[chezlaw]

[ QUOTE ]
[ QUOTE ]
There are no cases where one of two theories undecidable by normal evidence have been shown to be true, so if you apply meta-evidence you have no normal evidence that the meta-evidence approach is valid.

[/ QUOTE ]
I don't know what you mean by "normal" evidence, but there have been lots of examples of situations where competing theories existed, no experiments had been done to figure out which (if either) was right, and then later somebody came up with an experiment.

The general rule in those situations is that the simpler theory turns out to be the one that is correct. From that we can infer that, in general, simpler theories do better than more complicated theories.

This is not coincidence, by the way. The reason that more complicated theories often fare worse is that the traditional way of making a theory more complicated is to add more stuff onto it. There are more ways to add incorrect stuff than to add correct stuff -- so the more you add, the more likely you are to get some of it wrong.

This is why "the earth travels around the sun" is more likely to be correct than "the earth travels around the sun because it is being pushed by invisible angels." Whenever you add extra stuff on, you are increasing your exposure to potential error.

[/ QUOTE ]
I agree with all of that but its different. I'm talking about two theories that make no different predictions, not ones that cannot yet be tested.

This is the key I think. Rational belief on the basis of evidence is only meaningful if there is in principle some way of deciding which theory is correct.

chez