Let’s clarify some of the terms here. A “theory” is just a collection of one or more propositions. I will discuss what we count as “data” in section 2.3. For now, an intuitive understanding will suffice. “Elegance” is a measure of explanatory completeness, i.e., how much of what is posited by a theory is explained
within that theory. Go through the things a theory says are the case. Ask why each of those things is the case. Why does this exist? Why did this happen? The more a theory can answer, the more elegant it is. The more times a theory fails to answer, the less elegant it is. Elegance can be thought of as opposite of arbitrariness. The more that a theory posits arbitrarily, the more there will be unexplained things within the theory, the less elegant the theory will be. The fewer arbitrary things a theory posits, the more the theory will provide explanations for what it posits and hence the more elegant the theory will be. The Elegance Principle is not meant to be special or controversial. It is similar to principles that many people use, including the many variations of Occam’s Razor. Why do I use the Elegance Principle instead of another principle? I leave this question to my full discussion of my views on scientific methodology.
2.2. Applying the Principle
It is straightforward to see how to apply the Elegance Principle in general. If a proposition is included in the most elegant theory that explains all of the data, then it should be accepted as true and its negation rejected as false. It is also fairly simple to see what sorts of theories the Elegance Principle will favor. It is easier to ask more questions about complex theories, so in general, the Elegance Principle will favor simpler theories. These simpler theories will still need to explain all of the data, of course, and this will only be possible if the theories are sufficiently universal. As a result, in general, the Elegance Principle will favor theories that are more universal. Sometimes when simple, universal theories are used to explain data, the theories cleanly explain the data. Other times, the theories can only be made to explain the data by adding various additional posits (dust in the lens, unseen forces, etc.). Since these additional posits typically decrease the elegance of a theory, all else equal, the Elegance Principle will typically favor theories that require one to make fewer additional posits of this sort. Unfortunately, while it is possible to see how to apply the Elegance Principle in general and while it is possible to see broadly what sorts of theories the Elegance Principle will favor, I do not have a procedure for applying the Elegance Principle precisely in practice. I have no way to precisely rate theories in terms of elegance. I cannot provide them with a precise elegance “score”. I also do not know how to measure the amount of elegance a theory gains or loses as a result of particular modifications. A new posit is added and not explained – how much elegance is lost? Two semi-universal laws are replaced with a single universal law – how much elegance is gained? I do not currently know how to answer these questions. In practice, what I do is this. If I am assessing a theory, I consider the theory and get a general sense of how elegant its major components are. I then look at how well the theory explains the data. First, I look for data that directly contradict the theory. If there are any such data, the theory is refuted. Second, I look for data that the theory explains by means of arbitrary
Here and elsewhere in this document, I use the word “explain” so that “X explains Y” does not
imply “X correctly explains Y”. According to my use, it is consistent to have an explanation of something that is not a correct explanation of it.