Occam's razor: is it valid?

Occam's razor is applied to a situation in which two competing theories both account for all observable data; it eliminates the more complex one in favor of the more simple. For example, dualism must be discarded in favor of monism, because they both can fully account for existence.

Ontologically, this tool is a nightmare. It is kind of a mute acceptance that there IS NO WAY to know whether one theory is more true than the other.

Occam's razor is also an exercise in pragmatism. Recall that pragmatism is a method that examines the practical consequences of assuming an idea to be true. Since ontologically there can never be certainty, we should consider the fact that people need a systematic way to understand the world. Therefore it would practically be of use to them if we were to eliminate needlessly complicated theories.

Unfortunately it is a tool that must be applied when applicable. That's always a question of debate, however: when is it applicable? When two or more theories account for ALL observable data. If one theory BETTER ACCOUNTS for observable data than another, then that theory is obviously preferable; Occam's razor need not be applied.

I was wondering; what if two theories successfully account for 90% of the data? Obviously something's off, and you should probably do more observations until you can get either really close to 100% or change one of the theories so that it can account for the remaining 10%.

This tool is both ontological and epistemological, because it can apply to theories about the process of attaining knowledge as well as to theories about the world as we believe it to exist.

I think I ought to give an example of using Occam's razor in an epistemological situation, just to cement the title of my blog a little bit more.

The two main theories for how knowledge is attained are:
1) empiricism (knowledge derived from sense experience, associated with "Blank Slate" and Locke)
2) innate idea (ideas present at birth, associated with idealism and Plato)

Unfortunately I don't think these ideas are mutually exclusive (nature and nurture, anyone?). Furthermore, it seems that there are instances for both that problematize them. The two theories must satisfyingly account for all observable data before Occam's razor can be applied.

Darn, that didn't work out. Oh well, at least I tried.