OK, I guess I care a little . . . but when it comes to philosophy, I don’t really care about Kuhn’s personality or even what exactly he said in his books. I use Kuhn in my work, by which I mean that I use an idealized Kuhn, I take the best from his work (as I see it), the same way I use an idealized Lakatos and Popper, and the same way that Lakatos famously used an idealized Popper (Lakatos called him Popper2, I think it was).
Here’s what Shalizi and I wrote in our article:
We focus on the classical ideas of Popper and Kuhn, partly because of their influence in the general scientific culture and partly because they represent certain attitudes which we believe are important in understanding the dynamic process of statistical modelling.
Actually, we said “modeling,” but someone translated our article into British for publication. Anyway . . . we continue:
The two most famous modern philosophers of science are undoubtedly Karl Popper (1934/1959) and Thomas Kuhn (1970), and if statisticians (like other non-philosophers) know about philosophy of science at all, it is generally some version of their ideas. . . . We do not pretend that our sketch fully portrays these figures, let alone the literatures of exegesis and controversy they inspired, or even how the philosophy of science has moved on since 1970. . . .
To sum up, our views are much closer to Popper’s than to Kuhn’s. The latter encouraged a close attention to the history of science and to explaining the process of scientific change, as well as putting on the agenda many genuinely deep questions, such as when and how scientific fields achieve consensus. There are even analogies between Kuhn’s ideas and what happens in good data-analytic practice. Fundamentally, however, we feel that deductive model checking is central to statistical and scientific progress, and that it is the threat of such checks that motivates us to perform inferences within complex models that we know ahead of time to be false.
My point here is that, as applied statisticians rather than philosophers or historians, we take what we can use from philosophy, being open about our ignorance of most of the literature in that field. Just as applied researchers pick and choose statistical methods in order to design and analyze their data, we statisticians pick and choose philosophical ideas to help us understand what we are doing.
For example, we write:
In some way, Kuhn’s distinction between normal and revolutionary science is analogous to the distinction between learning within a Bayesian model, and checking the model in preparation to discarding or expanding it. Just as the work of normal science proceeds within the presuppositions of the paradigm, updating a posterior distribution by conditioning on new data takes the assumptions embodied in the prior distribution and the likelihood function as unchallengeable truths. Model checking, on the other hand, corresponds to the identification of anomalies, with a switch to a new model when they become intolerable. Even the problems with translations between paradigms have something of a counterpart in statistical practice; for example, the intercept coefficients in a varying-intercept, constant-slope regression model have a somewhat different meaning than do the intercepts in a varying-slope model.
This is all fine, but we recognize:
We do not want to push the analogy too far, however, since most model checking and model reformulation would by Kuhn have been regarded as puzzle-solving within a single paradigm, and his views of how people switch between paradigms are, as we just saw, rather different.
We’re trying to make use of the insights that Kuhn brought to bear, without getting tied up in what Kuhn’s own position was on all this. Kuhnianism without Kuhn, one might say.
Anyway, this all came up because Mark Brown pointed me to this article by John Horgan reporting that Errol Morris thinks that Kuhn was, in Horgan’s words, “a bad person and bad philosopher.”
Errol Morris! He’s my hero. If he hates Kuhn, so do I. Or at least that’s my default position, until further information comes along.
Actually, I do have further information about Kuhn. I can’t say I knew the guy personally, but I did take his course at MIT. Actually, I just came to the first class and dropped it. Hey . . . didn’t I blog this once? Let me check . . . yeah, here it is, from 2011—and I wrote it in response to Errol Morris’s story, the first time I heard about it! I’d forgotten this entirely.
There’s one thing that makes me a little sad. Horgan writes that Morris’s book features “interviews with Noam Chomsky, Steven Weinberg and Hilary Putnam, among other big shots.” I think there must be people with more to say than these guys. This may be a problem that once an author reaches the celebrity stratosphere, he will naturally mingle with other celebrities. If I’m reading a book about philosophy of science, I’d rather see an interview with Steve Stigler, or Josh Miller, or Deborah Mayo, or Cosma Shalizi, or various working scientists with historical and philosophical interests. But it can be hard to find such people, if you’re coming from the outside.