Algorithmic Monoculture: The Hidden Risk in Hiring AI

Simon Carver

For years, job seekers have blamed ATS systems for rejecting their resumes. The story became simple. Use the right keywords. Beat the algorithm. Optimize your resume. But what if the real problem isn't your resume? What if the real problem is that most companies are using the exact same algorithms to decide who gets seen, who gets interviewed, and who never gets a chance? A new Stanford-led study examined more than four million job applications across 156 employers and discovered something much larger than resume filtering. It found evidence of what researchers call algorithmic monoculture. Today we are pulling back the green curtain.

Lachlan Reed

And what we found isn't a story about technology, mate. It's a story about concentration of decision making. Because when the same AI model evaluates millions of people across hundreds of employers, a single flaw can become an economy-wide flaw. G'day everyone, and welcome to the show! I'm Lachlan Reed, coming to you from my backyard shed in Sydney where I've been tinkering with a leaky carburetor on my old Honda trail bike, and joining me is my co-host, Simon Carver. This is The Human Workforce!

Simon Carver

[laughs] Good to be here, Lachlan. Hopefully your bike carburetor is a bit easier to tune than these modern hiring algorithms. Because let's be honest, most people think ATS software is some giant robot reading resumes and instantly rejecting candidates. But if we look at the history, that wasn't the goal at all.

Lachlan Reed

Spot on! In reality, most ATS systems were originally built as databases. Just applicant tracking, workflow management, and recruiter productivity. The ATS itself was never the villain. It was just a digital filing cabinet. But what changed is the introduction of AI screening layers. Now, we've got these predictive models sitting on top of the cabinet, deciding which folders even get opened. [short pause]

Simon Carver

Right, we went from simple databases to active, automated gatekeepers. And that's where the Stanford researchers step in with one of the largest real-world hiring datasets ever studied. We are talking about over four million applications from three million unique applicants, all within one common hiring technology ecosystem used across 156 different employers. [matter-of-fact]

Lachlan Reed

Four million! [whistles] That is a massive chunk of the labor market. Even a kangaroo could trip over data that big. And the phrase the researchers used to describe what's happening there is "algorithmic monoculture." In cybersecurity, we know monocultures are a massive risk. If everyone runs the same operating system, one single vulnerability affects absolutely everyone.

Simon Carver

Exactly. If one company uses a flawed model, that's a company problem. But if hundreds of employers use the same model, that becomes a labor market problem. A candidate may not just be rejected by one company; they may be systematically locked out of an entire segment of the labor market because the exact same algorithm is making the exact same judgment everywhere.

Lachlan Reed

And this is where we need to look at what the Stanford study actually found, rather than the social media hype. The study didn't prove that AI automatically rejects everyone, or that ATS systems are inherently evil. But the numbers they uncovered on systemic bias are incredibly sobering. [reflective]

Simon Carver

They really are. The researchers reported that 26% of Black applicants and 15% of Asian applicants applied to positions where the AI screening system showed statistically significant discriminatory outcomes against those groups. Twenty-six percent of Black applicants. That is more than a quarter of that candidate pool facing an algorithmic headwind before a human recruiter even blinks.

Lachlan Reed

[reflective] That is a massive stat, Simon. And the thing is, those employers probably didn't intend to discriminate. But like my granddad used to say, if you build a crooked fence, it doesn't matter if you meant to make it straight -- the cows are still going to get out. Intent isn't the issue here. The outcome is.

Simon Carver

[warmly] That's a great analogy, Lachlan. And before we dive deeper into how leaders can fix this crooked fence, let's take a quick second to pause. For our listeners out there, I'm Simon Carver, here with Lachlan Reed, and you're listening to The Human Workforce. Our mission is to help leaders, executives, and everyday workers navigate the rise of AI without losing our humanity. If you're finding this episode valuable, please take a moment to like, share, and subscribe to our channel. It really helps us reach more leaders who want to keep the human in the workforce. Now, Lachlan, how do we start holding these systems accountable?

Lachlan Reed

Well, it starts with the bosses asking the tough questions. If you're a board member, a CEO, or a CHRO, you can't just buy a shiny new AI tool, hand it over to HR, and wash your hands of it. You've got to ask: Which AI systems are actually participating in our hiring decisions? How are they audited, and how often? Can recruiters override their recommendations, or are they just rubber-stamping whatever the software tells them?

Simon Carver

Yes! Can we actually explain why one candidate advanced and another did not? Because accountability doesn't disappear just because software made the recommendation. This isn't just about recruitment anymore. We are seeing organizations rely on AI for performance management, promotion decisions, and workforce planning. We are scaling efficiency, but we are also scaling our mistakes. [serious]

Lachlan Reed

Spot on. The lesson here isn't that we should bin the technology altogether. AI is a brilliant tool for summarizing and organizing. But when organizations stop questioning the tool, they create a massive governance problem. The future risk isn't just biased software, Simon. The future risk is identical software.

Simon Carver

[thoughtfully] That is the ultimate takeaway. The organizations that thrive in this next era won't be the ones that automate everything. They'll be the ones that know exactly where human judgment still belongs. Thank you all so much for listening to The Human Workforce. Don't forget to hit that subscribe button, share this episode with a colleague, and join us next time as we keep pulling back the curtain.

Lachlan Reed

Take care, everyone. Watch out for those crooked fences, and we'll catch you in the next one!