staugustine.net
entia non sunt multiplicanda praeter necessitatem,
Frustra fit per plura quod potest fieri per pauciora.
Don't be an undetectable creature.
Occam's Razor has become a basic tool for those who follow the scientific method. The primary activity of science — formulating theories and selecting the most promising ones — is impossible without a way of choosing from among the theories which fit the evidence equally well. For example, it's easy to think of alternative theories to Newton's famous statement that every action has an equal and opposite reaction. One such theory could be that for every action there is an opposite reaction of half intensity, but undetectable creatures magnify the reaction so it appears to be equal in magnitude; these creatures will all die in the year 2055, at which point the observable nature of the universe will shift. (Variations of this theory might mention 2056 instead, or 2057, and so on to infinity.)
Until 2055, the creature hypothesis will appear identical to Newton's law, leaving a scientist with no way of telling which, if either, is right. This is the problem of underdetermination: for every set of data, there are infinitely many theories which are consistent with it. Without a method to choose between these competing and theoretically possible theories, science ceases to function entirely and becomes useless for all practical purposes, and perhaps even useless in principle. Occam's Razor is such a tool, expressing a preference for theories with fewer unsupported elements. Thus, it deems the third law of motion in its original form preferable to the version with the added undetectable creatures.
William H. Jefferys and James O. Berger (1991) quantify this undesirable factor that in its extremity manifests as unnecessary assumptions into the degree to which a proposition is unnecessarily accommodating to possible observable data. Theories which specifically, logically entail the observed set of data or are similarly entailed by it are preferred over theories which are trivially consistent with it by mere virtue of being consistent with a wide range of possible data a priori or ad-hoc adjustment that is otherwise unjustified (see also Bayesian inference and falsifiability).
Go gold. |
