It took me until I read The Things They Carried for the third time until I realized that it contained something very valuable to rationalists. In "The Logical Fallacy of Generalization from Fictional Evidence," EY explains how using fiction as evidence is bad not only because it's deliberately wrong in particular ways to make it more interesting, but more importantly because it does not provide a probabilistic model of what happened, and gives at best a bit or two of evidence that looks like a hundred or more bits of evidence.

Some background: The Things They Carried is a book by Tim O'Brien that reads as an autobiography where he recollects various stories from being a story in the Vietnam War. However, O'Brien often repeats himself, writing the same story over again, but with details or entire events that change. It is actually a fictional autobiography; O'Brien was in the Vietnam War, but all the stories are fictional.

In The Things They Carried, Tim O'Brien not only explains how generalization from fictional evidence is bad, but also has his own solution to the problem that actually works, i.e. gives the reader a useful probabilistic model of what happened in such a way that actually interests the reader. He does this by telling his stories many times, changing significant things about them. Literally; he contradicts himself, writing out the same story but with things changed. The best illustration of the principle in the book is the chapter "How to Tell a True War Story," found here (PDF warning, and bad typesetting warning).

A reader is not inclined to read a list of probabilities, but they are inclined to read a bunch of short stories. He talks about this practice a lot in the book itself, writing, "All you can do is tell it one more time, patiently, adding and subtracting, making up a few things to get at the real truth. … You can tell a true war story if you just keep on telling it." He always says war story, but the principle generalizes. At one point, he has a character represent the forces that act on conventional writing, telling a storyteller that he cannot say that he doesn't know what happened, and that he cannot insert any analysis.

O'Brien also writes about a lot of other things I don't want to mention more than briefly here, such as the specific ways in which the model that conventional war stories give of war is wrong, and specific ways in which the audience misinterprets stories. I recommend the book very much, especially if you think writing "tell multiple short stories" fiction is a great idea and want to do it.

I apologize if this post has been made before.

EDIT: Tried to clarify the idea better. I added an example with an excerpt.

EDIT 2: Added a better excerpt.

EDIT 3: Added a paragraph about background.