J Thomas wrote:This relates to a question in philosophy of science. You appear to follow something like Kuhn's view. Scientists choose a single paradigm and try to relate everything to that. Some of them do things that tend to support it, while others do experiments looking for things that contradict it. As contradictions arise and get resolved by complications in the theory, eventually the whole thing gets so top-heavy that a new generation of physicists accepts a new theory.
I'm rather critical of Kuhn's revolutionary method as a normative method of science (i.e. as a prescription for how best to do science, vs a description of how scientists historically have done science), but if we take a pragmatist, instrumental view of science, then something like it does stand up very well as normative.
Our theories are essentially compression algorithms for observations. Instead of noting every . single . observable . phenomenon . ever, we note a pattern of of phenomena. We do this because it is much, much easier to work with; consider for analogy working with a vector shape in Illustrator vs pushing around individual pixels in Photoshop, it's way easier to change and transform and otherwise work with the abstract pattern than all the individual points it fleshes out to.
When our theories don't match up with every observation, we need to modify them somehow to keep them accurate and therefore useful. But the whole point of having theories in the first place is to make interacting with reality easier, so of course we're going to modify our theories in the easiest way possible to maintain accuracy. If we can make a small modification to an existing simple theory, that's much easier than coming up with a much more complicated theory, so if they provide the same accuracy, it's better to use the slightly modified simple theory than the complicated but perhaps more elegant theory.
However, when we end up having to make so many modifications to the simple theory that it and all its many exceptions and modifications is more difficult to work with than an inherently more complex theory, then it becomes pragmatically better to use the more complex theory than the simple one with all of its many exceptions. Of course, all this depends on someone coming up with a more complex theory in the first place -- if nobody has proposed a more complex theory to use, the simple theory with tons of exceptions is all we have to work with -- but as the pile of exceptions necessary grows, impetus to come up with a more elegant solution grows. If you pile on enough epicycles, geocentric astronomy is perfectly accurate -- but it's just so much easier to do calculations from a heliocentric model. Even accounting for all the rest of physics, besides just planetary motion, it's possible to construct an observationally equivalent mathematical model which has the Earth as the center of the universe, or even which has the surface of the Earth on the inside of a sphere and the entire rest of the universe contained in that sphere, and the innards of the Earth stretching to infinity outside the sphere. But the extra transformations you have to throw in all over the place makes that kind of model a giant pain to work with, so why bother?
And then even when we have a more complex theory, sometimes when we know we're working within a limited domain where the simpler theory is accurate enough, it still makes sense to fall back on that. Engineers use Newtonian mechanics still all the time every day because there's absolutely no need to do all that relativistic math when you're dealing with something the scale of a car driving over a bridge; once you round the results off to the necessary precision, the Newtonian results are the same as the relativistic ones.
My stand is that it is better to do scientific hypotheses a less restrictive way. When experiment shows that some idea is wrong, discard that idea. And everything that has not yet been discarded, is what's left to work with.
Nobody disputes this. The question is how much of a different idea do we have to move on to -- is the old idea with a few tweaks acceptable, or do we need something completely unlike it?
J Thomas wrote:Pfhorrest said something that nobody followed up on. He said that modern particles are not particles, but some sort of complicated mathematical construction that is utterly unintuitive. And my thought was that if you can say "particle" and it doesn't give you a false intuition but instead you think about the complicated unintuitive math, then it should be completely harmless. It's perfectly OK to call neutrinos particles if you remember that they are not particles. I think it's bad if you let classical ideas about particles influence your thinking about neutrinos, but as long as you avoid that and don't think of them as particles then it's all good.
The thing with that is, it's not that it turned out that something wasn't a particle as we once defined it but was instead something else. Instead, it turned out that everything that we called particles
were other than what we held "particles" to be. In a case like that, rather than saying "x is not a member of set S after all", we say "members of set S do not have property F after all". We didn't learn that photons (and neutrinos, and everything else) aren't particles; we learned that particles, in general, are not like what we thought they were. So we keep calling things the same name, and understand that the thing that that name names is different from what we previously thought it to be.