> Ask yourself how you'd apply the concept of rational expectations or the efficient market hypothesis to the simple game of chess. Would it be a useful analysis to put two players on opposite sides of the table and say "each agent will maximize their utility and engage in actions that are understandable as rational".

> I mean you could say that but what the fuck would you even mean?

Strange choice of example since "rational" forward looking optimizations rather than dumb heuristics is how we get dumb silicon to play chess in a very effective (and very human-like) manner.

I get the argument that the real world is less constrained and more chaotic than chess and that lots of people suck at playing chess anyway, but rational expectations in most models isn't the assumption that every individual in the economy is always right in their predictions, it's the absence of an assumption about the types of mistake the average human must always make in response to policy changes which was explicitly or implicitly embedded in all the non ratex models. A model of chess which assumes the opponent considers the best possible moves available to them will make fewer bad mistakes than one which assumes they always take the pawn even if the opponent isn't very good and likes taking pawns.

It's perfectly possible to argue against austerity policies with a rational expectations model or without any kind of model at all, it's just better we don't make them based on models with dumber assumptions like assuming nobody will predict that austerity will leave them with less money than before.

Models which are too sensitive to the assumption people get things right are bad, but not as bad as models sensitive to the assumption they'll never adapt to stop getting things wrong in a particular way.