On the Success of Scientific Theories

[watchamadoodle]

I wonder if this stuff would help?
http://en.wikipedia.org/wiki/Solomonoff%..._inference

Umm, maybe? It depends on what you think this is addressing in the OP. Or were you responding to the notion of approximate truth only?

I'm not quite sure how this fits in.

Say our current scientific knowledge is a set of rules.
Some new experimental results break those rules (with a high probability).
Scientists typically use Occam's razor to make the minimal change to the existing rules to explain the new experimental results.
Occam's razor has always bothered me, because it seems to be poorly defined.
Solomonoff's theory of inductive inference defines Occam's razor and shows why it works.
http://en.wikipedia.org/wiki/Solomonoff%..._inference

This is only based on a layman's reading of snippets in wikipedia. I was hoping some of you guys have the background in math to tell me if this idea makes sense? It seems like inductive reasoning, statistics, and Occam's razor are the essence of science? (Also I suppose another core idea of science is that the current state of nature predicts the future state of nature within the limitations of QM?)

There is also an updated version of Solomonoff's ideas

Hutter's notion of universal AI describes the optimal strategy of an agent that wants to maximise its future expected reward in some unknown dynamic environment, up to some fixed future horizon. This is the general reinforcement learning problem. Solomonoff/Hutter's only assumption is that the reactions of the environment in response to the agent's actions follow some unknown but computable probability distribution.

http://en.wikipedia.org/wiki/Marcus_Hutter