Saturday, February 7, 2009

Scientific Realism and Epistemic Justification

As far as I can see, much of the recent scientific realism debate revolves around the question of whether or not we are epistemically justified in believing our best scientific theories (or parts thereof) to be (approximately) true. In light of this, I've always found quite surprising that philosophers of science seem to be by and large uninterensted in the debates among epistemologists concerning epistemic justification. Among the few exceptions that I can think of are those scientific realist (such as Richard Boyd and Stathis Psillos) who more or less explicitly take an externalist/reliabilist stance on epistemic justification. However, I can hardly think of any scientific realist who takes an explicitly internalist/evidentialist stance. Yet, I don't see why an internalist/evidentialist view of justification wouldn't suit the purposes of (some) scientific realists. Do you think this is just an accident? Or do you see any serious reasons for a scientific realist to think s/he is better off being an externalist about epistemic justification (independenlty on who is right between internalists and externalists)?


  1. Certainly one could be a realist and an internalist, but I think both realists and externalists are partly motivated by skeptical worries, and in similar ways. The realist wants to say that not just any old theory would do, that our current theory is special, and in the face of difficulties explaining exactly what's special about it the realist appeals to its match with how the world really is. The externalist wants to say we really know what the world is like, and in the face of inadequate evidence says that it doesn't matter what we can tell about the evidence; it's fine as long as it really comes from the world. In both cases there seems to be confidence that the connection to the world which can't be explained is all perfectly all right and not something to worry about.

    Myself, I sort of share the "don't worry" inclination, but as a non-realist and not really a fan of many aspects of many versions of externalism I tend to think that if we're not worrying about it we should really not worry about it, rather than telling ourselves everything's fine; why do we need to tell ourselves that if we're not worrying?

  2. A famous argument in Arthur Fine's NOA seems to appeal to internalist scruples.

    In particular, Fine picks up on 'Hilbert's maxim,' that arguments about a theory must be 'more stringent' than arguments within the theory itself. In Hilbert's case, this meant that an argument for the consistency of everyday mathematics had to be carried out using purely constructivist/finitist methods. However (Fine urges), the same maxim applies to arguments for a realist interpretation of science: an argument for (or against) realism should meet a more stringent standard than those arguments that appear in scientific practice.

    And, of course, Fine charges that abductive arguments for realism fail to meet this challenge: if the abductive inference (e.g., to the existence of unobservable entities) is what's in question, then an argument for realism is trivialized if it relies on the same abductive inference strategy.

  3. I had forgotten about the Hilbert's maxim stuff. It seems odd that Fine invokes it, since it failed in its purpose for Hilbert. The incompleteness theorem guarantees that in order to prove consistency and settle some of the disputes in question, one must use a stronger language, not one with at best fewer resources. There isn't a general result of a similar kind that would give one cause to reject Hilbert's maxim in this case, but it seems weird that Fine would blithely invoke it.

  4. Don't forget that van Fraassen used a voluntarist epistemology to defend his anti-realism. I believe that he converted to voluntarism in the early 80's, shortly after writing The Scientific Image . It appears, e.g., in his responses in the Churchland and Hooker volume from 1985. I conjecture, but do not know, that his conversion to voluntarism coincided closely with his conversion to Christianity.

  5. Well, not exactly blithely... Fine does acknowledge that Hilbert's program failed. But this failure didn't have anything to do with Hilbert's maxim -- rather, the failure was in the assumption that there exists a proof of the consistency of arithmetic.

    Hilbert's maxim, on the other hand, still seems eminently reasonable -- it's basically a plea to avoid circularity.

  6. I agree with Bryan that Hilbert's maxim is "eminently reasonable", and that the incompleteness of arithmetic does not affect it. I may have missed the point, but is not clear to me, though, why this maxim is hard to satisfy.

    Consider the proposition that if an observed fact O is implied by T, then the likelihood of T given O is at least as high as the likelihood of T taken by itself. I take it that some abductivists cite this proposition in defence of realism. Well: isn't the proposition, being a priori, more credible than the empirical truths of science?

    Of course, it might be said that abductivism is committed to a lot more than the above proposition, and that the argument about Hilbert's maxim is therefore more complicated than I am making out.

    I accept that. My point, however, is that the comparison implied by Hilbert is between the first-order propositions of the target theory and certain second order evaluative propositions applied to the former. And I am not sure how you can get a general anti-realist argument out of that comparison.

  7. There are proofs of the consistency of PA. They happen to require transfinite induction to epsilon nought to carry out, but they are still somewhat informative. They can't be given in PA. They also aren't completely circular. The surprise about Fine invoking Hilbert's maxim (maybe just in calling it such instead of "anti-circularity principle" or some such) is that in the case Hilbert was concerned with, arithmetic, one cannot get more stringency (in the sense of weaker assumptions) in arguments about a theory than one has in that theory and get anywhere. The maxim might be better applied to other parts of science, which is there is that little hedge in my previous comment. Looking at the discussion, I agree with Bryan's assessment of the realist appealing to abduction, so, it seems to me like there might be something at issue with stringency and its relation to circularity.

  8. Well, you certainly don't want to be in Achilles' position (as in Lewis Carroll's homily) with regard to abduction. That is, you don't want to introduce abduction as a premise in an argument from abduction. But I took it that scientists don't do this.

    So the question is: If E is the evidence that justifies theory T in a science like physics, can a philosopher observe that this is justified by principle P (where abduction would be an example). Would this observation automatically be less stringent than the original thesis?

  9. Aaron,

    The analogy you suggest is interesting and I suspect that this is partly what has drawn (some) scientific realists to externalism. Yet wouldn't they their life be easier if they turned to, say, evidentialism? Couldn't they say something along the lines of S's belief that T is (approximately) true is epistemically justified because that belief fits S's total evidence? (Ok, I'm already hearing some of the standard objections against evidentialism, but let's put the epistemological debate aside).


    Could you expand on why you see Fine's argument to appeal to internalist scruples? I'm not sure I see exactly why you think that is the case.


    I'm not sure if I get the point you were trying to make. Is it that epistemic justification is largely internal for voluntarists? (S's belief that p is epistemically unjustified if and only if it is irrational for S to believe that p and whether it is irrational for S to believe that p is determined by S's epistemic state?)

    Aside to Shawn,

    'There are proofs of the consistency of PA'. Of course there are, there are at least as many as there are inconsistent theories, and as far as I can see this is the rationale for Hilbert's maxim. (Whether the fact that a theory is weaker than PA is evidence that is is consistent it's a different issue).

  10. What about the distinction between rule-circularity and premise-circularity? Clearly you don't want 'abduction is reliable' to be a premise in an argument whose conclusion is that abduction is reliable, but it seems to many of us that it's not such a problem to use abduction as a rule of inference in an argument that abduction is reliable.

    Compare Dummett on 'the justification of deduction' - even an argument for the conclusion that deduction is reliable has to use deduction along the way. Rule-circularity in the justification of inference rules seems to be something we have to live with, and I don't see that it undermines abduction.

  11. About proofs of the consistency of Peano Arithmetic -- Gentzen provided one, though it involves transfinite induction and a richer ontology than first-order arithmetic. Also Kreisel argued that one can show consistency with "informal rigour". But these are precisely the cases in which Hilbert's maxim has been invoked: the claim is that the assumptions are subject to at least the same level of prior doubt as the conclusion being proved.

  12. On proofs of consistency (Shawn on Feb. 8, Gabriele on Feb. 9): The gist of Gabriele's aside seems to be that Gentzen's proof of the consistency of PA was trivial or somehow involved an inconsistent theory. This is not the case:

    More generally, with both the search for proofs of consistency and arguments for limited forms of scientific realism, I do not think the goal is to refute skepticism or somehow get something for nothing. It should rather be to clarify how our mathematical or scientific methods work and what their scope is. If transfinite induction or the IBE needed for scientific realism is not different in kind from what we find in ordinary math and science, then this is an illuminating discovery.

  13. Chris,

    That was not the point I was trying to make, which was that part of the motivation for Hilbert's maxim is that he wanted the proof of consistency to be in a theory which was more likely to be consistent than the theory whose consistency is supposed to prove and that seem to be a pretty reasonable request on any proof of consistency, isn't it?

    More relevantly, I don't find the discovery that IBE is used ordinarily in science illuminating, for I don't see what the NMA adds to specific IBEs about specific empirically successful theories if you already have those in place. What I think *is* illuminating is that the scientific anti-realist can hardly defend his/her case against (the realist construal of) IBE without falling prey to external world skepticism. In other words, what I find "illuminating" is the discovery that IBE is ordinarily used by common people and that the scientific anti-realist's refusal to adopt it makes his/her position at risk of becoming epistemologically unstable, always tethering on the brink of collapsing into radical skepticism.