Photo courtesy of Igor Omilaev on Unsplash
Going in for a job interview no longer consists of speaking with a person. Instead, applicants find themselves conversing with computer-generated programs as they are examined based on keywords, eye contact and other mannerisms. Artificial Intelligence is choosing the workforce of the future and deciding what qualifies as a good fit for open positions without the invaluable input of a human executive.
After the impact of the COVID-19 pandemic, a Gartner Inc. survey among HR leaders stated that 86% of employers are now using AI technology in the hiring process. This software is specifically programmed to examine facial cues, tone of voice and timed responses to questions and pick out specific words or phrases within them. Rather than facing a human with empathy and patience, applicants may have their responses cut short or discussion topics abandoned at the will of increasingly unavoidable AI hiring systems.
Not only does this remove the entire aspect of human connection and charisma within interviews, but AI is imperfect due to its programming method. Almost 50% have been proven to show gender bias leaving applications from women to become automatic afterthoughts even in instances where women exceed qualifications. Additionally, evidence of both racial and gender bias are found in 25.7% of AI systems. Women, more specifically women of color, face unfair treatment due to this software and its defects, preventing them from having the same chances and opportunities as the most popular group chosen by AI: white men.
This is partially a result of the fact that men dominate 78% the field of AI creation. The programs are not only biased but also fail to perform as well for women and non-binary people. The voice recognition systems do not function correctly, and job opportunities and information are not distributed to women like they are distributed to men.
A Princeton study testing the extent of AI bias proved that these systems inherently push applications with European-sounding names over those with African-sounding names. European names were “significantly easier to associate with pleasant than unpleasant terms” while African American names were connected with words related to violence, sickness, crime and more. AI also struggled to understand different dialects spoken during interviews, repeatedly associating the African-American vernacular in English with Norwegian.
As AI incorporates the same biases and stereotypes its creators possess, it becomes just as, if not more problematic than gender and racial bias in the working world before the implementation of AI applications. While society continues to grow and change, there are means of defense against discrimination in the work world. And yet, there is no defense against programmed AI that is published with bias, as it is often not re-examined and continues to affect a countless number of applicants.
Furthermore, job listings themselves are appearing less and less in person and more through online platforms with confusing navigation systems and unfairly distributed opportunities. More importantly, these listings do not properly reflect the extent of actual job opportunities, as entire groups are excluded from being advertised certain openings due to racial or gender bias.
When examining the recruitment process through AI systems, it was discovered that the first wave of applicants often consists of white men and through interacting with them, the algorithm will recommend more applications from white men, eventually erasing applications from any other group of people.
AI isn’t working. While it might be more convenient for employers, it is significantly more harmful to applicants as gender and racial bias dominate which candidates are submitted for the recruitment process. Not only does the process create a total disconnect between employers and to-be-employees, but it also prevents opportunities from being distributed equally. AI has no place in the work industry and as long as it is biased, it should be removed from the hiring process entirely.
