A theory is a prediction-generating machine.
I have not even finished reading his post yet. I just thought I'd trot over here and quote him, because this is most succinct confirmation.
An example would be evolutionary theory applied to diet. People have cognitive difficulties with this because they think there is one paleolithic diet, and there is some clear record somewhere. No, what there is evolutionary theory, some anthropological evidence (but not enough to reconstruct an entire diet), and the ability to generate hypotheses based on this and other data.
Then you test the hypothesis.
Mostly, it seems like modern people use theories as creeds, or signalling devices. Believe in climate change? An incoherent question unless is it some sort of test to see whether or not you are in the club. Climate change is observable, but the theory of carbon dioxide causing global warming hasn't been very good on predictions. You get better predictions tracking sun activity. I don't know what this 'believe or not believe' is for, but it isn't of much use for science.
Maybe I shouldn't even use the word science, since that has been used to beat people about the head with. I want to pursue knowledge, and I see all these things getting in the way.
Anyway, ESR didn't just put up a good definition of theory in that post (which I did finish reading). He also came up with a new word:
A “cryptotheory” is a prediction generator of which we are normally unaware because it is not learned but rather wired into our brains and nervous systems.