As someone who has followed Crawford's trajectory for the past 15ish years I have to say his biggest problem may be the opposite. He is so tightly married to the idea that narrative simulations must produce concrete, written, book-like stories that he misses how effective modern simulation games (not just computer games, but also boardgames, RP, consims, "paxsims", etc) can effectively engage systems literacy - even though they do not process "natural-language" textual inputs, through a mathematical ur-model, into "natural-language" textual outputs.

I think the average group of analysts running a matrix game is considerably more "mathematically-minded" in that they are much more interested in what the simulation does than whether it produces a particular form of output.