Occam’s razor
In this recent Scientific American article, a well known historian of science provides a rather misleading description of Occam's razor or, more precisely, offers "counterexamples" that are not. The author's manifesto, that "the simplest explanation is often not the best one" is terribly ill-defined. I don't necessarily agree or disagree with that statement, it just needs a lot of context, and the context provided by the author is just plain wrong.
An "explanation" of an observation or phenomenon is a model, a simplified version of reality built for the purpose of falsifiably describing the "cause(s)" of the observation of phenomenon. Occam's razor is then a principle stating that the simplest possible model consistent with the observation and what is known about the phenomenon, using the least number of assumptions, is the best explanation. In the article, the author proceeds to give examples that are absolutely consistent with Occam's razor while trying to convince us that they are counterexamples.
- "While studying the motion of spiral galaxies, [astronomer Vera C.] Rubin discovered that the speed at which stars rotated around the center of their galaxies made sense only if these galaxies contained an additional mass weighing about 10 times more than the visible stars. The claim of a new form of “dark” matter—unseen and unseeable and present in far greater quantities than the visible matter of the universe—was not a simple explanation, but it turned out to be the best explanation." Sorry but the author herself tells us that the "only explanation" must involve dark matter, therefore no other explanation would explain the observation of the rotation of stars around the center of their galaxies.
- "Newton explained light as being made of particles, whereas other scientists of his era explained it as a wave. Quantum mechanics, however, tells us that light is, in some respects, both a wave and a particle. Newton’s account was simpler, but modern physics tells us that the more complex model is closer to the truth." Sorry, but Newtonian mechanics has not disappeared from higher education studies because it remains the simplest explanation of a lot of macroscopic phenomena. Quantum mechanics is required in a different domain to explain different observations.
Etc. More examples from smoking and COVID. Same fallacy.Understanding the context of an "explanation" means understanding what the model must be consistent with. You can't invoke Occam's razor without defining the constraints the model must satisfy. Some relatives of Occam's razor, such as the maximum entropy principle, have been wildly successful and take constraints into account explicitly. There would be no statistical physics or information theory without it.