It's important to get a perspective on "theory" and "model" in science, if only because it's pivotal to understanding Kuhn's work and how he reconciled it with

It's important to get a perspective on "theory" and "model" in science, if only because it's pivotal to understanding Kuhn's work and how he reconciled it with the more-famous account due to Popper that science works by falsification rather than confirmation. The point is that a good theory, like quantum mechanics or Newtonian mechanics or atomic theory or heliocentrism, is Turing-complete -- at least for its domain. These are explanatory frameworks which offer a systematic perspective on setting up models of physical phenomena. Kuhn wants to say, "Yes, scientists test and throw away, like Popper says. But usually, they test and throw away at the level of this model. They don't usually throw away the theory, they just say, 'oh, this effect is non-negligible and should be included in a more detailed model.'" Because a theory can compute anything which any other theory (in the same problem domain) can compute, there is no experiment which distinguishes the two theories. A good example is geocentrism versus heliocentrism: Newtonian mechanics fully allows you to transform to a coordinate system which places the Earth at the center of the universe with the Sun orbiting around it, by adding centrifugal and Coriolis forces. That's perfectly well allowed! You can accurately model all of the things which heliocentrism models with the Earth at the center of the solar system. How do we choose between these? Kuhn's answers are murky, having to do with "aesthetic criteria", and I prefer an account due to Imre Lakatos. This is a feedback theory, but since one prototypical example of a feedback theory is evolution by natural selection, you might think of it instead as theory evolution or so. What happens is, grad students are lazy and want to publish interesting results, so they select whichever theory makes it easiest for them to publish something unexpected. The Standard Model of Particle Physics is ad-hoc as all hell, and it turns out that we know that it's wrong, and it has oodles of parameters that could be tweaked to better conform with any experiment we wanted, including ways to tweak it so that there is a Higgs particle or ways to tweak it so that there is not a Higgs particle. It more or less violates all of Kuhn's aesthetic criteria. Why does anybody suffer such a thing to live? Because we can do insanely powerful calculations of real things with it, and those calculations come out correct to very high accuracy without too-complicated modelling.