29 August 2009

Thomas Samuel Kuhn

I haven't found the entire article online yet, but much of it is available through a Google Books preview. Skimming through the preview, it looks like we only have the first part in our book anyway. The article is called, "Logic of Discovery, or Psychology of Research?" It argues that Popper's view of science, while it has its place, is not how science is actually practiced on a day-to-day basis.

I haven't read enough of Popper or Kuhn to be sure that Kuhn isn't demolishing a straw-man, or even just dissembling about the meaning of "falsification" and "test." Kuhn argues that science is more about "puzzle solving" than it is about "problem solving." His main point seems to be that in the normal run of things scientists simply assume the currently accepted theory and work within it. At least in this excerpt, I get the feeling he's missing a crucial point.

Yes, in any run-of-the-mill lab, a groundwork of theory is assumed. The puzzle-solving involves figuring out how that theory applies to whatever work the lab does, whether the work is growing cells, or running current through vases, or whatever. But Kuhn seems to miss that the run-of-the-mill work will be a test of the deduced application. If it turns out to predict different results than those observed, there are three possible culprits: (1) Bad data; (2) Bad interpretation; (3) Bad theory. In the majority of cases, the first two are the most likely. Assuming that instrumentation problems can be sorted out, and enough trials are run to make chance systematic errors extremely unlikely, (1) can be dealt with. The next level to check is (2). How can the interpretation be modified to fit what has been observed? All of this fits in the realm of Kuhn's "puzzle-solving", but fits equally well with Popper's "falsification." Level (3) will only be engaged after levels (1) and (2) have been exhausted, but it takes more than one research group to really effect a change on level (3). Multiple groups have to have similar problems before anyone credible will seriously consider mucking with level (3).

My biggest issue with Kuhn so far is that I don't see how you can possibly have "puzzle-solving" without of necessity having "falsification." How do you know when the puzzle is solved, otherwise? Part of the puzzle solving is figuring out what the accepted theory says should happen in a situation, and that automatically creates a way to falsify the interpretation. I will grant that it takes a lot of broken puzzles before an accepted theory is itself called into question, but that's almost the point. If all it took was one "problem," then every time a machine malfunctioned we'd have a paradigm shift.

Some of Kuhn's examples also make me question his reasoning. He writes that "many theories, for example the Ptolemaic, were replaced before they had in fact been tested." Now, the Ptolemaic theory kept having to add in more and more epicycles to account for the observed astronomical data, and that, to me, would be a sort of test. The more data came in, the more the Ptolemaic system had to be modified. It was never sufficient as it was to explain every aberration. Also, the ad hoc adding of epicycles smacks of Popper's "irrefutability," since, given any conflicting data, more epicycles could be tacked on to "explain" it.

I'll be curious to find out in class if there's more to Kuhn in stuff that we didn't read. It's possible, and maybe likely, that we got the simplest version of his ideas, and that they don't really make sense until they're fleshed out a bit more.

Further reading on Wikipedia:


John said...

I'm not very familiar with Kuhn. My philosophical education is self-motivated, and I tend to get fed up and drop it when I encounter things that smell like sophistry.

Anyway, as to "(1) Bad data; (2) Bad interpretation; (3) Bad theory"

I would add "Bad experiment" ahead of (1). But otherwise, you are exactly right. Before tossing out a theory, you have to be absolutely
certain that something else didn't cause the failure. And a whole slew of independent researchers have to confirm your result.

Falsification is essential to science. If there is nothing, even in principal, that would falsify an idea, then it isn't scientific. However, repeatability is equally essential. The more independently run tests that support an idea, the stronger it is.

"many theories, for example the Ptolemaic, were replaced before they had in fact been tested."

This is completely wrong. Ptolemaic Theory had been failing test for centuries. Every time an astronomer had to add and epicycle, the theory had been tested. There was no better explanation, though, until Copernicus pointed out that if you let go of geocentricism, a few simple equations give a pretty good approximation. The superior theory quickly replaced the old one. (well, aside from some strong opposition by the Catholic Church)

Qalmlea said...

Yeah, that was pretty much my reaction. I think I lumped "bad experiment" in with "bad interpretation", but I can see listing it separately.

For the moment I'm leaving room to give Kuhn the benefit of the doubt, but the doubt is certainly weighing heavily.