Skip to main content

Command Palette

Search for a command to run...

AI and Human Biases (A Thought Experiment)

From Our Hands: The Biases That Grow Into AI

Updated
4 min read
AI and Human Biases (A Thought Experiment)
S
Breaking down complex systems into core principles. Exploring the intersection of Software Engineering, Data, and AI.

Close your eyes for a moment. Think of a person.. any person...

Now notice what happens next.

I say: a man. Something shifts in your mind.

I say: a woman. It shifts again.

I say: a background, a group, a community... and instantly before any story, before any detail your mind has already begun to draw a picture. A face. A personality. Maybe even a judgment.

Nobody taught you to do this in that moment. It just happened. Fast. Automatic. Almost invisible.

That's not a flaw in you. That's a pattern shaped by everything you've absorbed over a lifetime. The books, the news, the conversations, even the silences.

Now sit with this question:

What if AI does exactly the same thing because it was built from those very same patterns?

Where AI comes from

AI is not some mysterious intelligence born in a lab. It is built from us. Shaped by our hands, trained on our words. It learns from what we write, what we post, what we search for, what we argue about, and what we quietly choose not to say.

Every sentence it has ever read carries traces of how human beings think. The assumptions, the shortcuts, the blind spots - all of it gets absorbed, weighted, and passed forward.

So when AI produces something biased - when it assumes, excludes, or judges unfairly - the honest question isn't "what's wrong with the AI?" The honest question is: whose thinking did it learn that from?

We love AI when it sounds smart. We call it broken when it says something uncomfortable. But those uncomfortable moments may be the most honest ones and they are a mirror of patterns we’ve accepted for so long that we stopped noticing them..

The workaround we keep choosing

Today we try to fix AI bias with rules. We bake instructions into the system: be fair, treat everyone equally. And it works… to a point. The outputs get cleaner. The worst mistakes get caught.

But the underlying data doesn't change. And data is memory. It remembers everything we’ve said and done — including the parts we’d rather not look at too closely.

It’s like telling someone who grew up surrounded by bias to simply “act fair". They might manage it when they’re paying attention. But the old patterns learned over years don’t vanish just because someone told them to behave differently.

We've seen what happens when the patches aren't enough. Hiring systems that quietly rank candidates differently based on names or background.. Algorithms that make assumptions about people before they even know them as individuals. These aren’t random glitches. They’re echoes.

The data didn't invent these patterns. We did. The data just remembered them.

The harder question

We live in a world where we still need laws and workplace training to remind grown adults how to treat each other with basic fairness. Think about that. Left to ourselves, without those guardrails, we don’t always get it right.

And yet here we are expecting machines built from our own messy, unchecked output to somehow do better than us.

If we don’t truly believe, deep down, that every person has equal worth, why would we expect AI to act as if they do?

This isn’t an attack on technology. It’s a question about the source.

What would actually change things

In tech, we talk about the difference between a quick fix and a real solution. A fix treats the symptom. A solution goes after the cause.

Right now, most efforts around AI bias are quick fixes like filtering outputs, adding rules, patching problems. They matter, but they don’t touch the root.

The root is us. The data we keep creating. The assumptions we keep feeding into the system.

If we genuinely started thinking and acting differently toward one another, the data would slowly change. What we write, what we share, what we teach our children would change. And eventually, what AI learns from all of it would change too.

Yes, it's a loop. But it doesn't have to stay a bad one.

What we built, and what it's telling us

The next time AI says something that feels wrong or unfair, pause before you jump in to correct it. Ask instead: Where did it learn this? Whose words shaped it? Whose world is it reflecting?

We handed AI the raw material of human thought - our history, our habits, our blind spots and asked it to make something useful from it. What it shows us is often a record of who we were when we weren’t paying close attention.

The real question was never “does AI have bias?”

The real question is: now that something we built is showing us so clearly what our own hands have shaped… what are we going to do about it?

Practical AI Learning

Part 1 of 8

Bridging the gap between AI theory and real-world implementation. This series explores the end-to-end machine learning lifecycle—covering data preprocessing, supervised and unsupervised learning, neural networks, and model deployment. Each guide focuses on building functional AI solutions using industry-standard frameworks like Scikit-Learn, TensorFlow, and PyTorch.

Up next

Learning in the Age of AI FOMO: Choosing Deliberate Learning Over Endless Comparison

You open your feed for a quick scroll, and within minutes you’re staring at yet another post where someone claims they’ve rebuilt their entire workflow with AI, automated half their job, and now barel

More from this blog

S

Simplifying the complexities of modern software engineering

14 posts

I share clear, practical, and in-depth explanations of Python, Data Engineering, and Artificial Intelligence — turning complex concepts into easy-to-understand knowledge with real code and real-world insights. Whether you're a beginner or an experienced developer, you'll find valuable learnings to level up your skills. 🐍📊🤖