AI in hiring: Innovation or inequality in disguise?

Artificial intelligence has shifted from the periphery of business activities to the heart of business, transforming the way companies screen, assess, and recruit talent. Currently, AI is not just a technology; it is a gatekeeper. The latest projections, as cited by Brookings, suggest that close to 98.4% of the Fortune 500 use AI in the recruitment process, with one firm saving over a million dollars in a year through automated interviews.

In the case of smaller businesses, the rate of adoption is growing at a fast pace and is expected to increase from 51% to 68% by the end of 2025, thanks to the efficiency and cost-saving benefits. However, behind these promising figures, a disturbing question arises: What is the true cost of outsourcing human judgment to computers?

The promise and peril of AI in recruitment

Although AI holds the promise of speed and efficiency, it is anything but objective. “Proprietary recruitment software, shrouded in secrecy, can perpetuate and even exacerbate societal biases embedded in the historical patterns of employment.” An independent study is made difficult by a lack of access, so that claims of “bias reduction” remain largely unproven. But a closer examination reveals a disturbing truth: AI can perpetuate discrimination, and it does so in a subtle yet systematic way that impacts millions of job seekers each year.

Investigating Large Language Models in hiring

Our study focused on large language models (LLMs), increasingly used in both proprietary and open-source hiring tools. By simulating resume screening, the crucial first step in candidate evaluation was investigated by Brookings whether AI systems treat applicants differently based on social identity. Over 550 job descriptions and resumes were analyzed, each paired with 80 names associated with Black men, Black women, white men, and white women, to evaluate candidate selection patterns.

“Using names to signal social identity is a common approach which has revealed discrimination in mortgage lending, online ad delivery, and hiring,” the study notes.

Stark evidence of bias

The results were striking. Take a look at the results here:

• Gender bias: In 63% of tests, men’s names were favored 51.9% of the time, while women’s names were selected just 11.1% of the time.

• Racial bias: White-associated names’ resumes won 85.1% of the time, while Black-associated names’ resumes won only 8.6% of the time.

• Intersectional bias: When both race and gender are taken into account, the group that faced the greatest disadvantage was Black men, who were chosen only 14.8% of the time compared to Black women and 0% compared to white men.

These results point to a disturbing truth: AI not only mirrors existing imbalances, but it also has the potential to magnify them, ensuring that disadvantages are perpetuated in employment contexts where biases may not have existed before.

Balancing efficiency and fairness

The potential of AI recruitment is quite interesting in terms of efficiency, speed, and consistency. But the risk is also very real. If left unchecked, this technology could lead to a meritocracy that is blind to human concepts of fairness, preferring candidates whose names, gender, or racial background happen to align with the preferences of the algorithm.

The potential of artificial intelligence is quite interesting. No doubt. But we cannot avoid the risk being very real. If left unchecked, the technology can lead to a meritocracy that is completely blind

As AI is increasingly becoming an essential part of the hiring process, the challenge is clear: How to use AI not as a reflection of bias, but as a genuine amplifier of fairness.Ready to navigate global policies? Secure your overseas future. Get expert guidance now!