Wednesday, February 23, 2011

On The Structure of Scientific Revolutions. Part 2: Kuhn's Epistemology

Here's the second part of my post. These are the points that are more closely connected to the various discussions I had that motivated me to read Thomas Kuhn in the first place. I do want to reiterate that I think that his idea of paradigm- and revolution-based science history is usefully descriptive, and I mostly like it very much. I do take some exceptions here and there, however, and have some disagreements with respect to its universality. Mostly, I'm interested in challenging it against this epistemological paradigm that I've gone and developed in spite of myself.

Kuhn generally equates the current mature scientific paradigm to the stultifying stuff taught in textbooks. I have a few texts written before 1962, but those tend to be either highly specialized (not yet obviated I guess by new ways of looking at things) or else artifacts that I picked up and keep around as souvenirs instead of sources of information. Maybe things were a little different forty years ago. I mean, yes, textbooks serve to indoctrinate people into the current state of knowledge, but no, I don't think that these texts define very well what science is. To some smaller points of his, advanced book-writing isn't really frowned upon, and I also disagree somewhat that science separates itself from the larger community quite so much. There's a reputation of intellectual superiority that I think scientists vainly like to keep, but on the other hand, premier publications such as Nature and Science, really aim for general understanding of highly complicated fields. Or (I just added) go read the Feynman lectures (these clock in a couple years after Kuhn's essay). My introductory college textbooks often talked about past and current controversies, including the paradigms that stuck and the ones that didn't. The story of a gestalt switcheroo that turns a bug into a feature is an enduring favorite. The sort of triumphant narrative of toppling a progression of barriers that made Kuhn bristle? I don't know if I got that one quite as much, and when I did, it was more concerning the early discoveries. (We'll go back to classical waves, in other words, but not to a continuum of angels.) In my observation, the idea of thriving within a heady open-ended scientific crisis period is closer to the idealized self-congratulating story of many "top tier" scientists (as a colleague once liked to say) today. Even here in the dregs of applied science, "innovation" is the name of the game.

I think anyone working in a research field understands that scientific paradigms are articulated almost like a correspondence, a slow-motion argument consisting of innumerable published articles, conferences, and less-formal meetings. Underlying this communication is the normal science that Kuhn describes, but I don't think the subject matter is chosen solely to gratify a bunch of expected hypotheses. The popular sessions at a conference are the ones chasing after the sexy new field and lighting up the current controversies. Scientists, at least certain kinds of scientists, are just plain hungry for anomalies to fight about. They go looking for a crisis. And even for the bigger paradigm busters, there's plenty of room for brilliant kookery (e.g.) out there on the fringes.

Kuhn made a lot of hay about the seminal insights that John Dalton, a meteorologist, brought to the early days of chemical theory. The mode of thinking that he brought from a different discipline gave him tools to look at chemistry problems in a new way. Again, I don't know what it was like in the early 1960s, and maybe it's Kuhn that helped to begin this newer intellectual paradigm, but much like sexy research, digging around for nuggets in other fields is accepted, common, and encouraged these days. "Interdisciplinary" has become a buzzword too.

Kuhn often implies that not all paradigm changes are the same, that there's a gradation in revolutionary goodness. Roentgen did more than just articulate his paradigm when he discovered x-rays, but on the other hand, he was no Copernicus. We can scale down and down too. Every scientific experiment (or thought experiment) has a challenge and a reconciliation built into it. It's meant to test the paradigm and explain the results. Most researchers will be presented with anomalous measurements even in the course of normal science—if everything goes as expected on the first try, then you really are doing common engineering—which they might ignore, fail to notice, or suitably explain within the existing paradigm. Paradigm shift, Kuhn explains, is a consequence of this kind of normal puzzle-solving difficulty, a question of only how important and persistent the anomalies seem to the community. Kuhn also notes that one person's anomaly is another's puzzle problem, depending on what viewpoint they subscribe to. There is no bright line.

Furthermore, while "paradigm" describes the full body of communication, everyone carries around an individualized understanding of it. If other fields (perhaps even the impure ones) are allowed to come in and interact, it can be a source of competing ideas. As we introduce the idea of competing paradigms, subdivided fields, when we don't let the unpopular ones quite fade away, then we might observe that all of these ideas, dead and alive, can always be compiled into a paradigm-of-paradigms that we can never approach from the outside. I don't have much to add to that, except to note that it does give an unpleasant point from which to voice disagreement, and also from which to advocate, when a paradigm can have a broad or a narrow meaning as the discussion demands.

Prior to reading, as well as throughout the text, I imagined the description of scientific paradigms as a meta-construction built around the normal operation of science. Kuhn calls out normal science as the process of hypothesizing and delivering (until the point where this process fails to deliver) expected results. The anomalies he discusses, the ones that are seen as significant enough (and timely enough, and seen by the right eyes) to demand a new way to look at things, he stresses do come out of the normal operation of science. I don't think he means his views to invalidate this established investigatory process (even if they might require us think of it differently).

I tend to think of the scientific method mostly as a flowchart, starting with observation, study or review, followed by hypothesis, tests for agreement, and conclusions based on results of the test, adding and constantly revising the body of work. I've said that I don't think of this dogmatically, and see it mostly as a general guideline. Kuhn mentions that scientists tend to proceed day-to-day without thinking very much about the rules they're following, and this is true in my experience. I agree that research is goals-biased, and certainly test methods, standards of proof, and so forth are informed by (or are) the paradigm. The scientific method might count as a rough approximation of the quotidian work ("hey, let's see if this idea works"), and even if it's an imperfect decision-making hierarchy, it gets reinforced at the higher level in the conventions of scientific reporting (the customary sections of a paper—Introduction, Experiment, Results and Discussion, Conclusions—restate it outright) and also at the level of scientific funding decisions (write a proposal, and get money to see if it works). The scientific method is a beloved part of our current science philosophy paradigm, but much more than that, it is also part of a fundamental literary one. It maps the process of investigation on to a classic story: what is our subject like, what happened to him, and how did he change.

We like to construct narratives around science, just like everything else. If that can be seen as a template for the scientific method, then can the paradigm approach be mapped that way too? Is the articulation of normal science equivalent to a background study? Is investigating the anomaly the test of normal science? Do the conclusions and revisions amount to the delivery of a revolutionary new paradigm (or the reconciliation with an old one)? Well sure, if we are willing to speak broadly enough.

There remains a need to evaluate theory with respect to observations, and when Kuhn discusses the acceptance of new theories, he addresses this in terms of scientific validation. He denies a Popperian sort of straight-up falsification (rightly I think), and also more probabilistic sorts of validation (that is, accepting things more strongly when they agree better; extraordinary claims require extraordinary evidence). We might take on a new paradigm that's popular, or elegant, or simpler, or seems to promise richer articulation, or fits with the other ones, or maybe it's all just arbitrary. Kuhn speculates that what makes it science isn't necessarily the acceptance criteria, but maybe the fact that it's imagined as intellectual progress. I don't agree with that. I think what makes it science instead of something else is that it's evidence-based.

I remember a quote from my freshman physics book, paraphrasing, that in reality, electrons are neither particles nor waves, they're electrons. Kuhn eventually gets to a similar point and cites it as the resolution of a scientific revolution. I don't want to give him this one. I think that the probabilistic way of looking at things, more than Kuhn's, suggests that nature is an independent thing, and that more than one view can be held simultaneously, within some range of validity. Of course, that could be just me thinking like an engineer, bringing in a less-than-pure-science viewpoint of my own, which perhaps has more of a most-workable-understanding-given-the-data sort of culture. I prefer to couch my understanding of science as a series of known assumptions and constraining rules (maybe the same thing as a paradigm as Kuhn means it), under which some theory is known to be useful, sometimes only good-enough useful, and sometimes only preferred because it's consistent with other theory. I don't feel anyone has to take one rigid outlook to the table.

I concede that science revolutions may not always go to the best theory (it's too early to tell when they're busy being all radical), and certainly doesn't result in the best possible one, but to say that it's a competition between existing paradigms doesn't, to me, refute very well a probabilistic validation approach. At a minimum, there's a requirement of descriptiveness that contributes to the appeal of a new paradigm.

The old understanding of the scientific method is also useful for categorizing ideas. It's good to keep in mind that a "hypothesis" is a proposal, while a "theory" is well-understood within its definition and constraints. Mostly, this serves as a helpful tool for dealing with poorly-informed blowhards.

One thing that a probabilistic validation is good for (and which I think a paradigm model deals with less effectively) is to keep down the poorly supported competing theories. It's a continuation of the point, but it deserves a special heading. It's true that all iconoclasts don't fit in within the popular paradigm. On the other hand, just because you are out there taking a chisel to everyone's favorite statues doesn't mean you're a revolutionary. Maybe you're just an asshole. It's good to have some rules of thumb here. You'd better have a damn good argument if you want to show me your perpetuum mobile.

I am most comfortable spotting paradigms outside of science, as well I might be. Politics and economics seem, to me, to be filthy with the things, and far more than with scientific study, they are unburdened by the rigors of empiricism. Why do people come to suddenly believe in Communism, in consumerism, in American party politics, in popular revolution, in abolition? These are more gestalt-style shifts, nudged on, I often like to think, by events as well as the evolution of scientific paradigms, but colored more heavily by the whimsical human imagination. The failure of old networks to address perceived social crises, suddenly perceived broadly enough, precipitates revolutions of a more political (and generally violent) sort. Kuhn touches on this at the end (it may have been his starting point), but if you've ever witnessed a debate between an American liberal and conservative, then you've seen very clearly a failure to accept the other's set of assumptions and evaluations, not to mention a rather questionable concept of progress, in addition to a craptastic analysis of data (usually worse for the person with a threatened advantage). Living in a political climate that I loathe is difficult, especially when the tools I have for analyzing it are also the ones it provides. I give the social dissenters some major props, including, and maybe especially, those who can spot the system and find a way to conscientiously object to it. I think that much of the alternative social paradigms come from literature and art (and science may owe more to these than is usually acknowledged—I liked Kuhn's point that in the Renaissance, there was little distinction between science and art). I love to see when scientific principles are applied in a more honest manner than number-crunching your way to a foregone economic conclusion from dubious assumptions, and it's governed a lot of my reading in the past few years. My minor observation is that a more evidence-based approach would do wonders for the world.

Boring! But it's out of my head.


Penal-Colony said...

Your introdctory comments struck a chord with me: I too have saved earlier textbooks in my field, literature and philosophy, but not as curiosity pieces but because they're generally more competent and literate. In my field, say, literary studies, it's the past that presents the best models of greatness and the highest achievement in the individual genres.

The only paradigm shifts [if they could even be called that] are the ones in the secondary fields of criticism and commentary -- so that a Marxist literary critic might be replaced by a structuralist, a New Critic by a New Historicist, a Psychoanalist by a Relativist and a Darwinian by a Deonstructuralist. And on it goes, the secondary, scholarly literature subscribes to these fashions but the primary texts remain unchanged.

I guess the equivalent of the Scientific Method would be a kind of hermeneutical imperative dictated by any original text: You have certain parameters to work within but you may not [and mustn't] venture beyond those to make claims that are inconsistent with the facts of the text itself, its givens. New Historicists particularly stretch their analyses almost to revisionism, and I mean a revisionism of the text itself.

Maybe censorship does this too: If a reader's only experience of Huckleberry Finn is the new expurgated version, where all non-PC references have been removed, then what can such a reader truly say of Twain's intention to represent the prejudices of his time? Translation too presents analogous difficulties. You start altering the original in decisive ways then you're cuing and skewing the analyses later.

You mention this, of course, and you rightly elaborate your findings into questions about validity and Popperian falsification. The Procrustean urge is always somewhat present, is it not? You have a notion and you bend the evidence to your theory. Your definition of 'evidence' as a criterion for science is problemmatic and undefined: what is evidence?

Your disagreement with Kuhn is spot on: don't mathematicians sometimes approach a very difficult problem from a very different but valid point of view and still arrive at workable solutions? I don't quite get your point about the revolutionary asshole with his chisel, but I'll return to it again.

The application of this model to Politics and Economics shows the organic and progressive nature of the theory. Strait-jacketed paradigms aren't really paradigms at all, nor are they useful. They serve rather as doctrinaire instruments of oppression, dogmas to beat down, which is anathema to knowledge, learning, openness to discovery, evolution, change and progress.

Keifus said...

If I go back to a chemistry book, say, as a reference, it's helpful to retain a vague idea of the organization of a particular book. Easier to find stuff. (Maybe a parallel is keeping around your favorite text editions of literature.) I'm not quite the sort of guy to carefully reread old texts to pick out the nuances of emphasis, although I do like to skim for that sometimes. (Math books don't talk about slide rules anymore. Chemistry books I think used to talk more about exptl detail than they do now.)

Evidence is a difficult nut though. Kuhn points out that standards of evidence are really part of the paradigm (i.e., what we all agree to accept), and, once we get past our eyes and ears, the principles on which your measuring apparatus are assumed to operate are part of it too. All this is true. But I also believe that measurements have got to be more than arbitrary. I mean, I don't think nature behaves arbitrarily (random sometimes, or stochastic, but not arbitrary). For evidence-based understanding, I don't know if consistency of explanation of observations is sufficient. ("God did it" is perfectly consistent after all.) I don't know if prediction ability is enough either. As the conversation goes, science operates these days to challenge predictions, and the problem of induction is lurking somewhere in the depths of that too. I think the scope of the explanation is another important part. We can explain more of what we can see (although it make take different theoretical tools), and use those explanations to see more. An evidence basis seems to need all those things, and probably more.

Thinking along those lines, I don't think all paradigms (unless we invoke a gigantic paradigm-of-paradigms) really deal with those things equally. They're not always an even tradeoff, and some are better, by the above (admittedly problematic) definitions. Some are better in only limited circumstances, the superiority may not be the only deciding factor of a scientific revolution, and we may need more than one, but still, some just aren't very good at predicting, explaining, or expanding. The point I was trying to make was about dissent, skepticism, and the sort of bullshit equivalence we get a lot of around here. Just because you have an alternate view, doesn't necessarily mean it's an equivalent view. A lot of alternate views are lazy, incomplete, and very self-serving.

I like your points about translation and censorship and textuality. One of my favorite Borges stories was about an author writing the complete text Don Quixote in a different paradigm. (You know, now that I have that vocabulary.) It was pretty funny.

Yeah, I read some fraction of papers where a physics problem is solved in using a couple of different theoretical assumptions (such as a big computational solution of Maxwell's equations vs. scattering theory). When they agree, it's a hint you're on to something.

[I hardly deserve fine comments like yours. Thanks.]