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.

340<br />

new appendices<br />

able to show that, in the elementary theory, S is a (denumerable) Boolean<br />

algebra. (For another example see appendix *vi, point 15.)<br />

In Postulate 2, A1 is needed to establish that not all probabilities are equal<br />

(say, equal to 0 or equal to 1). The function of A2 is to allow us to<br />

prove ‘p(x, a) = p(x, b)’ for all elements a and b whose probabilities,<br />

given any condition c, are equal. This can be done without A2, but only<br />

under the assumption p(a) ≠ 0 ≠ p(b). Thus A2 is to enable us to extend<br />

the probabilistic equivalence of a and b to the second argument even in<br />

those cases in which a and b have zero absolute probability.<br />

A2 may be replaced by the following stronger formula:<br />

A2 +<br />

If p(a, a) = p(b, c) = p(c, b), then p(a, b) = p(a, c);<br />

or by either of the formulae (the weaker of which is B3 − ):<br />

B3<br />

B3 −<br />

If p(ab, c) = p(ba, c), then p(c, ab) = p(c, ba).<br />

If p(ab, ac) = p(ab, c), then p(ba, ca) = p(ba, c).<br />

Obviously, it can therefore also be replaced by the formula (which is<br />

simpler but much stronger):<br />

B3 +<br />

p(a, bc) = p(a, cb).<br />

But since B3 + is stronger than necessary—in fact, p(a,(bc)(cb)) =<br />

p(a, (cb)(bc)), though weaker, would suffice—it is a little misleading: its<br />

adoption would veil the fact that with the help of the other axioms<br />

alone, the law of commutation can be proved for the first argument.<br />

A2 + is preferable to the other formulae here mentioned in so far as it<br />

avoids (like the much weaker A2) using the product of a and b.<br />

However, we can make use of the facts here stated in order to reduce<br />

the number of our axioms to three, viz. A1, C + , and the following<br />

axiom B which combines B3 + with B + :<br />

B<br />

If p(ab, c) ≠ p(a, d)p(b, c) provided that p(a, c) � p(a, d)p(b, c) and<br />

p(a, d) = p(a, bc), then p(a, cb) ≠ p(a, d).<br />

Apart from being stronger than one might wish it to be, this system<br />

of only three axioms has all the advantages of the system of four<br />

axioms A1, A2, B + , and C + .

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

Saved successfully!

Ooh no, something went wrong!