In writing, I often notice that it's easier to let someone else come up with a bad draft and then improve it - even if "improving" means "rewrite entirely". Seeing a bad draft provides a basic starting point for your thoughts - "what's wrong here, and how could it be done better". Contrast this to the feeling of "there's an infinite amount of ways by which I could try to communicate this, which one of them should be promoted to attention" that a blank paper easily causes if you don't already have a starting point in mind.

You could explain the phenomenon either as a contraining of the search space to a more tractable one, or as one of the ev-psych theories saying we have specialized modules for finding flaws in the arguments of others. Or both.

Over in the other thread, Morendil mentioned that a lot of folks who have difficulty with math problems don't have any good model of what to do and end up essentially just trying stuff out at random. I wonder if such folks could be helped by presenting them with an incorrect attempt to answer a problem, and then asking them to figure out what's wrong with it.