3

William of Ockham would, in the 14th century, scarcely have imagined how much Philosophy of science could change. Then when Bacon gave us Inductivism in the 17th century, an immediate synergy can be seen. Inductivism with its bent towards verification, would be well served by the razor supplying the easiest to verify theory.

After Logical positivism, and especially the influence of Popper, we have today the general practice in Science to select theories on the basis of being (most easily) falsifiable. Yet Occam's razor is still held up as the gold standard to use for selecting theories.

Question: Have anyone commentated on this apparently seamless, but diametrically changed application?

christo183
  • 2,447
  • 4
  • 13
  • 33
  • See similar post. – Mauro ALLEGRANZA Nov 17 '18 at 13:09
  • Occam's razor is a very general "common sense" principle. There is no reason to think that it can be used in real life scienctific enterprise. – Mauro ALLEGRANZA Nov 17 '18 at 13:10
  • Your history of science and philosophy of science is too simple. The logical empiricists weren't advocates of falsificationism; that was Popper's central idea. Sometimes Popper is considered a logical empiricist, but often he isn't. At most he was one member of an intellectually diverse community. – Dan Hicks Nov 17 '18 at 15:49
  • 1
    We neither have today nor ever had the general practice of selecting theories on the basis of falsification. That was a theoretical fantasy that, upon scrutiny, did not work even for the field from which it was extracted, physics. For most, Quine's and Kuhn's critiques disposed of Popper just as they disposed of positivists. Subsequent structuralism presents a far more complex picture with parsimony being only one among many epistemic values that factor into selection of theories, see e.g. Kuhn's The Structure of Scientific Revolutions Revisited, or Nersessian's Creating Scientific Concepts. – Conifold Nov 17 '18 at 23:47
  • @Conifold See comments to Geoffrey Thomas below. I'm actually thinking of pulling this question, on the grounds of being trivial and uninteresting, based on commentary by you and Geoffrey. I'll give it a few days to see if someone has something useful to say. – christo183 Nov 18 '18 at 08:02

2 Answers2

4

So far as I can see, Ockham's Razor is simply a methodological rule, a principle of parsimony, that tells us not to assume more than we absolutely have to in order to explain something - an object, an event, a state of affairs or whatever.

Hence the old familiar, 'Ockham's razor shaved Plato's beard' - meaning that there was no need to assume the existence of Platonic Forms (eide or ideai) in order to explain the nature and behaviour of (say) objects in space and time.

The Razor is a plea for the removal of redundancy in explanation.

I can see a link with logical positivism which removed metaphysical entities, regarded by LPs as bogus, from science and everyday discourse. The LPs certainly, to put it mildly, saw no need to assume the existence of such entities.

It doesn't appear to follow that a theory or hypothesis which assumes the least necessary to explain something is the easiest to verify or falsify. Removing redundancies from a theory or hypothesis may still leave you with a collection of competing theories or hypotheses that are equally difficult to test. A theory purged of unnecessary assumptions does not have to have the 'least amount of assumptions' or be the 'least encumbered by variables to be accounted for' (christo). The other, competing theories may have more assumptions and be encumbered by more variables - only, they are not unnecessary.

Geoffrey Thomas
  • 35,647
  • 4
  • 43
  • 146
  • My conjecture here is that the Razor is used in the selection of theories, not the formulation or the testing of theories. Reason being that for the formulation of a theory should epistemologically only include parameters from the domain of interest, and conversely in experimental tests those same parameters should be the only concern, leaving only the choosing between theories as a workable application of the Razor... – christo183 Nov 17 '18 at 18:48
  • But as @MauroALLEGRANZA points out, the idea that it's applied as a general principal is tenuous. For all that the central tenet that the application of the Razor has changed from support of verification to falsification, without much commentary, seems intact. As far as selecting the "easiest to verify or falsify" theory, Occam's razor selects good theories incidentally, because the theory with the least amount of assumptions, is also least encumbered by variables to be accounted for. – christo183 Nov 17 '18 at 18:58
  • @christo. Thank you for your comments, acute and relevant as ever. I have amended my answer mainly in the spirit of Ockham by removing unnecessary statements. Best - Geoffrey – Geoffrey Thomas Nov 17 '18 at 20:24
  • Thanks for that perspective, it would seem selection of a theory for testing should be more involved than a naïve application of Occam's razor. Which makes me wonder, is this question philosophically interesting, or is it merely a trivial curiosity. – christo183 Nov 18 '18 at 07:49
  • @christo. Your question interested me, or I wouldn't have tackled it. I think, and remember I was a university lecturer for years, that you have a real gift for the unusual angle. You connect things that most others never think of putting together. In other words you have originality. Keeping going at philosophy and you will become a truly formidable figure. All the best - Geoffrey – Geoffrey Thomas Nov 18 '18 at 10:44
  • A valued comment. I couldn't stop if I tried. – christo183 Nov 19 '18 at 06:30
3

In machine learning and statistics Occam's razor has a precise mathematical interpretation: https://en.m.wikipedia.org/wiki/Occam_learning The basic idea is that by focusing on simpler explanations one is much less likely to overfit the data -- the phenomenon where one tries to explain the data with an exceedingly complex hypothesis that explains all the existing data but does not hold up on new data.

present
  • 2,500
  • 1
  • 9
  • 23
  • A very interesting addition! When we look at 'theories of action', it is quickly evident a strategy of discarding the simplest theory first is the most efficient, i.e. uses least resources, or yields the fastest solution, on average. If we accept: https://en.wikipedia.org/wiki/Theory-theory with the mechanism of Bayesian learning, we get a very close neuro-psychological parallel to Occam learning. You may say we are biologically predisposed to focusing on the simplest theory. – christo183 Nov 19 '18 at 06:28