roughly, yes. [...] If your mistakes are independent, then correcting one of them doesn't (much) correct your estimate of how many more mistakes you have to correct. Say you have 21 beliefs with 95% confidence and an argument clarifies a random one of them. You still have 1 expected wrong belief. By independence, we might as well say it's belief #1 that gets clarified. People who were wrong about it end up the same as people who were right about it. Yes, they gained more information, but they were really just unlucky to start with less information. This is exactly the lottery/inheritance model. Yes, your ignorance is a fact, but it's not a fact accessible to you. The argument decreases your estimate of your ignorance by the same amount, regardless of whether you win or lose. If you happen to know how ignorant you are, how many items you're wrong about, then the situation is different, but that's a lot less realistic than independence.