Hiring Prejudice in Application Screening Systems by Dr. Timothy Smith

Photo Source: Unsplash

With AI permeating every aspect of life from work and home to sports and entertainment, it stands to reason that AI will also command a growing impact on the job application process. The power of AI models to read job applications far faster than human screeners has significantly changed how human resources (HR) deparments handle job applications in both the public and private sector. A recent research paper titled, “Algorithmic Monocultures in Hiring,” authored by researchers at Stanford, Chapman, and Northeastern Universities uncovered a startling inequality in the applicant screening technology that affects millions of job applicants across the US and the world. (algorithmichiring)

Over 90% of employers in the US use computer programs to initially screen job applicants to sort through the flood of applications that arrive daily online. The online access to job applications has opened the door to a vast number of applicants from around the world. Sorting the promising applications from the unqualified ones requires significant attention from recruiters. A small team of recruiters cannot possibly read thousands of applications in a day, but AI-enabled tools can read thousands of applications in a matter of seconds. The high volume of applications justifies automated hiring algorithms also called Applicant Tracking Systems (ATS). According to research conducted by Jobscan, 97% of Fortune 500 companies use ATS. (jobscan)  

A deeper problem has emerged since most recruiting groups depend on only a handful of tools from the private sector. Popular ATSs include companies such as Workforce, SuccessFactors, and Oracle, making up nearly 60% of the ATS market share. With only a few companies processing most of the applications, the underlying algorithms impact application success across companies leading to systemic bias in the recruiting system. The research mentioned above in “Algorithmic Monocultures in Hiring,” uncovered deeper implications for millions of job seekers. The research looked at data from 3.4 million real job applicants submitting 4 million applications to 156 employers across 11 market sectors. By looking at individual applicants and specific jobs, the researchers uncovered "systemic rejections" in real hiring AI data. They found that applicants rejected by an algorithm at one employer appears very likely to be rejected at others too, because the same underlying system makes the call. In other words, an applicant does not apply to individual companies because the screening occurs in a central screening company not the company she applied to. 

With nearly all employers relying on a handful of companies to screen applicants, a few AI systems make the decision about who gets access to millions of jobs. The applicant may think that they have a fresh chance every time they apply for a new job at a new company; when they may get summarily rejected based on an earlier rejection. Additionally, these central algorithms compound racial bias in hiring. The researchers found increased recommendation for rejection for certain positions applied to by black and Asian applicants. Additionally, 4% of applicants that apply for 10 jobs get recommended for rejection. This rejection rate exceeds that of chance, implying a systemic problem. At the heart of the findings of bias in ATS systems lies a silent system quietly influencing an enormous share of who gets hired in America.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.