Wrong, Then Right: The Knowledge Underneath the Answer
We gave the same AI model the same incident twice. The first time it got the answer wrong; the second time it got it right. We didn't change the model, the prompt or the question, only the knowledge underneath it. Here is what the Study tested, why the answer changed, and what it means for anyone putting AI agents in charge of real decisions.
