Software developers are especially well-positioned to understand some of the controversial claims of Kuhn: his 'paradigm shift' would be to us a 'refactor'. Not just adding statements or fixing bugs (accounting for 'anomalies' would be the scientific version), but coming up with new primitive terms through which higher-level program statements are made (conversely, new terms/entities comprising the foundation of a theory).
My recollection when reading Structure of Scientific Revolutions was that toward the very end of the book he gets into some more traditional philosophy in relation to incommensurability, which centers around the technical impossibility of comparing paradigms and drawing metaphysical conclusions from that. But for most of the book he speaks very practically about making comparisons and talks about various measures that can be used to evaluate the efficacy of competing paradigms (e.g. accuracy with which theoretical claims match experimental measures, and number of claims which can be verified experimentally).
The article talks about a frustration of Kuhn's where "... Kuhn felt his critics disagreed not with his actual views but with distortions of his views — with the views of a fictional Kuhn." That has certainly been my experience in coming across discussions of Kuhn after reading Structure of Scientific Revolutions: "huh? Did we read the same book?" Most all of what he says is pretty common sensical, practical, illuminating and informed by history—but it seems that folks aren't interested in that and would prefer to ignore 95% of the book and respond to his concluding philosophical synthesis (which I mostly dismissed as an insignificant somewhat egotistical decoration tacked on to the book).
This bit from the article neatly clarifies the whole issue, IMO:
> The issue here is not the denial of reality, but the denial of an absolutely preferred way of talking about it. Statements can be true or false, but not whole languages. As Kuhn puts it, “The ways of being-in-the-world which a lexicon provides are not candidates for true/false.”
"Refactoring" is part of the current paradigm, so the act of refactoring is not a pardigm shift.
Consider, however, that you are a computer scientist in 2012 that has been working in the field of image recognition using hand written algorthms for 20 years. You are the best in your field, highly respected. Your code has been refactored so many times that you think it is near perfect.
Then along comes a young PhD student named Alex, and presents a completely different way of doing image recognition. Instead of talking about "refactoring", the talks about "retraining" or "rewireing". Instead of talking about "software architecture" he talks about "network topology". While you are "unit testing", he is "regularizing". He is still refactoring and building software architecture, you say, but for him, those are less important.
6 years later, everyone has forgotten about your life's work, and Alex's one paper has been cited 25 000 times and triggered a multi-billion $/year field of study.
I think THIS is the on-going paradigm shift in computer science, and it has just started.
> Consider, however, that you are a computer scientist in 2012 that has been working in the field of image recognition using hand written algorthms for 20 years. You are the best in your field, highly respected. Your code has been refactored so many times that you think it is near perfect.
> Then along comes a young PhD student named Alex, and presents a completely different way of doing image recognition. Instead of talking about "refactoring", the talks about "retraining" or "rewireing". Instead of talking about "software architecture" he talks about "network topology". While you are "unit testing", he is "regularizing". He is still refactoring and building software architecture, you say, but for him, those are less important.
Sounds like an intro to an unusually intellectual and a tad weird porn clip.
> "Refactoring" is part of the current paradigm, so the act of refactoring is not a pardigm shift.
The example you cited to justify that claim is a difference in magnitude, not intrinsic structure. I brought up refactoring as an analogy to highlight the key features of another structure less familiar to software developers.
The point of my example, is that the paradigm shift typically involves new people, with a new outlook, sometimes a different education or skill set, etc. They often are younger (but not always, sometimes they are just visionaries finally proven true, like Hinton).
When there is a paradigm shift, there is typically an "old guard" that is not able to or willing to let go of the old way, and that gets left behind. This is why the saying goes: "Science advances one funeral at a time".
In most cases (at least in my experience), refactoring is done either by the same team of programmers creating the original version, or a new team with a similar outlook to the old team.
This means that there is not really an "old guard" to leave behind.
But maybe you meant some other type of refactoring?
The key word in my previous comment would be 'analogy'. In any analogy, some elements in the analogous domain have no mapping. I was highlighting the isomorphism between code undergoing refactors and theories undergoing paradigm shifts, in terms of linguistic structure; I was not commenting on the social components of either.
> his 'paradigm shift' would be to us a 'refactor'.
Just refactoring existing code is not usually a significant enough change to qualify as a paradigm shift. A paradigm shift is more like a rewrite using a different, uh, paradigm - say, functional instead of object-oriented, or Big Data instead of relational.
I agree with the rest of your comment. I read SoSR in Philosophy of Science and have found it provides a very useful perspective. Although it's fun to contrast it with Feyerabend's "Against Method" to remind oneself of the range of viable perspectives here.
Sure--it would need to be a refactor of the foundations of the code base, and not just splitting a method in two or something. If you look closely at the word 'refactor' though, you'll see it's actually referring explicitly to what happens at the heart of Kuhnian paradigm shifts. Further, the impact of such refactors can certainly be of sufficient magnitude to be identified with a paradigm shift. You can find Stephen Wolfram speaking to this effect in regards to working on Mathematica. My personal experience also fits with the pattern described by Kuhn: because you couldn't think of better primitives initially, you need to introduce excessive complication which functions much like the classic example with circular orbits + epicycles. Then! it occurs to you, some new way of viewing the whole core of your problem; you refactor; ever thing is comparatively smooth as butter.
I think part of the reason that isn't so often brought up in typical software engineering blogs or whatever is that developers are not often in a situation conducive to it playing out that way. The only times I've seen it were when I was working with a very deep codebase, as opposed to wide--and it required me to keep the problem regularly in mind for weeks or months at a time before some insight emerged seemingly spontaneously.
That software development devolves into a debate of epistemology, and often a very political debate at that, is the saddest thing I've had to experience. It's made me interested in both politics and epistemology.
I have a background in both science and philosophy, and so I have thought a lot about Kuhn. Here are my conclusions.
To start, there are two parts to Kuhn's ideas. First there is a historical analysis of the process of an old theory being replaced by a new one. Second Kuhn tries to come up with a metaphysical explanation of what is going on here.
The historical analysis seems sound. I say that because it fits my own understanding, because scientists in general think it is right, and because philosophers of science, most of whom initially rejected it, have over the years come to mostly support it.
Where Kuhn gets into trouble is his metaphysical explanation. He has a sort of dualistic view that leads to a type of anti-realism, though he can't quite admit it.
What he should have done instead, in my view, is to explain his historical findings with a metaphysics from one of the anti-foundationalist philosophies, such as Pragmatism, existential-phenomenology, or the later Wittgenstein. This sort of approach makes possible a realism that fits his historical studies.
> First there is a historical analysis of the process of an old theory being replaced by a new one. Second Kuhn tries to come up with a metaphysical explanation of what is going on here.
Sometimes this is overblown. The canonical case of Relativity replacing Newtonian mechanics doesn't invalidate Newton. Rather, it makes Newton's theory a special case of GR that's still valid for most everyday calculations.
Relativity didn't replace Newton for a specific subset of engineering calculations.
But it absolutely destroyed Newton as a comprehensive scientific world view.
Newtonian philosophical fundamentals such as absolute universal time and absolute relative velocity are absolutely wrong - not just wrong for specific high-energy problems, but fundamentally incorrect conceptually, which makes them useless for the development of non-trivial future insight.
This is Kuhn's main point, and it's a shame so few people seem to understand it. A paradigm shift isn't just a new way of solving problems, it's a completely new conceptual map of reality, made of different entities connected by different relationships.
Exactly - and the same is true with regard to special vs. general relativity. Special relativity is another special case of general relativity for inertial reference frames. And special relativity itself is a necessary consequence of the theory of electrodynamics that Maxwell described holding true in every reference frame. But it no more replaces Maxwell's theory than brings it to its necessary conclusion. And even in the quantum world - which physicists famously can't reconcile with observations of general relativity - Maxwell's equations are true even though we have quantum electrodynamics - which include Maxwell's equations as their foundation.
So the paradigm shifts aren't often as earth-shattering as they seem to outsiders or as Kuhn makes them out to be, because they often subsume and contain the prior theory. The way Morris characterizes Kuhn makes it seem like the different worlds of theory, before- and after- revolution, are orthogonal to each other - or can't even talk to each other at all, because they are using the same terms to describe totally different concepts. But I don't believe that to be true in the slightest.
That said, I must admit that I haven't read Kuhn so I would have to rely on Morris' interpretation of Kuhn - which seems, ahem, somewhat slanted.
Kuhn is very explicit on that the typical behavior is this kind of subsuming rather than outright replacement. He even points out cases where the new paradigm causes us to lose the ability to say certain things we were capable of in the old one.
It's nearly always overblown. A new paradigm usually renders the evidence more probable, giving a more precise and accurate explanation, but it rarely fully contradicts old paradigms, since that would leave it with their replications and evidence to explain away!
But a new paradigm also changes the definitions of terms. Once Newton had distinguished "mass" (resistance to acceleration) from "weight" (on the earth's surface) then the meaning of "weight" has also changed. (Medieval scholars had previously refined "weight" however, which helped Newton.) You can teach this new worldview, but it isn't obvious, and it doesn't consist merely in an addition to previous evidence or propositions.
> Rather, it makes Newton's theory a special case of GR ...
Newtonian mechanics (and gravity) are a special-case approximation of relativistic theories. To me that's no biggie, but to some philosophers I've spoken to take the attitude that "approximately right is still wrong".
That sounds silly to me, but they have a point which might made better by emphasising the special-case nature of the approximation. Within Newtonian mechanics, there is no hint that it is a special case of something, it's validity seems to go on for ever. And it is that seeming which is not even approximately correct.
>Newtonian mechanics (and gravity) are a special-case approximation of relativistic theories. To me that's no biggie, but to some philosophers I've spoken to take the attitude that "approximately right is still wrong".
Well, those philosophers should stop expecting a world measured in terms of continuous variables to have a discrete causal topology.
As a working scientist a big problem with Kuhn is that he got a lot of people thinking that it is "boring" to do "normal science" and that the exciting thing is to overthrow the "current paradigm". So you get people with very minor discoveries claiming that they've created a new paradigm because they've found some discrepancy with current theory. It really is unhelpful and leads to misleading magazine covers like "Was Darwin/Einstein wrong?" when new discoveries in biology or physics get made.
Isn't Kuhn rather about false/wrong not being applicable when comparing paradigms, and the controversy about Morris taking this a bit too general?
Edit: On a semiotic level, we may describe Kuhn's approach as being about homonymous signifiers, which belong to different world-views or frameworks, by which they are also defined, while Kripke refers to the special case of names, which serve as an identifier pointing to an atomic entity, which may be extended or even merge with another one, but essentially remains this particular object, which is in turn generalized by Morris to apply to all classes and types of signified objects (rather reminding of the concept of the "referred" than of the more modern concept of cultural entities). Notably, 'truth' is in the first case more an effect of structure and social consensus, in the latter case a matter of discernible properties which may be assessed as facts.
Disclaimer: I'm not a historian nor a philosopher, though I do wonder (I like to think, at least) deeply about these things. I've also only just read the first two (three?) chapters so far, so this may be way to soon to opine. Still, he doesn't strike me as a "surprise twist" type of writer, so hopefully I'm doing the basic ideas justice.
Anyway... I was actually really surprised at how ad-hoc the go-to example of "science-without-paradigm" seems to be, namely the study of electricity and the scientists of that era. All he seems to be doing is an absurd amount of cherry-picking and handwaving as to why these people were somehow "without a paradigm" whereas modern physics (or whatever) is somehow 'more paradigmatic'. Yes, there were a lot of weird ideas (fluids, etc.) about in those days for what electricity was, but that's just because the set of observations was very limited, thus expanding the space of hypotheses beyond what we consider reason these days[0]. I honestly don't understand how we can elevate that to a more 'noble' or 'pure'[1] way of doing science or reason.
I might be a little to drunk to comment, but I'll do it anyway. Sorry.
[0] ... but the thing is that we have a reason these days to consider many of those ideas crazy, namely observation that contradicts them.
[1] This is what all of the book so far seems to be doing to me: "Floundering about without a set goal is much better!". Nevermind the fact that you always have to form a hypothesis before you can even think of anything to observe to confirm/deny it.
EDIT: Just to say: Maybe I've been lied to in my general education on the ancients, but I'm really confused as to why we can't consider the Greeks to have had a "paradigm" (geometry, "perfection" in some sense) or the Egyptians who constructed the pyramids (I hear they were big on astronomy and astrology). Maybe I should just read further into the book.
"there were a lot of weird ideas (fluids, etc.) about in those days for what electricity was, but that's just because the set of observations was very limited, thus expanding the space of hypotheses beyond what we consider reason these days"
That's pretty much what Kuhn means by pre-paradigm science: when scientists don't know enough about an area to have come up with general principles to guide their research. And Kuhn certainly doesn't think pre-paradigm science is better than science within a paradigm. He's primarily interested in trying to describe how science operates, rather than saying I way of doing science is better than another, but he's pretty clear that most of what scientists accomplish happens when they are working within paradigms.
> That's pretty much what Kuhn means by pre-paradigm science: when scientists don't know enough about an area to have come up with general principles to guide their research.
Isn't this extremely post-hoc reasoning? If you're on the bleeding edge, you'll formulate theories based on whatever context you have -- and experiment with that. I don't see how the idea of a "paradigm" helps here. It's what everybody does all the time.
Maybe I'm just confused about the definition of "paradigm".
... which Kuhn does define, but it's a definition just fuzzy enough that it's pretty easy to argue either way on any given area of science.
EDIT: Basically, I think more of a fan of the David Deutsch school of knowledge.
>Maybe I'm just confused about the definition of "paradigm".
Well, as the article points out, Kuhn himself never fully defined it. However, I think we can get a productive account out of reading it as "overhypothesis"[1].
Well worth the long read. Astonished that the author is a "mere" grad student. Keep an eye on this one.
Nice to see a sympathetic account of Kuhn from a scientist. I've always felt that he wasn't worthy of admission to the Sokal Rogue's Gallery. (I sometimes wonder if I've been too quick to judge some of the other big names in that gallery)
Almost surely an allusion to Sokal's eponymous hoax: He published a hilarious gibberish paper in "Social Text", with the glorious title Transgressing the Boundaries: Towards a Transformative Hermeneutics of Quantum Gravity.
Later, Sokal and Bricmont published a book, Intellectual Impostures (Fashionable Nonsense in the US) about postmodern "intellectuals" abusing scientific terminology (eg Jacques Lacan, Julia Kristeva, Paul Virilio, Gilles Deleuze, Félix Guattari, Luce Irigaray, Bruno Latour, and Jean Baudrillard); that would be the rogue gallery presumably.
It's an entertaining book, and I think those writers are guilty as charged.
The problem with Kuhn is that his central thesis of incommensurability as it is commonly interpreted is false. Classical mechanics arises as a specific limit of quantum mechanics. Galilean invariance arises as a specific limit of special relativity. Ray optics arises as a specific limit of wave optics. Newtonian gravity arises as a specific limit of general relativity. In fact, many of these theories were explicitly developed with the constraint that they reproduce the old theory in a certain limit, see e.g. Bohrs correspondence principle. Kuhn knows this, but many of his fans don't. That is partly the fault of the biases his fans, but partly the fault of Kuhn for intentionally deemphasising this most critical aspect of scientific revolutions. What's special about scientific revolutions in physics is not the supposed "incommensurability", but the mathematically precise way in which the old theories survive as limiting cases of the new theories. That doesn't happen for other paradigm shifts. The god-is-punishing-us-for-our-sins theory of disease did not survive as a specific limit of the germ theory of disease. That physics is not like that is what's surprising! Kuhn has his story exactly backwards.
If you want to understand those scientific revolutions, don't read Kuhn; get a physics bachelor.
It has been a while since I read Structure, but I don't quite see how his examination of Phlogiston theory vs. the contemporary understanding of oxidation fits in with your dismissal. It seems to me that Phlogiston theory isn't really a special case of oxidation, as currently understood.
In debates about the philosophy of ethics, it's not unusual to see a philosopher's personal character brought up in a debate about their ideas. But this is a debate about epistemology! Morris ties relativism to the idea about there being two (philosophical) Kuhns, then to there being two (temperamental) Kuhns.
Morris must have spent years thinking, "I know my advisor's theory is wrong. I must come up with the next theory to replace it." That's what grad school teaches you to do. (I'm still doing it.)
When the prevailing theory denies the existence of universal truth, you can't attack the theory from within itself. You have to step outside it somehow. Is attacking Kuhn's temperament the plan? Did Morris sit down and say, there's no rational way to attack this theory, so I'm going to make this fallacious argument about Kuhn's character, presented in a compelling movie? Using post-truth tactics to defend objective truth?
Tangential, but the author’s description of Kuhn’s paradigms as describing things incommensurably differently (“claims that were true in the old way of talking might be false in the new way, and vice versa”) reminds me of the way Hofstadter summarises Gödel’s incompleteness thereoms in GEB. That is that no set of axioms can be both mathematically complete and consistent, so we may need multiple theories (paradigms?) to reason effectively and discover all truths.
Morris seems to claim (based on the author’s summary) that necessary a posteriori truths (ala Kripke) provide some sort of anchor across these paradigms that proves newer paradigms are better than older ones because (and I’m a little unsure on his argument) newer theories are still describing the same thing in different detail and we can therefore see that a newer paradigm is objectively richer and more truthful.
However, I wonder as Kuhn did whether this argument is strong enough to handle all scenarios. As Kuhn argued, a concept of Heat is substantially different across paradigms. I would add to the list those qualitative problems like consciousness and qualia that seem, at least at this stage, to require non-scientific paradigms (ala Lyotard) to reason meaningfully about, but in taking on these paradigms we lose the ability, as Kuhn argued, to talk meaningfully about things we could previously.
The essay as a whole has several interesting moments. Though rehabilitations of Kuhn like this remain troubling:
> In Kuhn’s view, reality is out there, but it doesn’t speak our language. It remains forever alien, non-linguistic, regardless of how well we seem to describe its various parts.
"Reality is indescribable" becomes a bit of a liar's paradox. Why should anyone think the sentence is an accurate description of the world, if accurate descriptions of the world are in fact out of reach?
Of course, it's possible this essay is a sometimes insightful and sometimes poor critique of a separate critique, that in turn is sometimes insightful and sometimes poor, and in turn comments on a treatise that wavers between obvious and overgeneralized depictions of scientific discovery.
> "Reality is indescribable" becomes a bit of a liar's paradox.
There is a difference between an imperfect description and a lie. Kuhn does not say that all descriptions of the world are lies, or that "anything goes", although he have sometimes been misunderstood as saying that.
Ah. That's not exactly what I meant by liar's paradox.
I just meant it's a paradox of self reference. The liar's paradox is the most famous, but the set of all sets that do not contain themselves is another.
So it's true the Kuhnian proposition is much closer to "Every claim about reality is (somewhat) inaccurate." But we still have a massive issue.
I don't see the paradox though. The problem with the liars paradox is not the self reference itself, the problem is that the self reference leads to a contradiction. I don't see any contradiction the sentence presenting Kuhn's view.
Thank you to whoever made sure this news.ycombinator article title included "A review of Errol Morris's critique".
It was not super obvious from the article itself online that this article is a book review.
I thought the author was actually himself proposing some specific theory about truth (or science) (or whatever) and was getting really really frustrated wading through all the storytelling trying to find the specific theory that was being advanced.
Finally I realized it was a book review and that extra piece of context helped give the article (which is fun to read) some semantic meaning.
(I guess I take a different and much more naive view of the history of science and related fields, which is that (a) people are pretty smart, (b) people mostly make the best scientific decisions they can with the best info and ideas they have at the time, (c) a useful way to think about the history of science involves trying to reason about what people were thinking at the time and what was normal for them. Having read this book review I genuinely can't tell whether that's a Kuhnian or a Kripkean or a Morrisean Kripkean view or whatever.)
To explain Kuhn to the HN reader, I'd like to make an analogy between an established enterprise, and a disruptive startup.
The startup is disruptive, because it has re-thought a problem, and uses a different basic model to solve it.
This affects the language used to describe the problem and the solution. For example you still 'call' an Uber, but you actually don't call it anymore, you click on your phone screen to order it. So when a cab company and Uber talk about 'calling', they might have trouble agreeing what 'calling' really means (e.g., in Uber it already entails agreeing on a price, with a cab it doesn't). This is how I understand what Kuhn means with 'incommensurability'.
As the article points out too, it is not two completely separate languages. Rather, the shift in the underlying model also shifts the language, so that supporters of one model mean different things than supporters of the other model, when they make a statement.
> To me, these “endless textual revisions and supposed clarifications” sounded suspiciously like thinking.
How do you tell the difference between earnest thinking and flip flopping, "a sudden real or apparent change of policy or opinion by a public official, sometimes while trying to claim that both positions are consistent with each other... during the period prior to or following an election in order to maximize... popularity?"
Could we at least concede that when somebody says something that makes your bullshit detector go off, and then they change what they're saying, your detector isn't broken?
There are two reasons people change their stories. One is that they once believed something, but later discovered they were in error and now say something else they believe to be true. The other is that they don't care about what's actually true, and are just saying whatever they think will engage or impress their listeners. Frankfurt's "On Bullshit" is a readable survey of this topic.
The second kind is far more common. But the first kind is really important. When someone updates their views we should pay attention. Perhaps a member of the rationalist community could prove this using Bayes theorem.
When you revise your software, is it usually because the earlier versions were bullshit? Anyone who isn't constantly revising their ideas either isn't trying or is bullshitting.
Well the postmodernists certainly cite Kuhn in support of their anti-reality drivel, but the real problem with Kuhn is that he was just wrong.
There's a mature science, where knowledge is limited but accurate, where new findings cause a recontextualization but no paradigm shift, and the previous knowledge is preserved within its new context,, and immature science, where people are doing things wrong, reaching conclusions with inadequate evidence, and stuff gets thrown out during a "paradigm shift", i.e. when they start doing mature science.
Examples of mature science are physics with the incorporation of relativity and quantum physics, which preserved Newtonian physics, or biology, where the human genome project revealed that there were 1/3 as many genes in humans as was previously thought, at which the field barely batted an eye and shifted in a heartbeat to looking more at gene regulation. Clearly, these are huge changes in the fields, but they don't rise to the level of a Kuhnian "paradigm shift" since the old knowledge and vocab and understandings were preserved.
What's an immature science? Probably the social sciences, or those areas of other fields where there's reproducibility problems, p-hacking, and other dysfunctions. To the extent that these fields have overarching paradigms, they may suffer a "paradigm shift". But in mature science, there just aren't any paradigm shifts, in the Kuhnian sense, happening, because a mature science has sufficient evidence in hand before generating a "paradigm".
Kuhn was wrong. But "Paradigm Shift" is a flashy phrase and it captured the zeitgeist of the time it was written. At this point, anything that helps tame this pop-philosophy-of-science phenomenon is not unwelcome, especially because, yes, Kuhn's writings are supporting anti-realist postmodernist trash, without being especially misread.
First, note that you are conflating Kuhn's talk of theories with talk of disciplines. The introduction of quantum mechanics enhanced the explanatory power of physics; it also sent alternative theories, which proposed too many factors or more poorly explained the world, to the dustbin.
On that note, can physics be called a "mature science," then? To reconcile relativistic and quantum physics—both of which provide limited but very accurate and useful results—would be a paradigm shift; it would require the revision of large parts of one or both theories to be able to combine them without contradiction.
There are core ideas in each of these theories which are not simply revisable: they are connected with too many others. Peripheral ideas are malleable; the precise number of genes isn't very important if we can still explain protein variety (via alternative splicing and other modifications).
My recollection when reading Structure of Scientific Revolutions was that toward the very end of the book he gets into some more traditional philosophy in relation to incommensurability, which centers around the technical impossibility of comparing paradigms and drawing metaphysical conclusions from that. But for most of the book he speaks very practically about making comparisons and talks about various measures that can be used to evaluate the efficacy of competing paradigms (e.g. accuracy with which theoretical claims match experimental measures, and number of claims which can be verified experimentally).
The article talks about a frustration of Kuhn's where "... Kuhn felt his critics disagreed not with his actual views but with distortions of his views — with the views of a fictional Kuhn." That has certainly been my experience in coming across discussions of Kuhn after reading Structure of Scientific Revolutions: "huh? Did we read the same book?" Most all of what he says is pretty common sensical, practical, illuminating and informed by history—but it seems that folks aren't interested in that and would prefer to ignore 95% of the book and respond to his concluding philosophical synthesis (which I mostly dismissed as an insignificant somewhat egotistical decoration tacked on to the book).
This bit from the article neatly clarifies the whole issue, IMO:
> The issue here is not the denial of reality, but the denial of an absolutely preferred way of talking about it. Statements can be true or false, but not whole languages. As Kuhn puts it, “The ways of being-in-the-world which a lexicon provides are not candidates for true/false.”