I know what Occam’s razor is, but I’m confused as to how it is used. Very often I see people using Occam’s razor in a debate where they imply that through Occam’s razor, the simplest explanation must be true. But isn’t this a nonesense? Doesn’t Occam’s razor simply instruct us to test the simplest hypothesis first, as they tend to be correct, and then test others?
So, what is the correct way in which to use Occam’s razor in a debate?
You’re correct; Occam’s Razor doesn’t say the simplest hypothesis is necessary the true one.
In its Latin form, the rule says: “Entia non sunt multiplicanda praeter necessitatem,” meaning that entities (objects or phenomena whose existence is assumed) must not be multiplied without necessity. You shouldn’t assume something new to exist when you can explain all your observations with your current set of assumptions. Once this ceases to work, you have to make new assumptions.
A good example of this is the debate between the geocentric and heliocentric concepts of the universe. As long as astronomical observations could be explained with the hypothesis that the Sun orbits the Earth, Occam’s Razor implied that this was the most likely explanation. Once the orbits of other planets were observed more closely, and showed behavior that couldn’t satisfactorily be explained with this hypothesis but with the other one of the Earth rotating around the Sun, it was felt necessary to think about it again.
I think it would be more accurate to say that the acceptance of the heliocentric model was a success of Occam’s Razor in the absence of any strong evidence favoring one model over the other. AFAIK there were no appearances that led to the overthrow of the earth-centered model.
Ptolemy believed (for religious/philisophical reasons) that all heavenly motions had to be perfectly circular, so he placed the planets on epicyclyes–circles rotating on the edges of rotating circles. He needed many other kludges as well, and constructed a very elaborate mechanism that explained the motions mathematically, but necessarily had to ignore certain inconvenient facts. For instance, in these orbits on orbits, the planets were sometimes seven times closer to Earth than at other times, yet did not appear proportionally larger. This was explained away with certain assumptions about the otherworldly nature of the planets that made them different from sublunar objects.
Copernicus’ heliocentric model also relied on circular orbits (and epicycles, IIRC) and was therefore nearly as complex as Ptolemy’s. It did not gain immediate acceptance, in part because it was politically incorrect, as Copernicus knew–he gave orders not to publish until after his death–and that Galileo found to his peril, but also because it was essentially equivalent to the Ptolemaic model without providing major advantages. Today it seems much more “logical” to assume that the farthest objects–the fixed stars–don’t move at all, than that, in rotating around us once a day, they move faster than everything closer to us. But there were long entrenched belief structures to explain that and other philisophical objections. From a theoretical/mathematical point of view, Copernicus and Ptolemy were equal.
It was only when Kepler realized, a century later, that the orbits were better explained by elliptical, not circular, orbits, that the heliocentric world view gained significant power, precisely because it explained all the appearances without multiplying unnecessary entities. The power of Kepler’s view was that it was much. much simpler than either Ptolemy’s or Copernicus’. Thus it was preferred by Occam’s Razor.
Say you have a finite set of “true” statements T, that you call facts about a system S. Say you do not know any other “true” statements about the system S that are not in T. Any theory P that satisfies every statement in T is then “true”, until you find a statement p derived from P that is “false”, then you need a better theory. The way I interpret Occam’s Razor is that it dictates which “true” theory you should start falsifying first. Since they are all “true”, you should pick the simplest one and disregard others until that simple one becomes unsatisfactory (if ever).
I don’t know if that makes sense to anybody else.
Essentially it’s “the simplest true hypothesis is the one that should be accepted.” i.e. of all hypotheses that explain the data, the simplest one should be preferred. It doesn’t say anything about false hypotheses.
And by “simplest” I mean the one that makes the fewest a priori assumptions. I like groman’s way of thinking about it too.