Occam's razor is a logical principle attributed to the mediaeval philosopher William of Occam (or Ockham). The principle states that one should not make more assumptions than the minimum needed. This principle is often called the principle of parsimony. It underlies all scientific modelling and theory building. It admonishes us to choose from a
set of otherwise equivalent models of a given phenomenon the simplest one. In any given
model, Occam's razor helps us to "shave off" those concepts, variables or constructs that are not really needed to explain the phenomenon. By doing that, developing the
model will become much easier, and there is less chance of introducing inconsistencies, ambiguities and redundancies.
Though the principle may seem rather trivial, it is essential for
model building because of what is known as the "underdetermination of theories by data". For a given
set of observations or data, there is always an infinite number of possible models explaining those same data. This is because a model normally represents an infinite number of possible cases, of which the observed cases are only a finite subset. The
non-observed cases are inferred by postulating general rules covering both actual and potential observations.
For example, through
two data points in a diagram you can always draw a straight
line, and induce that all further observations will
lie on that
line. However, you could also draw an infinite variety of the most complicated curves passing through those same two points, and these curves would fit the empirical data just as well. Only Occam's razor would in this case guide you in choosing the "straight" (i.e. linear) relation as
best candidate model. A similar reasoning can be made for n data points lying in any kind of distribution.
Occam's razor is especially important for universal models such as the ones developed in General Systems Theory,
mathematics or
philosophy, because there the subject domain is of an unlimited complexity. If one starts with too complicated foundations for a theory that potentially encompasses the universe, the chances of getting any manageable model are very
slim indeed. Moreover, the principle is sometimes the only remaining guideline when entering domains of such a high level of abstraction that no concrete tests or observations can decide between rival models. In mathematical modelling of systems, the principle can be made more concrete in the form of the principle of uncertainty maximization: from your data, induce that model which minimizes the number of additional assumptions.
This principle is part of epistemology, and can be motivated by the requirement of maximal simplicity of cognitive models. However, its significance might be extended to metaphysics if it is interpreted as saying that simpler models are more likely to be correct than complex ones, in other words, that "nature" prefers simplicity.
