive nature of the recent study took the Equal Employment Opportunity Commission’s (EEOC) “four-fifths rule” under Title VII of the US Civil Rights Act into account. This indicates an adverse impact when a group is recommended at less than 80% of the rate of the most-dominant group of applicants.
The study found that 26% of Black applicants and 15% of Asian job seekers applied for positions where the AI systems 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,” the study explained. As things stand, applicants need to apply for 25 positions before they can ensure at least one recommendation for further evaluation.
AI hiring algorithms put workers at a disadvantage across jobs. The study also refers to the harms of “algorithmic monocultures,” where the same tools are used across organizations for screening. This might be good business for the vendor, but for applicants who apply for multiple jobs without any control over the screening tools used, this can mean higher chances of rejection across jobs, more so than if their applications were being reviewed manually.
Employers Should Be Responsible for the Tools They Use in Hiring
The decision to use AI in hiring comes with risks, both for candidates and employers. These tools may be adopted as a way to fairly screen candidates, but their impact still needs to be studied in greater detail to ensure they serve the purpose they are used for. Discriminatory suggestions, prompted by AI recruitment tools, may not be as apparent at first, but these algorithms are prone to absorbing skewed data and patterns that are best left in the past.
Locking employees out of job opportunities is detrimental to the landscape of employment as a whole, and it also introduces legal risks for employers who could be held accountable for potentially biased hiring. Even if federal regulations appear unlikely, states like Colorado and California are working to introduce worker protections to ensure that such a risk of bias in hiring via AI tools is restricted.
Regardless of the emergence of dedicated regulations, existing employment laws protect job seekers from discrimination in employment, and employers could soon find themselves in a tight spot without strict oversight on hiring practices.
Subscribe to The HR Digest for more insights on workplace trends, layoffs, and what to expect with the advent of AI.
