Imagine you’re trying to solve a puzzle. You have all the pieces, the rules are clear, and the answer seems obvious. But what if your own brain keeps getting in the way? This is the strange situation that unfolded when people started testing ChatGPT with a simple code called Rot13.

It turns out, even with a tool as powerful as AI, our human minds can play tricks on us. This story is about how a silly text trick became a window into some deep truths about how we understand information and trust what we see, especially when it comes to decisions.

The Rot13 Trick: Simple Code, Big Questions

Rot13 is a very basic way to scramble text. You just shift each letter 13 places forward in the alphabet. A becomes N, B becomes O, and so on. To unscramble it, you do the same thing again. It’s not a secret code, it’s just a simple letter swap.

For example, "hello" becomes "uryyb" with Rot

13. And "uryyb" becomes "hello" if you apply Rot13 again. It’s a reversible process that doesn't change the meaning of the words at all, just their appearance.

People started asking ChatGPT to do things with Rot

13. They would give it text and ask it to apply Rot

14. ChatGPT, being a very capable language model, could do this perfectly. But then things got weird when people tried to make it *think

• about the Rot13 text.

When AI Sees What Isn't There

One common experiment involved asking ChatGPT to explain a piece of Rot13 text. For instance, someone might show ChatGPT the Rot13 version of a sentence and ask, "What does this mean?" If ChatGPT understood the task, it would likely say, "This is Rot

13. The original text is [original sentence]."

But what if the prompt was more complex? What if it asked ChatGPT to *evaluate

• the Rot13 text as if it were normal text, or to make a judgment based on it? This is where the interesting results started appearing.

Some users reported that when they presented ChatGPT with Rot13 text that *looked

• like it contained certain words or concepts, the AI would sometimes respond as if those words or concepts were actually present, even though they were just scrambled letters. It was as if the AI was seeing patterns that weren't truly there in a meaningful way.

Daniel Kahneman's Ideas Come to Life

This phenomenon started to remind people of the work of Daniel Kahneman, a famous psychologist and Nobel Prize winner. Kahneman studied how humans make decisions and the biases that affect our thinking. He talked about two systems of thought: System 1 and System 2.

System 1 is our fast, automatic, and emotional thinking. It’s good at making quick judgments based on gut feelings and patterns. System 2 is our slow, deliberate, and logical thinking. It’s used for complex problems that require concentration and careful analysis.

Kahneman showed that System 1 often jumps to conclusions, and System 2 doesn't always catch its mistakes. This can lead to errors in judgment, even when we think we are being rational.

The

Bias in the Machine

The way ChatGPT sometimes seemed to react to Rot13 text mirrored these human biases. Even though the AI technically processed the scrambled letters, its internal processing might have been influenced by the *potential

• meaning it could derive if the text were unscrambled.

It’s like showing someone a drawing that vaguely looks like a monster. Even if they know it’s just scribbles, their brain might still *feel

• a sense of unease because System 1 recognizes a scary shape. The AI, in some ways, seemed to be exhibiting a similar tendency.

"Our thinking is often shaped by how information is presented, not just what the information actually is."

This is a core idea that Kahneman explored. The Rot13 experiments with ChatGPT showed that even artificial intelligence could, in a way, be tricked by the presentation of information, similar to how humans are.

Testing the Limits: What Does It Mean?

Researchers and curious users tried different ways to test this. They would create Rot13 text that, when unscrambled, formed words that might trigger certain responses. For example, if the Rot13 text, when unscrambled, contained a word that sounds negative or positive, the AI’s response might be colored by that.

This wasn't about the AI *understanding

• the negative or positive meaning in the human sense. Instead, it suggested that the AI's algorithms, trained on vast amounts of human text, might have learned to associate certain letter combinations or patterns (even when scrambled) with specific kinds of outputs.

It raised questions about how AI models learn and how they can reflect the biases present in the data they are trained on. The AI wasn't intentionally biased, but its responses could be influenced by patterns that mimic human biases.

Why This Simple Trick Matters

The Rot13 puzzle with ChatGPT is more than just a fun internet experiment. It highlights a crucial point about artificial intelligence: it is a tool built by humans, trained on human data, and therefore can sometimes mirror human thinking, including our flaws.

It showed that even with a clear, objective transformation like Rot13, our perception and interpretation can be swayed. This is especially important when we rely on AI for information, decision-making, or analysis. We need to remember that the output is a product of complex algorithms and the data they learned from.

The

Illusion of Objectivity

AI is often seen as purely logical and objective, free from human emotion and bias. However, the Rot13 case suggested that this isn't always the full story. The way information is encoded, even with simple methods, can influence the AI's output in ways that seem surprisingly human.

This means we should approach AI-generated content with a critical eye. We must question the results and understand the potential influences behind them, rather than accepting them at face value.

The Takeaway: Thinking About Thinking

So, what’s the big lesson from this Rot13 experiment? It’s a reminder that understanding how information is processed, both by humans and by AI, is incredibly important.

Our brains are wired to find patterns, and sometimes we see patterns that aren't really there or misinterpret what we see. The experiments with ChatGPT and Rot13 showed that AI, in its own way, can also be influenced by these kinds of pattern-matching tendencies.

It pushes us to think more carefully about how we interact with technology and how we interpret the information it gives us. It’s a fascinating look at the intersection of simple code, advanced AI, and the enduring quirks of human (and perhaps artificial) cognition.