One way to test scientific realism

One way of formulating Scientific Realism is as follows:
What our successful scientific theories say about unobservable entities and processes is approximately true.
This is not the only way to formulate scientific realism, but it is one of the more common ones, and it does effectively separate realism from versions of anti-realism which hold that we are not justified in believing what our theories say about unboservables.

Obviously, this version of Scientific Realism cannot be directly tested using our current theories and current technology, since what is currently unobservable can't be observed now.

However, what is observable shifts over time (at least in one important sense of the word 'observable'). This can happen either because (1) we develop the ability to reach new regimes of old variables (e.g. scientists create technology to make materials colder or hotter than we previously could, or we can study bodies moving at higher and higher velocities), or because (2) scientists develop new instruments that enable new types of observation reports (e.g. telescopes, microscopes, fMRI machines, or mass spectrometers).

This suggests a way to test realism diachronically, using the historical record. First, find something that went from being unobservable to being observable. Then find theories that were (considered) genuinely successful at that earlier time, and see what claims it made about the previously-unobservable-but-now-observable world. Finally, check those claims against the now-observable reality.

Scientific Realism (at least the version stated above) predicts that the old claims about the previously-unobservable things will usually approximately match the new observations of those things. (I say 'usually' instead of 'always,' because sensible realists are fallibilists.)

I have not run this test myself. To do it in an intellectually responsible way, a large survey of past transitions from unobservable-to-observable would have to be collected, and steps would have to be taken to make that sample of transitions representative. However, at first glance, it looks like at least some cherry-picked famous examples don't bode well for the realist's prediction:

• The telescope played a significant role in the scientific revolution
• The vacuum pump played an significant role in the scientific revolution
• The ability to cool things down further and further led to the discovery of superconductivity
• The ability to study bodies at higher and higher speeds was crucial in the transition from classical mechanics to special relativity

There are historical examples that run in the realist's favor too; I think one good example is that (on the whole, i.e. usually) phylogenetic trees generated via molecular data matched previously existing phylogenetic trees fairly closely (i.e. the old trees were usually 'approximately true,' which is all the realist wants). This is why, as I said, we need a large survey to figure out which historical transitions reflect the overall, general pattern, and which cases are outliers.
{ADDED LATER (May 2022): Simon Allzen's "From Unobservable to Observable: Scientific Realism and the Discovery of Radium" is another nice, detailed example that's intended as an example in the realist's favor. Here's a representative quotation: "an entity considered to be unobservable can be inferred at one stage in the process by virtue of its role as indispensable for predictive success [i.e. via IBE -- GF-A], only to change into an observable at a later stage, thus confirming the reliability of the inference. As a case study of the conceptual changes of entities I use the discovery of radium."}

Finally, in terms of already-existing arguments, this is not really very different from the Pessimistic Induction (if at all). I think of it as a specialized version of that argument, focusing on the realist's claim that the observable/ unobservable boundary does not mark an epistemically important distinction. For this reason, I think of the above as a diachronic version of Kitcher's "Real Realism" (which potentially comes to the opposite conclusion of Kitcher's view).

Indeterminism in developmental biology

There's a review article in this week's Science (v.320, April 4 2008, 65-68) that is potentially of philosophical interest, "Stochasticity and Cell Fate". The bumper sticker version: although a cell's transformation into a specialized subtype is deterministic in most cases, "[i]n some cases, however, and in organisms ranging from bacteria to humans, cells choose one or another pathway of differentiation stochastically, without apparent regard to environment or history."

Discussions of indeterminism in biology have usually been restricted to the 'random' mutations that drive evolutionary change. This, if it holds up, looks to be a quite different kind. And interestingly, the authors point out reasons why a certain degree of indeterminism may confer selective advantage upon organisms whose development contains stochastic elements.

Can a widespread local realist be a global anti-realist?--and the preface paradox

I have been thinking about whether there might be something like the preface paradox in the scientific realism debates. There is now a distinction being drawn between local (or 'retail') realism, in which one argues for the truth of particular scientific theories (e.g. quantum mechanics) or the existence of particular scientific entities (e.g. quarks), on the one hand, and global (or 'wholesale') realism, in which one argues for the approximate truth (or referential success) of mature scientific theories in general.

What I'm wondering is whether it can be justified and/or rational to be an everywhere local realist (so QM is approximately true, and general relativity is approximately true, and population genetics is (approximately) true, etc.), but still be a global anti-realist -- say, because you place a lot of weight on the pessimistic induction on the history of science. Or, on the other hand, whether everywhere local realism really pushes us towards global realism.

I'm currently guessing that one CAN be an everywhere local realist without being a global one, for the following two reasons.
(1) One standard response to the preface paradox seems perhaps even more applicable here than in the preface case: while the author assigns a high probability to each individual assertion in her book, the probability of (p & q & r & ...) will be low.
(2) Also, although if A is true and B is true, then 'A and B' must be true, it seems to me that even if A is approximately true and B is approximately true, then 'A and B' need not be approximately true, for A and B could be contradictory (for example, the prima facie conflict between quantum mechanics and general relativity).

I'm happy to hear any reasons for the opposite view, viz. that widespread local realism pushes us towards global realism.

Is arithmetic empirical?

One of the questions I've been wanting to think about (in part because of my interest in the Quine-Carnap relationship) but haven't really got around to yet is: Is there any important sense in which arithmetic is empirical? I know there is some good literature on the subject, but I've thus far only perused it without really digging into it.

For me, one consideration that makes me think it might not be crazy to think of arithmetic as empirical is what happened with geometry and general relativity. If Einstein can show that the space in which we live is non-Euclidean, isn't it at least imaginable that some future scientist will show us that the 'true' arithmetic of our physical world is non-classical (which I suppose means: it does not obey the Peano axioms). [There could still be a mathematical structure that obeys classical arithmetic, just as Euclidean space is still a mathematical object that obeys all five of Euclid's axioms.]

However, I've always had a hard time imagining what possible observation could cast doubt on classical arithmetic. In last week's Science, there's a report that at least might merit consideration as a candidate. Researchers found that if you add one photon to a light beam and then take one away, you observe a different end-state than if you reverse the order of operations, i.e., first remove one and then add one. In other words, x + 1 - 1 does not equal x - 1 + 1. Even stranger, the authors find that "under certain conditions, the removal of a photon from a light field can lead to an increase in the mean number of photons in that light field," that is, (roughly) that x-1>x. The summary and background for non-specialists is here, and the full technical report is here (both behind subscription walls).

Now, this effect depends on the failure of commutation relations ubiquitous in quantum mechanics, so it is quite possible that this in no sense makes arithmetic look empirical. But I'm not 100% sure about that. Any thoughts?

APA wrap-up: Kyle Stanford's "New Induction"

My time at the APA last weekend was pretty good: I learned a few new things, met some new people I've been wanting to meet, and got to catch up with a couple old friends. Particularly helpful/ enlightening presentations included Angela Potochnick on how the context of inquiry shapes explanation, Ken Waters (plus commenters Jay Odenbaugh and Michael Strevens) on causes that make a difference (not, I learned, to be confused with the conception of causes as 'difference-makers'), and the Author-Meets-Critics session on Kyle Stanford's Exceeding Our Grasp.

Stanford's basic claim is that current scientific theories are underdetermined -- not because we can generate empircially equivalent rivals to our currently accepted theories, but rather because at many, many times in the past, the scientific community has been unable to conceive of good alternatives to the then-current theory. The evidence that such alternatives exist is the fact that they are proposed and accepted centuries later: thus, Newton's mechanics did not consider special relativity as an alternative hypothesis; Newton's, when he proposed his gravitation theory, did not consider the general theory of relativity as an alternative; no classical physicists before 1900 considered quantum mechanics as an alternative explanation of the data, and so on. This is what Stanford calls the "New Induction" over the history of science.

The idea, as just presented, strikes me as a promising line to take. But there is one aspect of Stanford's presentation of the problem that I don't understand. Fiona Cowie asked (in part) about this in the question and answer session, but I still didn't follow the answer. Stanford says that (e.g.) in 1700, the special theory of relativity and Newtonian mechanics were "(roughly) equally well-confirmed". Similarly for the other cases: the future theory is supposedly just as confirmed as the old one -- even in the past.

I don't understand why Stanford says this for two reasons: (1) He doesn't need the theories to be equally well-confirmed for his point to hold (viz., scientists aren't even conceiving of a hypothesis that will later be accepted as superior), and (2) it seems false to me, on any reasonable (i.e., not hardcore hypothetico-deductive)notion of confirmation. In 1700, it is true that special relativity and Newtonian mechanics agreed on all the consequences that could then be observed. But someone who, in 1700, said "Newton is approximately right, yet when something goes really, really fast its length will contract and its local time will dilate from the point of view of slower-moving observers" -- there is NO evidence at all for postulating that further bit of theory. And it's the same with GTR (what evidence would there've been for gravitation being a 10-component tensor instead of a scalar?) and especially QM (what evidence was there for thinking a body cannot have a determinate position and momentum simultaneously?). In 1700, these now-accepted alternatives were consistent with the data, but they were not equally well-confirmed.

Note: a very similar line of objection is pushed at the end of P.D. Magnus's "What's New about the New Induction?" (Synthese, 2006), though he develops it slightly differently, I think. (As I understand him, P.D. claims that in 1700, STR, GTR and QM would look like 'gruesome' hypotheses.)

A difference between science and the humanities?

This post is going to be overly ambitious and overreaching, but isn't that what the blogosphere is all about? Classes started this past week, and I am teaching an upper-level course in philosophy of science. I started the class with the (overly ambitious) question 'What is science?' We had an interesting conversation, and I learned a bit from my students. There was an extended discussion about which particular parts of forensic science really do count as science, and the question 'What is science?' really matters there, because once something is declared accepted science, then it can be admitted into evidence in a court of law.

I had one thought concerning how to distinguish science from the humanities, which I didn't share with my students, but I figured I might try to articulate here. I think it may just be a slightly different way of putting a tired old point, but here goes.

Both the sciences and the humanities seek understanding; both offer explanations of various bits of the world. At a very abstract level, though, the kind of things each tries to explain is different. Obviously, (e.g.) Hamlet is a very different kind of thing than (e.g.) thermometer readings -- I have a more 'formal' difference in mind. When an English professor gives an interpretation of Hamlet, she has (basically: see (3) in next paragraph) ALL the stuff to be explained in front of her: the text is complete, finished. To put the point in terms of evidence instead of explanation, all the data/evidence she can offer for her preferred interpretation of the text is already in. This stands in clear contrast with (most of?) the sciences: new data is constantly being gathered, and new observations need to be explained. If a similar process were occurring in the Hamlet case, a new Act of that play would be produced every week, and various interpretations were shown to be stronger or weaker as new 'data' (i.e. texts) came streaming in.

I think many parts of philosophy are more like the Hamlet case. Take ethics for example. Murder, stealing, lying, etc. are morally wrong. This is in some sense the 'received/ established text' from which the ethicist works: an ethical theory has to explain why those things are wrong. Of course, there are thought-experiments designed to probe various parts of our ethical intuitions, but (1) these often yield contentious/ equivocal results, and (2) more importantly, I only want to say philosophy is more like the Hamlet case, not that it's identical. (Plus, (3) the English professor could potentially get some new information "around the margins" (to put it metaphorically): further historical details of Shakespeare's life, earlier drafts of texts, various facts about the circumstances of production, etc. -- there is clearly a continuum here with absolutely no new data on one end and lots of new data on the other.)

Another philosophical example would be philosophy of arithmetic. We've known that 7+5=12 for some time now, and the philosopher wants to explain how and why we know it's true. We're not getting a whole lot of new 'data' about arithmetic. As a final example, consider the enterprise that falls under the heading of 'interpretation' of various special scientific theories -- an enterprise which is frequently thought of as being particularly close to science. This industry is perhaps most developed in the case of quantum mechanics, but it thrives in general relativity, statistical mechanics, population genetics, etc. as well. On my criterion, these projects (usually) fall much closer to the 'humanities' end of the spectrum: the game here is to take some standard formulation of the theory in question (e.g. of quantum mechanics) and provide an explanation of that theory. The formulation of (e.g.) quantum mechanics is, I think, like the text of Hamlet, insofar as it is taken to exhaust what needs to be explained. However, sometimes this explanation will (either by itself, or with further specification) generate new experimental predictions. When that happens (as in some cases of spontaneous-collapse interpretations of quantum mechanics), then the project becomes more scientific on the above criterion, and less like the humanities.

Perhaps the criterion I've suggested here can partly explain why many (though by no means all!) physicists express exasperation with the philosophy of physics, and consider the enterprise pointless.

Should a naturalist be a realist or not?

Unsurprisingly, the answer to the question in the title of this post depends on the details of what one takes 'naturalism' about science to mean. The shared conception of naturalism is something like 'There is no first philosophy' (I think Penny Maddy explicitly calls this her version of naturalism) -- that is, philosophy does not stand above or outside the sciences. "As in science, so in philosophy" is (one of) Bas van Fraassen's formulations.

Each of the following two quotes comes from a naturalist, but the first appeals to naturalism to justify realism (about mathematics), while the second appeals to naturalism in support of anti-realism (about science).

In his review of Charles Chihara's A Structuralist Account of Mathematics in Philosophia Mathematica 13 (2005), John Burgess writes:

"If you can't think how we could come justifiably to believe anything implying
(1) There are numbers.
then 'Don't think, look!' Look at how mathematicians come to accept
(2) There are numbers greater than 10^10 that are prime.
That's how one can come justifiably to believe something implying (1)." (p.87)

Compare van Fraassen, in The Empirical Stance (2002):

But [empiricism's] admiring attitude [towards science] is not directed so much to the content of the sciences as to their forms and practices of inquiry. Science is a paradigm of rational inquiry. ... But one may take it so while showing little deference to the content of any science per se.(p.63)

Both Burgess and van Fraassen are naturalists about their respective disciplines (mathematics and empirical sciences) -- but they disagree on what the properly scientific reaction to questions like "Are there numbers?" and "Does science aim at truth or merely empirical adequacy?" is.

The mathematician deals in proof. And proof is (at least a large part of) the source of mathematics' epistemic force. The number theorist (e.g.) assumes the existence of the integers and proves things about them; that's what she does qua mathematician. People with the proclivities of Burgess and van Fraassen would agree thus far, I think. But they part ways when we reach the question "Are there integers?" A Burgessite (if not John B. himslef) could say "If you're really going to defer to number theorists and their practice, they clearly take for granted the existence of the integers." A van-Fraassen-ite could instead say: "What gives mathematics its epistemic force and evidential weight is proof, and the number theorist has no proof of the existence of integers (or the set theorist of sets, etc.). Since there is no proof of the integers' existence forthcoming, asserting the existence of the integers (in some sense) goes beyond the evidential force of mathematics. Thus, a naturalist about mathematics should remain agnostic about the existence of numbers (unless there are other arguments forthcoming, not directly based on naturalism)."

Is there any way to decide between these forms of naturalism -- one which defers (for the most part) to the form and content of the sciences, and the other which defers only to the form? (Note: van Fraassen's Empirical Stance takes up this question, but this post is too long already to dig into his suggestions.)

Pessimistic induction + incommensurability = instrumentalism?

One popular formulation of Scientific Realism is: Mature scientific theories are (approximately) true.
One of the two main arguments against this claim is the so-called 'Pessimistic (Meta-)induction,' which is a very simple inductive argument from the history of science: Most (or even almost all) previously accepted, (apparently) mature scientific theories have been false -- even ones that were very predictively successful. Ptolemy's theory yielded very good predictions, but I think most people would shy away from saying 'It is approximately true that the Sun, other planets, and stars all rotate around a completely stationary Earth. So, since most previous scientific theories over the past centuries turned out to be false, our current theories will also probably turn out to be false. (There are many more niceties which I won't delve into here; an updated and sophisticated version of this kind of argument has been run by P. Kyle Stanford over the last few years.)

The kind of anti-realism suggested by the above argument is that the fundamental laws and claims of our scientific theories are (probably) false. But we could conceivably read the history of science differently. Many fundamental or revolutionary changes in science generate what Kuhn calls 'incommensurability': the fundamental worldview of the new theory is in some sense incompatible with that of the older theory -- the changes from the classical theories of space, time, and matter to the relativistic and quantum theories are supposed to be examples of such changes. Communication breaks down (in places) across the two worldviews, so that each side cannot fully understand what the other claims.

Cases of incommensurability (if any exist) could result in each side thinking the other is speaking incomprehensibly(or something like it), not merely that what the other side is saying is false in an ordinary, everyday way. An example from the transition from Newtonian to (special) relativistic mechanics may illustrate this: Suppose a Newtonian says 'The absolute velocity of particle p is 100 meters per second.' The relativistic physicist would (if she is a Strawsonian instead of a Russellian about definite descriptions) say such a sentence is neither true nor false -- because there is no such thing as absolute velocity. [A Russellian would judge it false.] If she merely said "That's false," the Newtonian physicist would (presumably) interpret that comment as 'Oh, she thinks p has some other absolute velocity besides 100 m/s; perhaps I should go back and re-measure.' To put the point in philosophical jargon: presuppositions differ between pre- and post-revolutionary science, and so the later science will view some claims of the earlier group as exhibiting presupposition failure, and therefore as lacking a truth-value, like the absolute velocity claim above. (Def.: A presupposes B = If B is false, then A is neither true nor false)

This leads us to a different kind of pessimistic induction: (many of) the fundamental theoretical claims of our current sciences probably lack a truth-value altogether, since past theories (such as Newtonian mechanics) have that feature. (If you want to call claims lacking truth-values 'meaningless,' feel free, but it is not necessary.) This is hard-core instrumentalism, a very unpopular view today; most modern anti-realists, following van Fraassen, think that all our scientific discourse is truth-valued (though we should be agnostic about the truth-value of claims about unobservable entities and processes). But this instrumentalism seems to be a natural outcome of (1) taking the graveyard of historical scientific theories seriously, (2) believing there is something like Kuhnian incommensurability, and (3) holding that incommensurability involves presupposition failure. And none of those three strike me as crazy.

Disclaimer: This argument has probably been suggested before, but I cannot recall seeing it anywhere.

realism and the limits of scientific explanation

Long time, no blog. I finally got back a few days ago from the last of my visits to schools for final job interviews. It was very interesting and instructive to observe non-Pittsburgh philosophers in their native habitats. I should know by the end of this week where I'll be next year.

In lieu of an actual post, I am putting up the handout I used at a couple of my job talks. As a result, it looks programmatic/ bullet-pointy; but I tried condensing this into a normal post, and it was just far too long. If you can make out what's going on, I would really appreciate any feedback/ comments/ eviscerations from readers.

REALISM AND THE LIMITS OF SCIENTIFIC EXPLANATION

The argument

(P1) Scientists do not accept explanations that explain only one (type of) already accepted fact.
(P2) Scientific realism, as it appears in the no-miracles argument, explains only one type of already accepted fact (namely, the empirical adequacy or instrumental success of mature scientific theories).
(P3) Naturalistic philosophers of science “should employ no methods other than those used by the scientists themselves” (Psillos 1999, 78).

Therefore, naturalistic philosophers of science should not accept scientific realism as it appears in the no-miracles argument.

Explanation and defense of (P1)

Explanations that explain only one type of already accepted fact
(i) generate no new predictive content, even when conjoined with all relevant available background information [‘already accepted fact’], and
(ii) do not unify facts previously considered unrelated [‘only one type’].

Evidence for (P1): Scientists reject
- Virtus dormativa-style explanations
- ‘Vital forces’/ entelechies as explanations of developmental regularities
- Kepler’s explanation of the number of planets, and the ratios of distances between them, via the five perfect geometrical solids
- ‘Just-so stories’ in evolutionary biology

The no-miracles argument for scientific realism

Abductive inference schema
(1) p
(2) q is the best explanation of p
Therefore, q

No-miracles argument for scientific realism
(1) Mature scientific theories are predictively successful.
(2) The (approximate) truth of mature scientific theories best explains their predictive success.
Therefore, Mature scientific theories are (approximately) true.

Proponents of the no-miracles argument (Putnam, Boyd, Psillos) accept (P3), appealing to naturalism to justify their abductive inference to scientific realism. Putnam claims that scientific realism is “the only scientific explanation of the success of science” (1975, 73).

The argument for (P2): Scientific realism (i.e., the claim that mature scientific theories are approximately true)
(i) generates no new predictions,
(ii) unifies no apparently disparate facts, and
(iii) explains only one previously accepted fact, viz., science’s predictive success.

Einstein and the Units of Selection

No, the title of this post is not a typo. I just finished reading through the first three articles in the most recent issue of Philosophy of Science. They are an argument-response-rebuttal between Elisabeth Lloyd ("Why the Gene Will not Return"; "Pluralism without Genic Causes?") and Ken Waters ("Why Genic and Multilevel Selection Theories Are Here to Stay"), who is one of her targets in the original essay. As the biologically-inclined among you will have inferred, this is the latest installment in the long-standing units of selection debate; very roughly, the question in these debates is: Upon what does natural selection operate? Organisms? Genes? Groups of organisms?

Ken Waters' basic response to this question -- which he first articulated in "Tempered Realism about the Force of Selection" (Philosophy of Science 1991) -- is that there is no determinate fact of the matter about whether selection is really acting at the level of the gene or the organism/ genotype. Mathematical models can be constructed in terms of genes and in terms of genotypes, and both kinds of model suffice to represent the facts of dynamic changes in populations. (See "The Dimensions of Selection," P. Godfrey-Smith and R. Lewontin, Philosophy of Science 2002, for an excellent treatment of the niceties of of the situation.) Since these different models do not represent different facts, Waters concludes that we will choose between them on pragmatic grounds. In the language of his current paper, Waters says that different models "parse" the causal structure differently.

For the purposes of this post, I will assume Waters is correct to maintain that there is no fact of the matter about whether the true cause of any particular evolutionary change lies at the level of the gene or the genotype. What I want to do is to compare this situation with Einstein's reaction in (what I consider) an analogous situation.

At the beginning of Einstein's 1905 paper that introduces special relativity, he asks us to imagine a conductor and a magnet in relative motion with respect to each other. If the take the conductor to be at rest and the magnet moving, then Maxwell's theory says that an electromotive force is generated in the conductor, which gives rise to an observable electric current C. If, on the other hand, we assume the conductor is moving and the magnet is at rest, then Maxwell's theory says that no electromotive force is generated in the conductor, but an electric field is generated around the magnet -- and this field induces exactly the same electric current C as before. Einstein's conclusion is that we are not actually dealing with two physically different situations here; rather, our theoretically distinct models are representing one and the same set of facts. This is exactly Einstein's argumentative maneuver in his famous elevator thought-experiment as well: though the pre-Einsteinian theory would distinguish between the cases in which I am being uniformly accelerated through a gravitation-free region and in which I am at rest in a homogenous gravitational field, Einstein maintains that there is in fact no difference between these two cases. This is (one version of) the Principle of Equivalence.

Note that Einstein does not say is that 'we choose between the competing descriptions of the magnet-and-conductor case on pragmatic grounds,' or that 'we parse the causes differently: either as an electromotive force or as a electric field.' Rather, he re-arranges the permitted causal structures of the theory to eliminate these pseudo-differences, so that the theory no longer "leads to asymmetries which do not appear to be inherent in the phenomena." He replaces the separate categories of 'inertial effects' and 'gravitational effects' with a single category (which we could call gravitational-inertial effects) via his principle of equivalence.

What I am curious about is whether Einstein's maneuver can be carried over into the biological case. I am hoping someone better-informed than I am can tell me why this has no prayer of working, or why Einstein's cases are not analgous to the situation in evolutionary biology. Of course, I wouldn't mind hearing suggestions for how this might work, either.

Editorial note. Posting here will probably be sporadic for the next few months: I am going on the job market this year, and that process has been (and, I imagine, will continue to be) time-consuming.

Specialization and collaboration, again

Yesterday Paul Hoyningen-Huene presented a talk entitled "What is Science?" at the Center for Philosophy of Science here. He intends the question in his title to be taken in a very general way, so his target is one of those Big Questions that, in my last post, I bemoaned as a dying breed in our climate of increasing specialization.

Prof. Hoyningen-Huene pointed out a discouraging fact for anyone who wants to attempt an answer to the Big Questions in the philosophy of science and simultaneously remain reasonably close to actual scientific practice: according to Thomson ISI (the citation management company), there are 170 categories of natural science, 54 in the social sciences, and 15 in the formal sciences -- not including subdisciplines, which can vary widely. So if someone makes a general claim about science or scientific practice, and wants to check that claim thoroughly, then 239 different categories of scientific activity -- most of them complex and varigated -- must be checked.

I feel pulled in two directions by the existence of these 239 categories. On the one hand, it seems that collaboration is the only means to make headway on the Big Questions. On the (not-quite-mutually-exclusive) other, it seems likely that the Big Questions just won't admit of anything approximating a (reasonably) general answer. (Hoyningen-Huene's strategy is to describe several examples drawn from across several scientific disciplines that support his thesis, and assert that these examples are paradigmatic.)

Finally, Kieran Setiya has also recently posted about specialization in philosophy over on his blog, Ideas of Imperfection. Since he's much smarter than I am, I recommend you read his post.

Specialization and collaboration

Over the past few decades, philosophy -- and philosophy of science in particular -- has become increasingly specialized: we have philosophy of quantum field theory, philosophy of developmental biology, etc. It seems that even the so-called "generalists" in philosophy of science are becoming a more and more self-contained group. (For example, I went to a session entitled "Confirmation" at the last Philosophy of Science Association meeting, and I had a very difficult time understanding what was being discussed, at least in part because there was a lot of specialized jargon and assumptions shared by the experts used without explanation -- though my limited brainpower certainly played its part in my incomprehension.)

In general, I think this trend of specialization is a Good Thing, primarily because it has led to specific results that we might not have found otherwise. (Thus I disagree with Karl Popper's claim: "For the scientist, specialization is a great temptation, but for the philosopher, it is a mortal sin.") But I think specialization also has its costs -- in particular, we tend to bypass answers to bigger questions. The question "What is a scientific explanation?" is replaced by "What is explanation in quantum information theory?" or "What is an evolutionary explanation?" and so on. (I think both of those questions are very interesting and philosophically important ones!) The philosopher of biology is uncomfortable talking about explanation in the physical sciences, and the philosopher of physics feels likewise about explanations of biological phenomena -- and the generalist is busy worrying about 'grue'some predicates, the barometer and the thunderstorm, or the irrelevant conjunction problem to deal with explanations in particular sciences. (I think this may in part explain why philosophy of science survey classes often begin with writings of logical empiricists: they tried to give genuinely general accounts of notions central to science.)

In keeping with the generally naturalist spirit of philosophy of science and this blog, we can ask ourselves: What Would Scientists Do? Scientists these days are hyperspecialized, and publish their hyperspecialized research in increasingly specialized journals. However, they also answer bigger, broader questions as well, via collaboration with scientists outside their specialty. So I wonder whether the time is ripe now for philosophers of science, armed with the insights about their particular sub-disciplines amassed over the last few decades, to begin collaborating to answer some of the bigger questions again. And the collaborations need not end there -- philosophers of science could also collaborate more with folks working within epistemology and metaphysics proper, or other fields.

I imagine many will say that we have overthrown the logical empiricist myth that there is a single thing, explanation, or confirmation, or even science. I am open to the idea that these might be myths. But I think we should check whether this is the case -- and if they are mythical, we can at least gain clarity and specificity about what the differences are between e.g. the explanatory patterns of physics and biology.

Underdetermination and equivalence modulo p

Since the description of this blog states that it deals with "issues in logic" related to philosophy of science, I figure that, for the sake of truth in advertising, I should post something logical. (Though I don't feel particularly rushed: there are already a fair number of smart logicians actively participating in the blogosphere -- check my blogroll. For reasons I don't understand, the situation is different in philosophy of science. Any armchair anthropologists have an explanation?)

Underdetermination arguments occur in many quarters of philosophy: Descartes' demon is perhaps the most famous, but they have also played a leading role in discussions about scientific realism during the last few decades. In this post I want to characterize a particular sort of underdetermination using elementary logical notions. (This form of underdetermination either is -- or is closest to -- the Quine-Duhem variety, I'm not sure which at the moment.)

Consider two sets of sentences, A, B such that neither set is a logical consequence of the other. Now suppose there is a third set of sentences C such that:
If C then (A iff B).
That is, if we assume that C is true, then A and B are logically equivalent. (In all models where C is true, either both A and B are true, or both are false).
Then we say A and B are equivalent modulo C.

How does this relate to underdetermination? We can have two theories that are not logically equivalent (and thus are not 'the same theory'), but do become logically equivalent if we make some further assumptions (C above) -- and these further assumptions can be taken to be "auxiliary hypotheses" or "background knowledge" (or whatever one wishes to call the other claims a theory uses, besides its own, to make predictions). If we are committed to the truth of the background knowledge, then we cannot decide between the two theories.
(If this is a bit abstract, here's a toy example:
A = M and (if p then q) [assume M says nothing about p or q]
B = M and (if p then not-q)
C includes the sentence 'not-p';
so neither of A and B implies the other; if C is true then A and B are logically equivalent, while if C is false then A and B are inconsistent.)

My question: is anything philosophically interesting going on here? If we hold r to be true, do we really need to choose between (r or s) and (r or not-s)? I think not -- though they differ in logical content, they are not rivals (or are they?). At least, if we take r to be true, then they are definitely not rivals, though they might be considered rivals ‘on their own’. They certainly are genuine competitors when we hold r false -- though then they are no longer equivalent in any sense.

Comments:
1. This is not the usual sort of underdetermination situation. First, the notion of "empirical content" (or "empirical equivalence," i.e. identity of empirical content) does not appear, so the much-maligned observable/ unobservable distinction is never mentioned. Second, and more importantly, the two theories A and B are not incompatible: the Cartesian demon, on the other hand, is either deceiving us or not (at least on the usual interpretation). The demon-hypothesis is incompatible with the 'real-world' hypothesis. On the other hand, 'if p then q' and 'if p then not-q' are not inconsistent -- we need simply hold that 'not-p' is true.

2. How does this relate to a ‘real’ example, e.g. Bohmian mechanics? It is empirically equivalent to standard quantum mechanics as long as absolute position is undetectable -- but not if absolute position is detectable. In other words, the standard theory and the Bohmian theory are empirically equivalent modulo the claim that absolute position is undetectable. And that is structurally similar to the toy example above. (Of course, there is the difference in this case that the two theories are 'empirically equivalent modulo p,' not 'logically.')

3. Lastly, it is probably considerations akin if not identical to the above that prompted philosophers to move to ‘total theories’ (i.e. theories PLUS all their auxiliary assumptions) as the proper objects of epistemic evaluation. See e.g. (Leplin, Erkenntnis, 1997).

Is 'p is a priori' itself a priori?

My most recent posts have been far too long by blogging standards, so I am determined here to be brief. Most philosophers (though not all, e.g., anyone who holds the Quine-Putnam indispensibility thesis) believe the following sentences are true:
(1) "'2+3=5' is true a priori"
(2) "'The earth is round' is true a posteriori"
The question is: are (1) and (2) true a priori or a posteriori? Put metaphorically, does experience teach us that logical theorems are known independently of experience? I am not sure there are any non-question-begging arguments to be given one way or the other; if there is one, I would love to hear it.

That's all I wanted to say. If you are curious about what motivated me to think about this question, keep reading. There are two motivating sources:
First, my dissertation, which deals with the academic year Carnap, Tarski, and Quine spent together at Harvard in 1940-41. In a private conversation Carnap had with Quine, one way they formulate the difference between themselves is as follows.
Carnap: 'p is analytic in language L' is itself an analytic statement.
Quine: 'p is analytic in L' is a synthetic statement, to be settled by a behavioristic investigation into the linguistic habits of L-speakers.
Granted, 'analytic' is not identical to 'a priori' -- but for Carnap, they were extensionally equivalent, and the question above is very close to this issue.
Second, for the last 10 years, van Fraassen has been suggesting that we think of empiricism not as a theory or assertion but as a stance. (See "Against Naturalized Epistemology" (1995) in On Quine and 2002's The Empirical Stance.) The primary argument he offers is that empiricism, if conceived as an assertion, is self-defeating: "All knowledge about the world is a posteriori" (or any other slogan intended to capture the empiricist's thesis) will be difficult to construe as having experience as its source. And, van Fraassen says, if the empiricist thesis cannot be justified on the basis of experience alone, then it fails to live up to its own standards, and is therefore self-defeating. But van Fraassen is assuming that "such-and-such is a posteriori" must itself be an a posteriori claim. And that is taking for granted an answer to my question above.

On Rosenberg and Kaplan's "Physicalism and Antireductionism in Biology"

I've just finished reading Alex Rosenberg and D. M. Kaplan's "How to Reconcile Physicalism and Antireductionism about Biology" in the current issue of Philosophy of Science. The title describes its contents perfectly, and I can unequivocally recommend it to anyone interested in the topic (which is more than I can say for my foray into the subject). I'll quote the key part of the introduction:

In considering the relation between the PNS [Principle of Natural Selection] and physical science, three alternatives suggest themselves:
a. The PNS is an underived law about biological systems, and is emergent from purely physical processes. ...
b. The PNS is a derived law; it is derivable from some laws of physics and/or chemistry. ...
c. The PNS is an underived law about physical systems (including non-biological ones), and from it the evolution of biological systems can be derived... This is an alternative no one has canvassed, and one which we shall defend here.

Before continuing, let me give their version of the PNS:

For all x, y, and E: If x is fitter than y in environment E at generation n, then probably there is some future generation n', after which x has more descendants than y.

Note that the domain of quantification has no restrictions.

Their basic rationale for calling the PNS a law of physical science -- in particular, of chemistry (59, 62) -- is this: if there is a kind of chemical molecule that (in some sense) replicates itself and has higher rates of 'survival' than other molecules in a given reaction (say, as a reaction moves towards equilibrium, this kind of molecule is favored), that molecule will be subject to the PNS. (It's not a law of physics simply because (sub)atomic particles cannot be construed as replicating.) Calling the PNS a law of chemistry instead of biology "is just a picturesque way of drawing attention to the fact that selection for effects only begins to operate at the level of chemical interactions... Similarly, we call the second law of thermodynamics a law of physics, even though it obtains for all systems -- physical, chemical, and biological -- since it is at the level of the physical that it begins to operate" (61). As an example of such chemical natural selection, they point to current models of origins of life research.

So far, so good. But I'm less comfortable with the other half of their claim in c. above, viz., that from the physico-chemical PNS a fully biological PNS can be derived. They rephrase this point later in the article:

According to this view, at each level of the organization of matter there turns out to be a PNS, and each one should be in principle derivable from the PNS for the immediately lower level or some other lower level(s), all the way back down to the PNS for molecules. (61)

The problem is that they do not explain how this (in principle) derivation would proceed. They phrase the point slightly differently elsewhere (the PNS's "operation at higher levels of the aggregation of matter is a consequence of the operation of the underived PNS for molecules together with the rest of physical law" (62)), but they never actually spell out the derivation beyond this -- as far as I can tell. (In one place (top of p.62: "The rest is natural history"), they appear to hint that any higher-level PNS is a 'consequence' of chemical PNS in a historical sense: because the PNS acted on the primordial soup, today's organisms came into existence -- yet that is completely unlike any sort of reduction any philosopher of science that I know of has talked about. So I assume they can't mean that.)

So the question to Rosenberg and Kaplan is: does natural selection operate on the biological realm (whether it be genes, individuals, or even groups) because it operates on the chemical-molecular level? -- where that 'because' has the same force as the one in 'This mole of gas has a higher temperature than that one because this one's molecules have a higher mean kinetic energy.' Here's one way to press this worry. Look at their definition of the PNS quoted above, and convert it into 'the PNS for molecules':
For any molecule x and molecule y, if x is fitter than y, then...
Now they say that from this (and perhaps other PNSes), we should be able to derive higher level PNSes:
For any gene x and gene y, if x is fitter than y, then...
For any organism x and organism y, if... then...

But these higher-level PNSes don't appear to follow at all. Certainly, each follows from the general PNS quoted at the beginning; but from the fact that all chemical molecules behave a certain way, you cannot infer that organisms will behave a certain way -- unless organisms are chemical molecules. Rosenberg and Kaplan might say at this point: but all we are is an aggregate of chemical molecules: that is just the physicalist thesis which we profess in the title of our paper. But 'Aggregates of As are B' does not in general follow from 'As are Bs,' even for the most determined physicalist.

Naturalism and the alleged continuity between philosophy and science

There are probably as many varieties of naturalism as there are naturalists. Either despite or because of this variety, many philosophers today are naturalists (though noteworthy dissent also exists). Philosophers of science, in particular, seem to be uncompromising naturalists. Here is one slogan that I think does a decent job of expressing the basic naturalistic position in a minimal enough form that most naturalists would accept it:
For any question, approach and answer that question the way a scientist would approach and answer that question (leaving the particulars open at this level of abstraction).
In other words, philosophers should defer to scientists on both questions of both method and factual content.

Another tenet of many modern naturalisms is "science is continuous with philosophy," a phrase which traces back to Quine. There is certainly something right about this notion, for we can point to current thinkers whose work does not fall neatly into science or philosophy: the late, great Rob Clifton is a paradigmatic example, as well as many other technical philosophers of physics working today. There are similar cases in other sciences too: should Sober and Wilson's Unto Others be classed as biology or philosophy?

However, it seems to me that "Science is continuous with philosophy" is also in some tension with the original naturalist slogan. If you asked an interdisciplinary team of scientists to answer the questions:
"Am I a brain in a vat (or being deceived by an evil demon, etc.)?"
"Is knowledge justified true belief?"
"Is the meaning of a sentence identical with its truth-condition?"
"Is the fundamental aim of science truth or empirical adequacy?"
they would (I think) say that such questions are not scientifically tractable -- in such cases, the 'scientific approach' (whatever exactly that might be) cannot answer that question. I think this would be the scientists' answer on the grounds that none of these questions -- or the vast majority of the others that appear in The Journal of Philosophy, Nous, The Philosophical Review, etc. -- are ever addressed in Science, Nature or other leading scientific journals.

In short: scientists do not view science and philosophy as continuous (even though there are borderline cases in technical philosophy of the special sciences), so a philosopher who views them as continuous is not fulfilling the naturalist's commitment to defer to the sciences. (Though the first stirrings of this discontinuity were felt earlier in the 17th C., I would guess that it becomes explicit with Newton and his rules of philosophizing.) So what is a philosopher (who is not Rob Clifton) to do? I think part of the appeal of naturalism is that science is seen as epistemically privileged, and if we can lump philosophy in with science, then that epistemic privilege and prestige will rub off on philosophy. I think the moral to be drawn from the discontinuity is just that philosophy lacks science's epistemological privilege -- and that is a conclusion a naturalistically-inclined philosopher might happily accept anyway.

First real post: abduction from limited evidence

I am currently teaching a summer session class called "How Science Works." The aim of the class (according to the course description) is to "introduce science and scientific ways of thinking to students who have not had much contact with science." One of the big-picture themes I am trying to develop in the course is that, in many cases, scientific reasoning is just extra-careful common-sense reasoning made precise. Thus, for some of their homework assignments, I ask students for an example from their lives where they (or someone they know) have used whatever scientific strategy we are studying that day.

On their last homework assignment, I asked them to give an everyday example of an abductive inference that is supported by independent sources of evidence, a very common kind of inference in good scientific practice. There were some very clever answers; the most common one was the cheating significant other: I never see the significant other with someone else, but I do observe her staying out late, hanging up the phone immediately when I stop by unexpectedly, and various other secretive behaviors.

But, unsurprisingly, some students did not quite understand the requirement for evidence from multiple independent sources, and gave an example of an abduction from just one. (These examples often involved some sort of roommate malfeasance: "My favorite CD is no longer in its case, so my roommate must have taken it without asking.")

But this mistake raised one version of a question I've been wondering about off and on for a while now. This version of the question is:

What is the epistemological status of abductive inferences (also known as 'inferences to the best explanation') that have only one (independent) source of evidence?

To make this a slightly more well-posed question (or at least a question more amenable to philosophers of science), we can pose the question more naturalistically: how does science regard and treat such inferences?

My gut-feeling answer is that such inferences are regarded as 'merely hypothetical' or otherwise second-class citizens -- like atomic theory in the 19th C, for example (for an account of this historical episode very congenial to my point, see Penelope Maddy's "A Problem in the Foundations of Set Theory, Journal of Philosophy Nov. 1990, pp. 623-625). Often such abductions from limited evidence are pronounced ad hoc. And, in biology, this also seems to be the reason many people call certain adaptionist accounts of certain biological traits "just-so-stories": if the only evidence we have for that adaptationist account is the way the animal looks or acts now, then we just don't have strong reason to accept that account. If we found enough sources of independent evidence for our conjecture, even the strong anti-adaptationists might come around.

I think this question (of the status of abductive inferences based on only one evidential source) is an interesting one in its own right. But it seems it also has implications for philosophy proper: many philosophical positions purport to explain only one thing, and lack the kind of consilience that good abductive inferences in the sciences enjoy. To take just one example, scientific realism aims to explain the success of science -- and as far as I can tell, that's it. (At least some other realisms will definitely fall in the same boat.) Certain critics of scientific realism (van Fraassen, for one) criticize all instances of "inference to the best explanation" (IBE), and use that criticism as an argument against scientific realism (which is supposedly the result of an instance of IBE). However, we may not need to throw out all instances of IBE in order to have a good argument against the realist -- we need only act like the scientists (as I portrayed them in the previous paragraph): accept only those instances of IBE for which we have more than one independent source of evidence.