Friday, February 20, 2015

Alan Musgrave on the Success of Sciences as an Argument for Realism

The Popperian Critical Rationalist philosopher Alan Musgrave points to the incredible success of the natural sciences in explaining the world, predicting the behaviour of the natural world, and in designing technologies that work so well. Why is this so? In essence, it is likely to be the result of science getting to be truth, or in some cases close to the truth, about objective reality.

Musgrave lays out the “miracle” argument for scientific realism:
Musgrave, Alan. 2006–2007. “The ‘Miracle Argument’ For Scientific Realism,” The Rutherford Journal 2
http://rutherfordjournal.org/article020108.html
First, Musgrave sets out a Popperian version of inference to the best explanation, as the form of reasoning we must use to explain why the natural sciences are so successful.

The inference that it is rational to accept that there is an objective reality with objective truths follows via inference to the best explanation from the overwhelming reality that some scientific theories are extremely successful at explaining the world and doing things while some are worthless or almost completely worthless.

In essence, there if there wasn’t an objective reality with objective truths, the astonishing success of science would be (bizarrely) wholly accidental or some kind of miracle.

14 comments:

  1. It's a ridiculous argument and flies in the face of all the progress made in philosophy even among (ex)-Popperians who have come to realise that the humanities and the natural sciences are entirely different fields. I might as well say that since Bayesian probability theory works well in determining whether a person has HIV then it must also work well to determine the guilt or innocence of a murder suspect. (BTW some would be authoritarians wearing lab coats do argue just this...).

    https://fixingtheeconomists.wordpress.com/2013/09/08/probabilities-keynesian-legal-versus-bayesian-mathematical/

    This is, at base, an authoritarian, anti-democratic argument. It is endorsed by over-zealous technocrats who would gladly dismantle the best of our social institution. The new authoritarianism will not come, pace Popper, from those influenced by continental philosophy; but rather from those thinking that they can restructure our best social institutions in line with Science. As Foucault has demonstrated: this began a long time ago and it is only getting worse and worse; it is now a huge part of neoliberal discourse.

    ReplyDelete
    Replies
    1. I am not sure why you drag social sciences into this, as if you refuted Musgrave's argument.

      The argument is about the incredible success of the natural sciences.

      If you refuse to accept the idea that there is an objective reality with objective truths which natural science studies, then you need an explanation of why natural science is so very successful. It would appear that, if no objective reality exists (whether the reality is a pure mental idealist world or an external materialist world that is the causal origin of human sensation), the success of science becomes:

      (1) a pure accident, or

      (2) some kind of miracle.

      Delete
    2. You are again confusing the idea that positing an objective reality is synonymous with assuming that we derive 'objective truths' about this reality. One simply does not follow from the other.

      Statement 1: The earth exists and I live on it.

      Statement 2: Therefore I can pick up the earth.

      That would be somewhat equivalent to your assumption.

      In fact, it is becoming increasingly clear that natural science doesn't deal with 'objective truths' but merely what works well. Example: Newtonian physics 'works' pretty well but is now not though to be 'objectively true'. There is every likelihood that Einstenian physics 'work well' for new problems encountered but will seen not to be 'objectively true' in 300 years.

      Delete
  2. The NMA argument is a dreadful argument for trying to shore up scientific realism. In short, it commits a 'base-rate fallacy,' and is unable to account for actual scientific practice, in which many now-known-to-be-false theories were, for a time, perfectly serviceable and important theories which undergirded scientific progress.
    Even Stathis Psillos's qualified attempt to rescue something out of the NMA has recently been (IMO) successfully demolished by Colin Howson. See the latter's 'Exhuming the No-Miracles Argument,' Analysis, Vol. 73, No. 2, April 2013, pp. 205-211.
    I have no truck with post-modernism (it's mostly nonsense with some honourable exceptions). But scientific realism is mostly untenable as well, except in some very moderate and restricted form, as with the excellent Hasok Chang.

    ReplyDelete
    Replies
    1. Am I to take it that you reject the idea of objective truth and an objective reality?

      "unable to account for actual scientific practice, in which many now-known-to-be-false theories were, for a time, perfectly serviceable and important theories which undergirded scientific progress. "

      It can take account of these things perfectly well. E.g., by looking at how science gets corrupted by religion (think creation science) or other forms of irrationalism, and how earlier theories were a reasonable approximation of truth but needed revision, or how they only seemed to be true before much better knowledge was available.

      As for Colin Howson, "Exhuming the No-Miracles Argument," Analysis (2013) 73.2: 205-211, I see no refutation of Alan Musgrave's Popperian version of the "no miracles" argument. Howson's critique is directed against an inductive version by S. Psillos.

      Worse still, Colin Howson is apparently a subjective Bayesian, a theory deeply flawed.

      Delete
    2. He's talking about, say, Newtonianism being perfectly serviceable despite the fact that it is today recognised as not being True.

      This has nothing to do with science being "corrupted by religion" (bugbear, much?). It has to do with the nature of science itself.

      Most studies of the practice of science show that it operates along pragmatic lines. Not along the lines of Absolute Truth. It operates on the basis of Absolute Truth only in the minds of those that utilise a discourse based on scientific discourse to further their truth-claims in the humanities; i.e. scientism-ists.

      http://en.wikipedia.org/wiki/Scientism

      Delete
    3. I have no opinion on the concept of 'objective truth' nor 'reality' as it relates to scientific practice, for neither of these metaphysical ideas is needed to make sense of the shape of scientific knowledge. If you ask me do I 'believe in reality,' then certainly so, for whatever that's worth. If you ask me is such a belief necessary to be a scientist, then I would answer in the negative, since IMO good science aims at useful prediction and not the discovery of 'truth' per se (hint: I'm a constructive empiricist when it comes to Phil of Sci).
      Accept or reject that as you may, your idea that 'approximate truth' can be used as a way to corroborate the NMA was shown long ago to be false. For someone who claims a Popperian critical rationalist stance on science you seem ignorant of the critiques of Popper's account of verisimilitude, approximate truth, and truth-likeness, set forth in his Conjectures and Refutations in 1963. Two of Popper's very own colleagues, David Miller and Pavel Tichy, each gave rigorous formal proofs that Popper's ideas were flawed.
      See Miller, D - 'Popper's Qualitative Theory of Verisimilitude,' in The British Journal for the Philosophy of Science, 25, pp. 166-77.
      Tichy, P - 'On Popper's definitions of Verisimilitude,' The British Journal for the Philosophy of Science, 25, pp. 155-160.
      Neither of these papers was ever answered by Popper, or his allies.
      After Popper, other accounts of 'approximate truth' were attempted in the form of 'possible worlds' and 'type hierarchy' approaches. The notion of 'similarity relations' took centre stage. However, these ideas were also undermined by Miller, in his 1976 paper 'Verisimilitude Redeflated,' British Journal for the Philosophy of Science, 27, pp. 363-380.
      In a review article some years later, Peter Urbach summed up the shortcomings of the whole idea of 'approximate truth' as a grounding for scientific progress.
      See Urbach, P - 'Intimations of Similarity: The shake basis of verisimilitude,' The British Journal for the Philosophy of Science, 34, pp. 266-275.

      Modern scientific realists like Niiniluoto admit that truthlikeness is not a purely semantic idea, but also has a 'pragmatic' dimension, and that 'the concept of similarity is practically ambiguous.' (Niiniluoto, Critical Scientific Realism, 1999, p. 77).
      Psillos himself admitted around the same time that 'according to the 'possible worlds' approach, truth-likeness turns out to possess odd features such as dependence on the number of states of the world. True propositions end up having the same verisimilitude as false ones.' (Psillos, Scientific Realism: How Science Tracks Truth, 1999, p. 268.)
      He also says: 'I cannot see how the type-hierarchical approach to scientific theories can offer a realist account of verisimilitude, and of truth (if truth is a limiting case of verisimilitude.) (ibid. p. 273).
      More recently Anjan Chakravartty (another prominent realist and fine philosopher) has concluded in reference to these types of arguments that 'The difficulties reviewed and suggested here for extant account are serious, and in some cases fatal.' (Chakravartty, A Metaphysics for Scientific Realism, 2007, p. 218).
      Ironically, Psillos's proposal that I mentioned above was motivated precisely because of the shortcomings of the Popperian approach that you appear to think is unimpeachable. As for your swipe at Howson's methods, I look forward to your refutation of his argument in a forthcoming paper. As for your confidence in Popper's ideas, it is sorely misplaced. In fact, you might say it is a perfect example of an irrational and unscientific stance which appears to contradict inference to the best explanation...

      Delete
    4. "Newtonianism being perfectly serviceable despite the fact that it is today recognised as not being True."

      That is not correct, Phil.

      Newtonian mechanics as a theory that has a good claim to be true or a very good approximation of the truth about reality has been merely refined and limited to a certain domain by modern science: the macroscopic world. The macroworld is a higher level world and an emergent phenomenon from the lower-level quantum world.

      There are, moreover, many theories in science that have never been refuted or falsified, but merely developed and refined, e.g., heliocentric theory of the solar system, Darwinian evolution, etc.

      "Most studies of the practice of science show that it operates along pragmatic lines. Not along the lines of Absolute Truth."

      A very strange claim given that science is an empirical discipline and as defended by empiricists never makes claims to 100% truth: its truth claims are always probabilistic and open to testing and falsification.

      But none of this refutes the inductive argument that many of its theories are very probably true or close to the truth about an objective reality.

      Delete
    5. "Newtonian mechanics as a theory that has a good claim to be true or a very good approximation of the truth about reality has been merely refined and limited to a certain domain by modern science: the macroscopic world."

      Simply untrue with regards to it being an extension. The two are as distinctly different as, say, idealism and materialism in philosophy. In Newton, for example, gravity is a constant. Whereas in Einstein it is no such thing. Instead the speed of light is the constant.

      The theories are fundamentally different at a structural level. To give a simple analogy from economics take the quantity theory versus endogenous money: in the QTM velocity is held fixed as is output; but in endogenous money all the variables are allowed to fluctuate and (generally speaking) the wage-level is said to determine the price-level.

      To say that endogenous money is a 'refined' or 'limited' version of the QTM is misleading in the extreme. Same thing can be said for Newton vs. Einstein. If you disagree with this, of course, please feel free to remove your 'refutations' of the QTM on this very blog.

      "A very strange claim given that science is an empirical discipline and as defended by empiricists never makes claims to 100% truth..."

      Really? Newton's laws were never thought to be 100% true? I would imagine most people did indeed think this.

      More importantly, however, you seem now to have shifted from science discovering 'objective truths' about 'reality' to science constructing probabilistic estimates that seem to fit with reality. That is a far weaker claim. Probably somewhat closer to the truth -- if I may use that expression with a dash of irony -- but certainly far from the stronger claims made above.

      "But none of this refutes the inductive argument that many of its theories are very probably true or close to the truth about an objective reality."

      That is not an inductive argument. The inductive argument would be that many of the theories appear to explain certain aspects of reality to some, probably limited, extent and may well be completely overthrown in the future.

      Delete
    6. "More importantly, however, you seem now to have shifted from science discovering 'objective truths' about 'reality' to science constructing probabilistic estimates that seem to fit with reality. "

      There is no change in my view:

      (1) we have good reason to think that it is highly probable that science discovers objective truths about objective reality, and

      (2) any particular scientific theory as described by its propositions is always an empirical truth known by empirical evidence and inductive argument.
      -------------
      I have been saying this from the very beginning. If you did not realise this, you have been labouring under a misapprehension.

      And, yes, you can make an inductive argument to support this.

      I've just given you one: the argument that if our best scientific theories weren't objectively true and describing an objective reality, their success would be sheer accident or a miracle.

      Incidentally, you seem to think that just because all scientific theories are open to revision in principle, this means they all will in the future.

      There is no rational reason to think that the heliocentric theory of solar systems or the germ theory of disease will ever overthrown -- the evidence for them is just too strong.

      Delete
  3. "For someone who claims a Popperian critical rationalist stance on science "

    I am not personally endorsing it, merely linking to one Popperian critical rationalist defence of scientific realism.

    And you have not refuted Musgrave argument either.

    You cite a paper (Colin Howson, "Exhuming the No-Miracles Argument," Analysis (2013) 73.2: 205-211) that is an attack on an inductive defence of the "no miracles" argument, obsessing over subjective Bayesian priors, when Musgrave's argument is framed as deductive argument and its notion of probability only needs to be epistemic.

    ReplyDelete
  4. "As for your swipe at Howson's methods, I look forward to your refutation of his argument in a forthcoming paper"

    Oh, I don't need publish any paper.

    Perhaps you are unaware that there is already a rich literature exposing the flaws and untenable nature of subjective Bayesianism:

    https://larspsyll.wordpress.com/2014/07/23/bayesianism-preposterous-mumbo-jumbo/

    https://larspsyll.wordpress.com/2014/07/22/bayesianism-a-dangerous-religion-that-harms-science/

    ReplyDelete
    Replies
    1. "I am not personally endorsing it, merely linking to one Popperian critical rationalist defence of scientific realism."

      Sorry, I have no idea what that's supposed to add to the debate.

      "And you have not refuted Musgrave argument either.
      You cite a paper (Colin Howson, "Exhuming the No-Miracles Argument," Analysis (2013) 73.2: 205-211) that is an attack on an inductive defence of the "no miracles" argument, obsessing over subjective Bayesian priors, when Musgrave's argument is framed as deductive argument and its notion of probability only needs to be epistemic."

      Um no. I've cited about half a century of scholarship on NMA (and I can cite tons more), a good chunk of it by scientific realists. You want to obsess over Howson and Bayesianism to deflect from that, then that's your business. I don't need to 'refute Musgrave,' the refutations are in the literature I gave you, as well as the work of van Fraassen, Laudan, K. Brad Wray, PK Stanford, and many others.
      You seem not to understand the chronology of the NMA as it has developed over time. The Bayesian turn against it is very recent, and if you don't like it then fine but it won't help you in supporting Musgrave-style arguments. The papers I cited by Miller and Tichy are devastating against precisely the type of deductive argument that Musgrave tries to make. It is not me that needs to 'show how Musgrave is wrong,' he is an outlier in the Phil of Sci community in this sense. Alexander Bird (a metaphysical realist) has even gone so far as to say that the NMA is a bit of an embarrassment for realists, and should be dropped. There are plenty of other more interesting arguments.
      It's rather notable that the NMA was developed by Hilary Putnam, who then went on to drop it in favour of his internal realism. Putnam is a fine philosopher, and even finer for showing himself to be capable of admitting error.
      The main and most simple reason that the NMA fails is because all NMA arguments have a heavy reliance on the notion of 'approximate truth.' If you actually bother to look at the papers I cited you'll see that this notion is useless as a guide to the pattern of scientific knowledge as we historically know it. There are fairly simple philosophical reasons for this - theory underdetermination, which means that for any empirically adequate theory (in van Fraassen's terms), there may be other unconceived alternatives. NMA cannot decide which of these is the bearer of 'approximate truth,' and so it fails as a support for realism, either axiological or ontological.
      I'll pass over the ridiculous idea that two blog posts constitutes 'a rich literature' against Bayesianism. Especially since Bayesian arguments against NMA are actually just a small subset of a much larger literature, as I've said earlier. In other words, refuting NMA doesn't require Bayesian methods, and never did.

      Delete
    2. So it appears you now admit that Colin Howson ("Exhuming the No-Miracles Argument," Analysis (2013) 73.2: 205-211) does not refute Musgrave.

      "It is not me that needs to 'show how Musgrave is wrong,' he is an outlier in the Phil of Sci community in this sense. "

      That is an appeal to invalid authority fallacy. It does not matter if Musgrave is in a minority: his arguments need to be refuted, not attacked and dismissed by some informal logical fallacy.

      "NMA cannot decide which of these is the bearer of 'approximate truth,' and so it fails as a support for realism,"

      NWA isn't meant to decide: scientists or rational human beings do.

      As for "approximate truth", the concept is valid: e.g., "It is about 4 o’clock in the afternoon" (when it is 3.58 pm) or "John is about 6 feet tall" (when John is actually 6 feet half an inch tall) are clear, valid examples of approximate truth.

      Delete