25.01.2013 Views

popper-logic-scientific-discovery

popper-logic-scientific-discovery

popper-logic-scientific-discovery

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

400<br />

new appendices<br />

thus becomes something like formal simplicity; but otherwise my<br />

theory of simplicity agrees with that of Wrinch and Jeffreys in this<br />

point.) They also saw clearly that simplicity is one of the things aimed<br />

at by scientists—that these prefer a simpler theory to a more complicated<br />

one, and that they therefore try the simplest theories first. On<br />

all these points, Wrinch and Jeffreys were right. They were also right in<br />

believing that there are comparatively few simple theories, and many<br />

complex ones whose numbers increase with the number of their<br />

parameters.<br />

This last fact may have led them to believe that the complex theories<br />

were the less probable ones (since the available probability was somehow<br />

to be divided among the various theories). And since they also<br />

assumed that a high degree of probability was indicative of a high<br />

degree of knowledge and therefore was one of the scientist’s aims, they<br />

may have thought that it was intuitively evident that the simpler (and<br />

therefore more desirable) theory was to be identified with the more<br />

probable (and therefore more desirable) theory; for otherwise the aims<br />

of the scientist would become inconsistent. Thus the simplicity postulate<br />

appeared to be necessary on intuitive grounds and therefore a<br />

fortiori consistent.<br />

But once we realize that the scientist does not and cannot aim at a<br />

high degree of probability, and that the opposite impression is due to<br />

mistaking the intuitive idea of probability for another intuitive idea<br />

(here labelled ‘degree of corroboration’), 10 it will also become clear to<br />

us that simplicity, or paucity of parameters, is linked with, and tends to<br />

increase with, improbability rather than probability. And so it will also<br />

10 It is shown in point 8 of my ‘Third Note’, reprinted in appendix *ix, that if h is a<br />

statistical hypothesis asserting ‘p(a,b) = 1’, then after n severe tests passed by the hypothesis<br />

h, its degree of corroboration will be n/(n + 2) = 1 − (2/(n + 2)). There is a striking<br />

similarity between this formula and Laplace’s ‘rule of succession’ according to which the<br />

probability that h will pass its next test is (n + 1)/(n + 2) = 1 − (1/(n + 2)). The numerical<br />

similarity of these results, together with the failure to distinguish between probability<br />

and corroboration, may explain why Laplace’s and similar results were intuitively<br />

felt to be satisfactory. I believe Laplace’s result to be mistaken because I believe that his<br />

assumptions (I have in mind what I call the ‘Laplacean distribution’) do not apply to the<br />

cases he has in mind; but these assumptions apply to other cases; they allow us to<br />

estimate the absolute probability of a report on a statistical sample. Cf. my ‘Third Note’<br />

(appendix *ix).

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!