I mentioned the other day the problem of mechanics being undetermined in its meaning. (Yes, I know - I'm behind in posting) That is we had three rather conceptually different ways of thinking about mechanics but no real way to choose between them. However science often is even more undetermined than this. That is the problem isn't underdetermination of how to take a theory. Rather given a set of evidence it is underdetermined which theory or mathematics we should use to describe it. This notion of underdetermination actually comes in several different forms. Unfortunately loose talk often conflates these very different conceptions. One of the most referred to forms of underdetermination is labeled the Duhem-Quine thesis after one of the prominent early philosophers of science from the beginning of the 20th century and of course the famous logician and philosopher who wrote about this in his wildly influential, "The Two Dogmas of Empiricism." Unfortunately even this more limited underdetermination thesis is misunderstood. I thought for my third post for the week of science blogging it would be useful to talk about it. There's simply a lot of misinformation out there about this important topic.
The first notable undetermination thesis can be found in Hume. This is the claim that for deductive logic there are an infinite number of mutually incompatible theories that explain the evidence. It's easy to prove and is probably the most influential form of underdetermination. Unfortunately it's also perhaps appealed to far too often. After all even if we can't deductively distinguish between theories we can appeal to other rational processes to judge theories. The problem of induction can come into play here. But while this is perhaps a problem for those demanding certainty in epistemology it isn't a problem for science which makes no such demand and which appeals to induction quite regularly.
Often appeals are made to what amount to variations of Humean underdetermination. In general most of these appeals, for whatever reason, tend to simultaneously reduce the rational to deduction. This is unfortunate. It tends to result in caricatures of positions and typically provides strawman arguments. You'll especially find variations of this in what are loosely called postmodern theory. (Not all that goes under the rubric of postmodernism is bad - for a while I used the term myself until it collapsed under the the weight of too many using the term while appealing to shoddy thinking)
It's important to note that Humean underdetermination is primarily a logical thesis. It deals solely with the problem of deduction.
The next major work on underdetermination was Duhem. Duhem rightly criticized Newton's claims to have developed his laws of gravity from Kepler's laws. Kepler's laws, you might recall, were the recognition that planets followed ellipses with the sun in one of the foci. The problem is that if we take Newton seriously then with all the bodies in the solar system none of the planets follows an ellipse. That's because each extra planet will slightly disturb the orbit of all the others due to gravitational interaction. There are a few other similar problems mathematically (such as all bodies having mass and not just the sun) One can not empirically move between the two. Thus if Newton is right Kepler is wrong and vice versa.
Duhem's thesis is that physical law is holistic. One can't consider a physical law in isolation from all other physical laws. Of course physicists do this all the time as simplifications. But one has to recognize that empirically these are simplifications. When judging laws one must judge the entirety and not just a few isolated theories.
It's important to keep in mind (as most don't) that Duhem was writing only about physics. While one could claim that say biology can be reduced to physics this need not be the case. Further biological laws as held might not be so reducible. So how one judges the applicability of Duhem to science as a whole tends to be tied to how one views the relationship of physics to the rest of science. Whatever ones position there, Duhem certainly never made wider ranging claims.
Where the practical implication for underdetermination comes in is that since we can't falsify theories except by examining the full system of physics, a creative thinker can with enough creativity figure a way to fix a failing theory. Let us say, to take a modern example, that we have a theory of gravity for the whole universe but that doesn't predict the motion that astronomers see. Now we could say that the theory (in this case General Relativity) is wrong. Or we could postulate some new, invisible, matter making up most of the universe. Guess which choice physicists picked? Given this though in most cases we can't really pick out which solution is best. There are numerous other examples, such as Einstein's fudge factor to GR which he later regretted (but which some physicists later on have added back).
Duhem's claim was much more a practical one. He thinks that what lets us choose between these competing theories which can't be decided logically (roughly deductively) is "good sense." It's not entirely clear what he means by good sense. It's roughly Peircean abduction from what I can see. But he feels most scientists have it and that it is rational.
Moving on to Quine, in his famous work he provided an epistemological thesis. While still very tied to logic Quine simply made a more far reaching claim than either Hume or Duhem. In "Two Dogmas of Empiricism" he said, "any statement can be held true come what may, if we make drastic enough adjustments elsewhere in the system." (This differs from Duhem by being true of all knowledge rather than just formal theories in physics) Now it's sometimes a bit unclear what Quine meant by this. Typically he's taken to be embracing a kind of holism where every term in the universe depends upon every other term. So Quine is thus asserting that one can rationally defend any statement due to this holism. However it's important to note that Quine puts forth this idea as a postulate. He never actually provides a proof nor really a strong argument for this belief. Roughly his argument is that we reject theories with a "falsifying instance." (Roughly Popper's approach) However Quine suggests that any evidence of falsification can be discounted rationally even if, in the extreme case, we can only do this by pleading that it was all a hallucination. (I'd note as an aside that this is often taken as a helpful method of discounting religious evidence)
If we consider Quine formally we can see that there are two aspect to his claim.
(1) for any theory T1 and body of supporting evidence E there is at least one competing theory T2 that is as well supported by E as T1 is
(2) any theory T1 is as well supported by evidence E as any other theory t2
Now our reading (2) simply seems problematic on the face of it when we think about it for a moment. Perhaps if we limited ourselves to deductive reasoning we could justify (2). But then it's really just Hume's theory taken epistemically>. It is also too trivial. Clearly reason involves more than deduction and we can distinguish rationally between the idea my missing keys are due to my dropping them from my missing keys being due to a conspiracy of key thieves hiding out in my house. Some ideas are simply more reasonable than others. So we can probably reject this reading out of hand. This kind of relativism, even if it was what Quine meant, simply doesn't have any argument supporting it. (And, given the quote in Two Dogmas that we could hold onto any statement come what may, it does seem to be what Quine was asserting) The problem is that it is demonstrably counter to our practical reasonings. To counter Quine's example, it is not always rational to claim that one had an hallucination.
Even if (2) is indefensible given that our rationality is more far ranging than deduction we still may be able to defend (1). However it is not clear what is meant by "well supported." Larry Laudan in "Demystifying Undetermination" provides an interesting argument that applies equally well to both (1) and (2). Consider a Bayesian approach to determining what is "well supported." In Bayesian methods one can only consider evidence from when one begins to calculate. That means that if T2 is submitted after T1 then according to Karl Popper any evidence from before its submission doesn't count as confirmation for T2. Thus while both T1 and T2 account for the same evidence equally the evidence doesn't support both equally.
Now of course many will cringe at Bayesian probability being used in this fashion. However even if we reject that use it demonstrates a problem. We often judge the success of theories in large part due to their predictions. But this simply means that older theories that make predictions are valued more than later theories that may be equally empirically adequate. Put an other way, the methods of reasoning need not entail that equal empirical adequacy entails equal rationality.
To see this go back to our discussion of the three mechanics. Even though Newton's formulation of mechanics, the Lagrangian and the Hamiltonian were equally adequate empirically people tended to consider "right" the Newtonian formulation. Further for a wide variety of reasons one can see that this was reasonable. First off Newton came first. Secondly Newton's "objects" accorded better with our common sense and practices. Now we can still appeal to this issue of empirical adequacy but unless we carefully constrict what it is to be rational one can't say that empirical adequacy entails epistemic equivalency.
Clark interpreted Quine: "In "Two Dogmas of Empiricism" he said, "any statement can be held true come what may, if we make drastic enough adjustments elsewhere in the system."
You interpret Quine to be talking about some logical system or physical system. I believe he is talking more generally about belief systems. What he is saying, as I have read him, is that when we are faced with conflicts in a system or way of viewing a problem or even something as general as our experience in general, we have a tendency to want to make changes to that system rather than abandon it. We will continue making adjustments within a given system until we recognize that abandoning the system is warranted because there is a better way of making sense of the data. So those who have an investement in a way of seeing things will continue to create epicycles rather than abandon a helio-centric view for instance; whereas if we can see that there is an alternative way to make sense of the data we may just come to the conclusion that a way of dealing with the data just won't work and the new system is superior. But the same practice of making adjustments to a belief system would work the same way with more general belief systems as well. At least that is how I have read him.
I don't think he's talking about a physical system. He's clearly talking about knowledge or beliefs. But I don't think he's merely talking about human tendency but rather what can be empirically verify the meaning of statements. It seems to me that it falls out naturally from his holism and his Peircean verificationalism. The big issue (and interestingly we're talk about this at HOPOS-L) is how Quine's nominalism falls into all this.
The key issue in philosophy of science though is the degree to which one can do this rationally.
If Quine is merely saying that we can irrationally make arguments to maintain our beliefs then he's making a rather trivial point with little philosophical significance. (This is why Laudan rejects the reading you take fairly early on in his criticism of Quine)
I think part of Quine's point is that beliefs about what a "rational" move is is also something which is up for revision. Thus his remarks about the revisability of logic (if we feel the need to go /that/ far, which he doesn't think we have good reason to; Quine never backs down from classical logic).
Of course, whatever background beliefs we hold (whatever beliefs we aren't actively considering for revision at the moment) will rule out the vast majority of possible theories we might decide on. But all beliefs are revisable, in principle, and so any statement can be "held true come what may" in principle.
Imagine that the statement in question is what we take the world to be insisting upon, in our experience of it, and I think this becomes easier to accept. This sort of thing occasionally happens in sci-fi/fantasy stories; So-and-so has a mind-boggling experience (say, of the Inner Chambers Of The Illuminati) and is convinced from that point on that what's ordinarily taken for reason is a bunch of bunk.
I think that's certainly true - thus his point about even logic being up for debate with the example of the law of excluded middle and quantum mechanics. However even if logic itself may be up for revision it doesn't follow that one can defend anything. As you yourself point out - Quine doesn't think we have a good reason to reject classic logic. Thus reason can adjudicate between claims. Which was ultimately Laudan's point.
Well he doesn't think we have any good reasons to reject classical logic on account of other beliefs that he's not holding up for revision at the moment. But he might come around eventually and, say, stop asserting bivalence because of fretting about vagueness. In which case some of the beliefs he has previously not held up for revision will have to be revised.
I took Quine's point in the end of "Two Dogmas" to be basically what Sellars says a bit later (about science, but I think the point can be taken more broadly): Any element in the web of belief can be revised, but not all at once.
Yeah, but I think the question that makes it problematic is whether this revision can be rational and how that affects what can be revised.
I've closed comments in order to avoid spam since I don't check this older blog as much anymore.
and Theology
Number of unique visitors:
Blogged by Clark Goble