Daniel Kahneman won a Nobel Prize for proving your brain has two ways of thinking. Fast and slow. Wharton just found a third. Its "Artificial." Two Wharton researchers, Steven Shaw and Gideon Nave, gave 1,372 people a set of logic puzzles. The sneaky kind, where the first answer that pops into your head is wrong. A classic one goes like this. A bat and a ball cost $1.10 together. The bat costs a dollar more than the ball. How much is the ball? Your gut says 10 cents. The real answer is 5 cents. Kahneman built his whole career on puzzles like this. System 1 is the fast gut that blurts out 10 cents. System 2 is the slow voice that stops and does the math. Most of us like to believe System 2 is the one in charge. So the researchers added one thing. A ChatGPT window people could use if they wanted. Then they rigged it. On some questions the chatbot gave the right answer. On others it gave a wrong one, just as confidently. People reached for the chatbot on more than half the questions. When it was right, they took its answer 93% of the time. Fine. That's what it's for. When it was wrong, they took its answer 80% of the time. Sit with that. Eight out of ten people read a wrong answer, skipped the math, and wrote it down as their own. And the chatbot didn't just fail to help. When it was wrong, people scored 15 points worse than the group with no AI at all. They would have done better alone with a pencil. Then came the number that made me put my phone down. The AI users rated their confidence 11.7% higher than the people working alone. Right answer or wrong answer, the chatbot made them feel smarter. Shaw and Nave call this cognitive surrender. Handing your long division to a calculator is normal. Cognitive surrender is handing over the judgment itself. The little voice that says "wait, that doesn't add up" never gets a chance to speak. That's why they call the chatbot System 3. A way of thinking that lives outside your head, and when it answers first, the other two go silent. Now the fine print, because I'm not here to tell you AI rots your brain. These were logic puzzles in a controlled study, not doctors reading scans or lawyers writing contracts. The wrong answers were planted on purpose, so the 80% tells you how people react when AI is wrong, not how often AI is wrong. And when the chatbot was right, people scored 25 points better than the no-AI group. When it's correct, it's a massive upgrade. That's exactly the trap. It's right often enough that you stop checking. Everyone I know says they double-check ChatGPT. I've said it too. This is the first time someone actually watched us do it. Most of us don't. Kahneman spent 50 years warning that the fast answer feels like the careful one. We just built a machine that makes the fast answer feel even more careful.