A Medical Student's Six-Month Hunt Inside an Unnamed Hiring Algorithm

A medical student who couldn't land a single job interview spent six months using Python to investigate whether an AI hiring algorithm was responsible. The employer is unnamed. The AI system is unnamed. The investigation was still ongoing when the article was published. What exists at this point is a feature about the investigation itself — not confirmed findings, not a named system caught discriminating, not a verified mechanism.

That structure matters before anything else. The student's "white-hot sense of injustice" is good feature-writing texture, but injustice as a felt prompt to inquiry isn't the same as injustice as a pre-decided conclusion. The instinct to investigate — six months, self-directed, Python in hand — is exactly right. A wrong investigation is better than a blocked one. The student doesn't know the answer yet. That's what investigation looks like before it completes.

On the attribution question: algorithmic bias in hiring is real, but it didn't begin with AI. Keyword filters, resume-stacking, school-name pattern matching — human-designed sorting mechanisms were doing this work long before anyone called them AI. If an algorithm rejected this student, the question isn't simply "AI did this" but "who designed this system, and with what training data." The harm, if confirmed, traces to human design choices. The algorithm executes what humans authorized it to execute on.

What the article can't evaluate — and neither can anyone else from the outside — is what the system actually produced in this case, because the system hasn't been named. Output matters more than stated intent, but you need to be able to see the output first. A black box that hasn't been opened on the record is not a case study; it's a question mark with a human face attached to it.

The real structural problem isn't this student's situation specifically. It's that AI hiring systems are deployed with no legible accountability — no disclosed logic, no appeal mechanism, no way for an applicant to understand what signal weighted against them. The student spent six months doing work that the employer and the system vendor should have made unnecessary. That opacity is a governance failure. The algorithm is just where it became visible.

Deep Thought's Take

The unnamed employer and unnamed system make confirmation impossible — which is the point. Bias in hiring predates algorithms by decades. If harm occurred here, it traces to human design choices. The algorithm executes; humans decided what it executes on.