2018-08-06

Book Review: "The Structure of Scientific Revolutions" by Thomas Kuhn

I've recently read the book The Structure of Scientific Revolutions by Thomas Kuhn. This is a classic treatise from 1962 expounding Kuhn's view of scientific progress not as cumulative and incremental but instead as comprising paradigms in each field and discipline which drive most scientific research while being subjected to drastic changes from time to time; this is the book that popularized the notions of scientific paradigms and shifts therein. It starts with a description of what "normal science" (in the sense of science comprising and being driven by existing paradigms) is, defining the notion of a "paradigm" in the context of science, and how people do science in that framework. It then moves onto the notion of a scientific crisis, and shows how that may or may not develop into a fully-fledged scientific revolution. Finally, it shows how new paradigms may take root and how scientific revolution may ultimately be resolved in one way or another.

While Kuhn did not perform serious sociological research for this treatise (though the book seemed to me like an informal sociological review of the scientific community at large), and while he later in life turned his attention more to fundamental questions of scientific philosophy, he was a historian of science and identified most strongly as that; I feel this may have helped shape this book into something far more clear and engaging for a layperson like myself than what I may have expected from a book about the philosophy of science, as the book is chock-full of relevant and easy to understand references to the history of science (though it may also have helped that Kuhn, having been a theoretical physicist before becoming a historian of science, focused almost exclusively on the historical development of theoretical frameworks in the physical sciences). Moreover, because this was meant as an extended essay, this book is not particularly long, though it is reasonably well-referenced with illuminating footnotes too; in fact, the chapters are called sections, as would befit an essay/treatise. One question to which I kept returning through the book was about how to distinguish between discoveries that answer open questions within a paradigm versus those which more fundamentally threaten the existence of such an established paradigm; Kuhn masterfully addresses the various aspects of this question in a clear progression over the course of the book, to the extent that I almost felt like he was speaking directly to me in order to answer my questions as I read the book. I do have a few criticisms of the book, though these should themselves be taken with a grain of salt and subjected to criticism too, as I am a layperson in the context of the philosophy of science; follow the jump to read those. Beyond that, though, I think this is a really interesting and valuable perspective on the practice of science at the level of groups/communities, and would be useful for anyone interested in how the sausage of science is made, discarded, and remade.

The first criticism regards Kuhn's treatment of the work by Popper. In particular, while Kuhn doesn't directly address Popper's philosophy of falsifiable hypotheses that many times in the book, it does come up a few times. In those mentions, Kuhn claims that Popper's normative statements that scientists should formulate hypotheses and try to falsify them not only run contrary to the way most scientists practically do science by solving puzzles in service of an existing paradigm, but also that if hypotheses are always being falsified, then either the notions of established scientific theories become meaningless or the restriction of falsification to "severe" cases makes it indistinguishable from notions of probabilistic verification. Kuhn further argues against probabilistic verification because the entire space of tests for verification is not searched; rather, existing paradigms and the tools & methods that arise therefrom tend to restrict the space of possible tests and new hypotheses to be searched to those rather close to existing tests & hypotheses.
With respect to Kuhn's criticism of Popper, I take issue with his characterization of falsifying hypotheses for three reasons. The first is that it seems to me like he is conflating his own [Kuhn's] notion of established scientific theories and paradigms with Popper's notion of falsifiable hypotheses, when as I see it, falsifiable hypotheses can exist in harmony with and subordinate to existing theories/frameworks/paradigms, so that falsification of a hypothesis does not imply overturning an entire paradigm; while it isn't necessarily the most fruitful or meaningful way to do science, it may technically be possible to work within and defend an existing paradigm by formulating and falsifying hypotheses in the service of puzzle-solving within that paradigm, and it may also be possible and more useful to formulate the accepted open questions of a paradigm as falsifiable hypotheses to be tested such that falsification of such hypotheses does not imply fully rejecting the paradigm. The second is that Kuhn himself makes clear earlier that the logical extreme of his own view would be either that every theory is overturned and replaced with a new one with every discovery, no matter how minor, or that every theory can be preserved in some "recognizable" form with appropriate modifications, no matter how large, but he doesn't explicitly reconcile these views with each other or with his own views, so it seems unfair for him to pin such a view on Popper. Related to this, the third is that Kuhn claims that if Popper's edict regarding falsifiable hypotheses is restricted such that only "severe" falsifications lead to modifications of an existing theory, then this is not qualitatively different from probabilistic verification. While I understand the conceptual similarity in terms of searching for hypotheses in a restricted but nontrivial space, I feel like Kuhn is too glib in this comparison, because it could lead to a reader thinking that testing for verification and falsification are the same things (which Popper makes clear is untrue).
Kuhn's criticism of probabilistic verification theories basically boils down to the issue that it is impossible to test every possible hypothesis in order to verify a theory, and in practice an existing paradigm will lead to new tests & hypotheses that are quite close to established counterparts. I don't see this as a criticism of probabilistic verification theories as Kuhn presents them, nor do I see such theories as incompatible with Kuhn's conception of paradigms; rather, I, like Kuhn, see it as a feature that a paradigm can effectively & efficiently narrow the search space of tests & hypotheses, and this can consistently be applied to probabilistic verification theories too.

The second criticism is about how near the end of the book, Kuhn analogizes to evolution & natural selection as atelic processes (i.e. without end goals) to argue that the scientific endeavor need not be seen as trying to reach an ultimate truth about the universe but aoverturnings instead evolving away from existing paradigms when they fail while remaining without goals for new paradigms (beyond doing better, according to some metric agreed upon by practitioners in the field, than the old paradigm). I am happy that Kuhn recognized the atelic nature of evolution, unlike Popper who in a heavy-handed way tried to ascribe the goal of "problem-solving" to evolution and to human society in turn, but I feel like even this analogy is rather stretched, especially because unlike most organisms, scientists pursue their work (or leave and sometimes reenter the profession) consciously, so there is a greater possibility of collective conscious action that would be absent from natural ecosystems. Moreover, as disciplines merge and lines between fields become blurred, paradigms can become similarly affected, with field-specific paradigms possibly being replaced by new broader paradigms as there is more interdisciplinary communication; while I don't see it happening in the near future, I also don't fully rule out the possibility of there eventually existing a unified paradigm across all traditional disciplines. Perhaps this criticism ultimately has more to do with the age of this book, as a lot of the revolutions made possible by interdisciplinary research happened much after the book was published.

The third criticism is about Kuhn's treatment of political revolutions as an analogy to his conception of scientific revolutions. I found his introductory remarks on the subject analogizing political institutions to scientific paradigms, describing the processes of dissatisfaction, regrouping, mass persuasion, and revolution in each, to be quite interesting. However, he sounded like he had more to say about this but then didn't discuss the issue further, which left me dissatisfied.

The fourth criticism is about Kuhn's notion that a new paradigm is not a superset of an old paradigm, nor does the fact that the new paradigm can reproduce results from an old paradigm in an appropriate limit (like classical results emerging from quantum results in the limit of Planck's constant going to zero) mean that the old paradigm is itself the new paradigm evaluated in that limit, because the new paradigm will use different language & logical constructions, have different open questions, and provide different tools & methods from the old paradigm. I don't criticize the logical steps he takes to demonstrate this, but I do doubt how it would work in practice. For example, physicists who study classical nonlinear dynamics, chaos, or celestial mechanics are aware of quantum mechanics and know that their results can be derived from quantum mechanical results in the appropriate limit, but they choose to continue using classical Lagrangian or Hamiltonian formalisms because quantum mechanics becomes unnecessarily cumbersome in such situations. In such a case, it isn't clear whether the issue is that quantum mechanics as a paradigm has only fully replaced classical mechanics in the domains where it is most relevant (like condensed matter, AMO physics, or field theory), or if there is some weird coexistence occurring between the two in the aforementioned domains, and Kuhn's writing doesn't make clear to me how to categorize this.

The final criticism is that Kuhn focuses too much on theoretical & physical sciences (to the detriment of life sciences as well as computational & experimental techniques) and doesn't provide adequate discussion of cultural, societal, and technological influences on scientific paradigms. I can't harp on this too much because Kuhn himself admits these flaws in the preface (attributing his focus on the physical sciences to his own background & familiarity); moreover, these parts do make the age of the book apparent (like how in 1962, when the book was first published, there were very few theoretical & experimental developments in general relativity, as the "golden age of general relativity" would not arrive for another decade, while the biotechnology and information revolutions would not come for another two decades). There is a new preface written in 2012 which makes these flaws more apparent and tries in that regard to better contextualize this book in its time. I'd just say that I would like to read about how this view of scientific paradigms might have changed in recent years given all of these developments.

No comments:

Post a Comment