Friday, February 6, 2009

Scientific Realism and "Consilience"

This post might just be motivated by my ignorance of the topic, my apologies if this is the case - I just decided to express some vague intuitions that have been in my mind for some time.

The intuition is that a good argument for scientific realism would come from a consideration of the convergence of independent methods and theories towards the same hypothesis. There’s something like this, of course, in Whewell’s ‘consilience of inductions’, and in subsequent work on the robustness of inductive methods. However, the Whewellian line of argument, as far as I understand, is mainly concerned with the reliability of inductive practices, e.g., with how varying experiments corroborates the trustworthiness of their outcomes.

What I have in mind is, instead, some general argument that the convergence of totally independent experimental and/or theoretical procedures towards the same ‘result’ (i.e., the postulation of the same entity, or the same theoretical structure) is best explained by saying that those procedures discover truths (perhaps all this could be put in terms of probabilities, with the probability, say, of the reality of x raising as more and more independent reasons for positing x emerge). As far as I know, something in this sense was said by Salmon (if I remember well, he considered the fact that there are 12 completely different methods of measuring Avogadro’s number as a reason for realism about the latter). But I don’t know of any more systematic treatment.

One could think of it in terms of a ‘detective story’ analogy: in the same way in which independent signs all point to the same (unknown) culprit and - after a certain (clearly vague) threshold is reached - they are sufficient for (quasi-)certainty about the identity of the murder, so independent scientific hypotheses might work together. Of course the objects and the murderer are observables, while the problem with scientific realism concerns the unobservable, but the important point here is that coherence seems to be an indicator of truth, and that the anti-realist arguments in favour of a commitment to mere empirical adequacy seem correspondingly weakened. Here, existing work on coherence and truth (e.g., Bovens and Hartmann), as well as on unification as a theoretical virtue (e.g., Myrvold), becomes relevant but, again, I don’t know of anyone who developed this in the specific domain of scientific realism.

One could say that we simply don’t have evidence of such convergence except in rare cases, such as that considered by Salmon. But this is a historical point- perhaps open for discussion - and doesn’t contradict the conceptual possibility of an argument for realism based on convergence.

Another objection I anticipate is that the convergence towards the same entity is not surprising because we typically tend to use theoretical structure that already exists, maybe in some other field, to explain new problematic evidence. I don’t think this reply has force: consider an alien scientist coming to Earth and illustrating all alien theories to us. It could happen that these theories are very different from ours, and yet (some of them) are based on (some of) the same entities we postulate. Surely, there wouldn’t be a possible pragmatic source for the convergence there, and yet - I think - we would have a strengthened belief in the existence of the entities in question.

Can anyone tell me what is wrong with the idea? Or, if it is not obviously incorrect and was explored before, where I can find something? Even just hearing people’s opinions on this would be nice. (I realise that I have been ambiguous between entity and theory realism here, but I guess it is because the intuition can be developed in slightly different ways).


  1. I don't know. I've always suspected that the notion of "truth" in the Realism/Anti-Realism debate was a red herring. The real debate seems to be one over the level of committment one owes to science. Realists are like Baptists, whereas Anti-Realists are like Presbyterians. They believe in science just fine, as long as it gets the poor fed and the keeps people from robbing each other. Beyond that its anyone's guess. Realists find that lack of faith disturbing. The truth is right there in the book (or the lab?) and anyone who fails to start shaking and crying before its awsomeness should be viewed as a threat.

    (Apologies to Arthur Fine)

  2. the convergence of totally independent experimental and/or theoretical procedures towards the same ‘result’ (i.e., the postulation of the same entity, or the same theoretical structure)

    The problem here is that scientific models need never postulate the existence of "real" entities. Some models do, but it's quite straightforward to take a model which postulates real entities and reformulate it so as to expunge the realism without altering the model's empirical predictions.

  3. I wanted to add: The basic anti-realist position, as I understand it, is models and their theorectical entities merely describe patterns in our observations. They need not correspond to "real" objects in some sort of external reality. In this view "convergence" merely signifies that these patterns continue to hold for more and more observations.

  4. Matteo,

    I vaguely remember Michela (Massimi) suggesting something similar in a paper (I think she was talking of some theories being overdetermined by the evidence, which sounds somewhat like what you suggest), but I might be wrong and the best thing might be to ask Michela.

    I think you clearly identify one of the main objections to the view, but I am not sure your Alien Scientist reply carries much weight in absence of any evidence that Alien scientists (with a conceptual apparatus suitably different from ours) would in fact come up with the same theories (I sometimes use the alien scientist when discussing grue in class).

    Also, I don't think that one needs all this if s/he is an entity realist. If, on the other hand, one is an (epistemic) structural realist, then isn't this what s/he's already claiming (the main difference being that typically the ESR focuses on diachronic rather the synchronic examples of "consilience")?

  5. At first sight, there two questions here. The first is whether consilience of theories is an argument for realism.

    For example, is it an argument in favour of the reality of evolution that it plays an explanatory role in both biological taxonomy and functional physiology?

    The second is whether verification by different experimental procedures is an argument for realism. For example, is it an argument for the reality of evolution that it is confirmed by both the fossil record and biogeography?

    Perhaps, these reduce to the same thing, since successfully playing an explanatory role is a form of confirmation. But prima facie there is some difference.

  6. Even accepting that IBE is a truth-conducive inferential strategy -- and needless to say, many anti-realists will not accept this -- I don't see how one would argue that truth (or existence) would be the _best_ explanation in such a case of convergence.

    We know only that the procedures are independent, not that applying (any of) them reliably (e.g. with high objective probability) leads to any particular sort of result. So even if we allow that convergence requires explanation (and I don't see any plausible argument that it should), then why shouldn't we say that it happens as a result of ways that humans prefer to conceptualise things?

    And I would disagree that in your alien example: 'there wouldn't be a possible pragmatic source for the convergence'. Might the alien and human scientists (in virtue of their cognitive similarities) not prefer the same set (or similar sets) of theoretical virtues (such as Kuhn's simplicity, scope, consistency, accuracy, and fruitfulness)? And why aren't those _pragmatic_ virtues?

  7. Some of the comments seem to have missed the point I was trying to make. A typical realist argument is that the reality of an entity is the best explanation for a successful account of certain phenomena based on that entity. But this can (legitimately, in my view) be rejected as question begging by the antirealist/instrumentalist (and of course the realist claim is not built in the model/theory). But the convergence of totally independent procedures (c.f. Salmon on Avogadro´s number) seem to pose a challenge to the antirealist position, much like having 99 heads after 100 coin tosses makes the hypothesis that the coin is REALLY biased very plausible. Of course, the antirealists could just insist that the coincidence needs no explanation (what the realists call a miracle isn´t really a miracle), but intuitively their position seems weakened in this case (this was the point of my detective story analogy, I guess).

    Gabriele: I don´t think entity realism works as it stands, and I do beleive that ESR is too restrictive and 'too diachronic', so to speak. Hence, you identify exactly two points that might motivate an argument from convergence. (I will ask Michela about all this, next time I see her...).

    Darrell: the point in the alien science example was that even assuming that the same pragmatic virtues are preferred by both aliens and humans, they could come up with radically different theories postulating some of the same entities. Forget about how unlikely this is to be case: isn't the mere conceptual possibility sufficient to require an explanation of the convergence in existential commitments (implicit in the theory, not explicitly made by the scientist)? If, on the other hand, you claim that convergence on the same theoretical posits is forced by the choice of the right pragmatic virtues, then wouldn't it be likely that these virtues are themselves truth-conducive? After all, the only thing shared by the two communities are their methodological preferences PLUS the reality they live in...

  8. Hi Matteo,

    Just to address your last question; even if reality plus certain preferences 'forces' particular results, why assume that those results are _representative_ of reality (or some aspects thereof)?

    And even if the argument can somehow be made that they are representative, might it not then be the case that the form of representation is far weaker than that required by a full-bloodied scientific realist or entity realist?

    In short, I think that even if it's clear (A) that the agreement is a result of applying some method, and (B) that reality determines what the method yields, it still doesn't show that (C) reality determines that the method yields truths (or truth-like statements), or discloses fundamental entities.

  9. Hey Darrell,

    Well, I guess we have reached the basic point of disagreement between realists and antirealists. To my mind, if we agree that the method is the same and it leads to the same result in a way determined by the same reality, AND there are many independent ways of reaching the result, this supports the realist thesis. This form of coincidence (unlike the mere success of science) would indeed be a miracle. In your terms, (A) plus (B) plus the sort of overdetermination I am pointing at should be sufficient for (C). But we are clearly going round in circles here...

    Thank you for your comments, anyway!

  10. Mohan is right to distinguish the two arguments here. Salmon's argument is a version of the second, concerning the convergence of independent experimental results. Specifically, Salmon's argument applies in the case where the value of a particular physical quantity can be independently determined. Norton calls this “the method of overdetermination of constants”.

    Here are some references where the argument is discussed:

    Salmon, Wesley C. 1984. “The Common Cause Principle and Molecular Reality”, in Scientific Explanation and the Causal Structure of the World, Princeton University Press, Princeton, pp. 213–227.

    Achinstein, Peter. 2002. “Is There a Valid Experimental Argument for Scientific Realism?”, in The Journal of Philosophy, Vol. 99, No. 9, pp. 470–495. [JSTOR]

    Both are reprinted in:

    Achinstein, Peter. 2004. Science Rules: A Historical Introduction to Scientific Methods, Johns Hopkins University Press, Baltimore MD.


    Norton, John. 2000. “How We Know About Electrons”, in After Popper, Kuhn and Feyerabend: Recent Issues in Theories of Scientific Method, Robert Nola and Howard Sankey (Eds), Vol. 15, Australasian Studies in History and Philosophy of Science, Kluwer, Dordrecht, pp. 67–97. [PDF]

  11. Mohan,

    I think your distinction between "experimental" and "theoretical" consiliecne is important, but, if I understand Matteo's point correctly, he has in mind a third kind of consilience. Matteo is thinking of the consilience among logically incompatible theories (that share some ontological committments) while your "theoretical consilience" is among logically compatible theories.

  12. ... btw, as far as I can see, this is why Matteo's argument is not exactly the same as the standard realist IBE/NMA arguments as some of the comments above seem to assume.

  13. Another paper on this topic:

    Wimsatt, William C. 1981. “Robustness, Reliability, and Overdetermination”, in Scientific Inquiry and the Social Sciences: A Volume in Honor of Donald T. Campbell, Marilyn B. Brewer and Barry E. Collins (Eds), Jossey Bass, San Francisco, pp. 124–163.

    (Reprinted in his Re-Engineering Philosophy for Limited Beings: Piecewise Approximations to Reality, Harvard University Press, Cambridge MA, 2007, pp. 43–74).

  14. I'm concerned about the extrapolation of theories from their original contexts into alien domains. A good example is autopoiesis, which works fine in theoretical biology but has been seized on by some systems theorists and "applied" to sociology and economics. Commercial enterprises are uncritically analyzed as autopoietic systems. A fence around a property becomes a "membrane" and so on.

    I have much less of a problem with people in the harder sciences nosing out isomorphisms, if they do it responsibly. For example, Diederik Aerts, a quantum physicist at the VUB, is a maestro at that kind of thing. But too many people just muddy the waters. Reality isn't about metaphors, science isn't about metaphors.

  15. A brand new paper on Perrin's work on Avogradro's number and arguments for scientific realism:

    van Fraassen, B. C. (2009), "The perils of Perrin, in the hands of philosophers", Philosophical Studies, 143:5-24.

    This is van Fraassen's contribution to the proceedings of the 2008 Oberlin Colloquium in Philosophy, which was devoted to the philosophy of science. I'll post a general announcement about the volume separately.

  16. Peripherally related but amusing, a quote from Fodor in the LRB reviewing EOW:

    “If this is what consilience is like, I recommend the assorted antipasti.”