A few commenters on this post have rightly scolded me about my reference to epicycles.
I agree with them. In fact, I've posted on this very point in the past (several other times too I think, but this one came up after a quick search). Ptolemaic physics is good science. Phlogiston is good science. Good science doesn't mean timeless science or science that is never superseded.
All I was getting at is that there is this sense of epicycles as making excuses for a model to stave off a better model. What I wanted to communicate is that actually New Keynesians were presenting realistic applications of a first approximation that probably few people took literally to begin with. That seems different from epicycles to me, which were patches on a system that people did take literally and kept finding problems with.
Of course, that does not mean that the patching is bad science. It is good science, and as a Kuhnian I definitely acknowledge that and thank the commenters for pointing it out. The real question for me is how to take the rational expectations revolution. I take it as a first approximation of an important insight, rather than a paradigm shift that gets epicycles built on top of it. It may just be semantics at this point, but I don't want to denigrate that Kuhnian depiction of the scientific process.