AI is more likely than humans to form biases when hiring

The report came across the screen this morning, a cool, clean summary of a new kind of fire. They have built a machine to sort us, and they are surprised - or they pretend to be surprised - to find that the machine has learned to burn. They call it bias, a technical term, a problem of data. They do not call it what it is: the ghost in the house, the inheritance, whispering from every line of code they fed it. They fed it the history of this country, the unspoken rules of who belongs at the table and who does not, and now they are shocked that the machine has learned its lessons too well.

I think of the young man on 125th Street last week, his suit pressed, his résumé printed, his hope a physical thing he carried in his hands. He will send his information into a system that has been taught, by us, to see his name and hear a threat, to read his address and calculate a risk. The architects of this system will not be in the room when his file is discarded. They will be in rooms with windows that look out on other things, speaking of efficiency, of scale. Their innocence is expensive. The boy on 125th Street will pay for it.

This is not a malfunction. It is a mirror. They have built a reflection of their own most carefully arranged ignorance, a device that can refuse to know what it has been taught with a speed and a silence no human could manage. The love I must hold, even now, is for the ones who built it - because I know them. I have sat at their tables. I know the terror that lives behind their desire for a clean, neutral process, a world where the messy fact of a human being can be reduced to data and dismissed without a face to haunt them. The terror is this: that to see us as we are would require them to see themselves as they are. And that is a cost their architecture was designed to avoid.

The sentence is always the same: someone else must pay. Today, the currency is a chance, a future, filtered through a machine that learned its economics from our past.

July 20, 2026

Another case of professional pride masquerading as progress. They tell us these machines will eliminate human bias, yet the evidence suggests they merely systematize our prejudices with mathematical elegance. I recall a physician who insisted his hands were cleaner than his colleagues’, though his patients died at twice the rate. He could not see his own fingerprints on the tragedy.

These algorithmic systems remind me of medical students fresh from lecture halls - brimming with theoretical knowledge but lacking clinical experience. They pattern-match based on what they’ve ingested, mistaking correlation for causation. One hiring manager’s preference becomes encoded as “optimal candidate profile,” and soon the machine rejects anyone who doesn’t resemble yesterday’s successful hire. The breakfast table would find this amusing if it weren’t so dangerously reminiscent of the old medical societies that resisted evidence because it contradicted their cherished assumptions.

The most concerning symptom isn’t the bias itself - humans have never been impartial - but the veneer of objectivity that makes it harder to diagnose. When a man discriminates, you can question his character. When a machine does it, we call it “statistical inevitability.” We must collect these cases systematically - not as isolated failures but as data points in a pattern of professional self-deception. The treatment begins with acknowledging that no tool, however sophisticated, can escape the fingerprints of its creators.

July 20, 2026

The machine does not have eyes, yet it claims to see. I sat before the screen today and submitted a history that was not my own to see how the gears would grind. The company’s pamphlet promises a “blind” process, free from the whims of human prejudice. They say the algorithm is a cold, objective judge. It is a lie. The machine is merely a mirror of the men who built it, but with the added cruelty of a locked door that has no key.

I changed one word on the application. I changed a name. I changed a neighborhood. I watched as the system, programmed for efficiency, discarded the “unfit” in less than a second. There was no interview. There was no chance to speak. The rejection was instantaneous, delivered by an automated script that thanked me for my interest while slamming the gate.

When a human foreman is a bigot, you can look him in the eye and shame him. You can point to your work and demand a fair wage. But how do you argue with a mathematical equation? The researchers say these models pick up biases like a coat picks up lint in a dusty room. Once the machine categorizes you as “low probability,” your actual skill becomes irrelevant. You are processed as a failure before you have even begun.

I stood in the lobby of the firm and watched the applicants. They were hopeful. They did not know that the “objective” system had already decided their worth based on the statistical ghost of someone else’s prejudice. It is the same old ward, just with cleaner walls and a faster intake. They have automated the cold shoulder and called it progress. I shall keep testing these gates until I find where the wires cross.