Tuesday, February 13, 2007

Occam’s Razor

I was reading a debate between Apolonio Latar and an atheist. They were discussing the question, “Is it Rational to Believe in God?” Apolonio does a fine job of making the case that it is rational to believe in God. The atheist made a comment that I would like to address. The atheist makes the following remark...

I think the answer is the failure to properly appreciate the necessity of Occam’s Razor. The mere fact that God is not necessary is, in and of itself, all the rational defeater one needs.

I will not be addressing the claim that it is a “fact” that God is not necessary. I want to focus instead on the idea that Occam’s Razor can be a decisive method in disproving God. To do this I will quote a online discussion on the topic outside the context of proving/disproving God.

(Chad Orzel) Occam’s razor is not science. The scientific method does not consist of looking at the world, constructing multiple theories which might explain the observed facts, and then choosing the simplest.

Occam’s razor is meta-science at best, and not a decisive argument in any way. Sometimes, the more complicated theory is the right one.
(Tommaso Dorigo) Well, well, well, while I usually concur with Chad’s opinions, I am not sure I agree with him on this one. Occam’s razor does not consist in constructing multiple theories, but rather to advise succintness in construction of a scientific theory.
(Chad Orzel)I wouldn't say that Occam's razor is unscientific, but it is not by any stretch a decisive principle of science. It's like mathematical elegance-- a hint that you're on the right track, but not conclusive evidence of the rightness of the theory.

If the only tool you have to distinguish between two theories is that one is simpler than the other (or that one is more mathematically elegant than the other) then you need better experiments.

(There's an anecdote about some famous physicist-- for some reason, I think it was Bethe, but it might've been Feynman or Fermi or one of those people-- who was presented with a bunch of experiments that some people were claiming as evidence of a new particle discovery. Each time they showed him a bubble chamber picture that they said showed their new particle, he had an alternate explanation-- stray fields, particles not detected in the picture, all sorts of things.

(Finally, in frustration, they said "Look, we've shown you a dozen pictures, and you've given a dozen different explanations of the track. We have a single, simple theory that explains all of them."

("Yes," was the reply, "and the difference is that each of my dozen different explanations is right, and your single simple explanation is wrong." And he was right.)
(Tommaso Dorigo) I know Feynman's story Chad, but it proves my point - entia non sunt multiplicanda: it is the particle they were claiming which was what Occam's razor would have slashed, and it was precisely by applying Occam's principle that Feynman ended up being right: his explanation were more economical, since they relied on known effects.
(Chad Orzel)This is a nice illustration of the reasonwhy Occam's razor is not a decisive scientific principle, though: Prior to the experimental confirmation, you could perfectly well argue that either side had the more economical explanation. The people doing the experiment thought that a single particle was a simpler experiment, while Feynman (was it Feynman? I can't recall the source) thought that using known effects was simpler.

In hindsight, we can easily say that the correct explanation was also the simpler one, but that's because we know it was correct. Relative simplicity is very much in the eye of the beholder, though, and does not provide a useful basis for determining which theory is correct.

I don't object to Occam's razor as a sort of useful heuristic, but it's not decisive. You can say of two theories "Well, this one seems simpler, so I think it's on the right track," but that doesn't prove anything. It might be a good way to determine which of two expensive and complicated sets of experiments would be better to undertake (the simpler theory stands a better chance of being confirmed), but experiments are the only real way to distinguish between theories. (LINK)

No comments: