Reconsidering Occam Razor Fallacy: Discussion?

[Andre]

I searched PF for "occam razor" and found 20 pages of threads. Hence it appears to be a very important argument. However, I think in reality that's is one of the most elaborate fallacies.

Suppose that A causes B and corrolates directly and that A causes C and corrolates directly, then is follows that B and C also corrolate.

Suppose that we discover B and C and its correlation but we do not know of not A. So logically we would like to explain the correlation between B and C and there is always a deducted hypothesis to explain why B causes C (or vice versa).

Searching for evidence we also may find D and E, which are all caused by interactions of the unknown A with B and C and we consider our hypothesis substantiated and we promote it into the theory of B, causing C,..E.

Even if F,G,H,I, which are discovered later, practically refute the theory, we consider that anomalies, measuring errors, contamination of evidence, or it's simply ignored.

Then somebody stumbles upon A and realizes that A causes B and A causes C and that B does not cause C and even that FGHI could be logically explained with A as ultimate cause.

Now the idea that our theory is wrong that B is not causing C appears to be hard to accept so what happens?: "Go away, we have already B and C, who needs A, Occam Razor".

So, here is the appeal to this very special rule, which sounds excellent but in reality is a fallacy since it's only recommending a course of action: "choose the simplest". However, the truth is completely independent of our choice and will not adjust to it. Choosing the more complex solution (if available - since A was not) would have had the advantage to help understanding the complexity of the matter and would have allowed for more flexibility fine tuning the hypothesis to the complex reality.

I think we should reconsider the logic of Occam razor.

Discussion?

 

[Office_Shredder]

Actually, a cobbled theory of why B, C, D, and E appear to be related (since we can't verify B causes C given what we know) is more complicated than a simple single cause for all of them

 

[Gelsamel Epsilon]

Funnily enough you're simplifying Occam's razor too much. Occam's Razor basically says "Choose the simplest yet suffice (or the simplest out of all theories which are equal)". The discovery of A would render previous theories inadequate and therefore more must be added for the theory to suffice. However if A was unobserved, but thought of, and a theory was proposed which explained everything IF A exists but there is no observation of A then it will be cut down with Occam's Razor because until such A is observed you're unnecessarily complicating the theory.

For instance, Einstein was taken over Lorentz because Lorentz speculated about ether which has so far been unobserved, so when choosing between Lorentz and Einstein the math between them comes out with the same results, but Lorentz has ether, therefore Lorentz is dropped and Einstein is picked up.Edit: Without Occam's Razor you'll have many abstract and complex stupid theories about Z Y X T Q R S and definitely P, you may also get A, but who is to say which one of these is right especially if they contained things that are unobserved? B and C are fine, if A is observed then the theory changes to be suffice but until then why complicate things?

 

[Andre]

Funnily enough you're simplifying Occam's razor too much. Occam's Razor basically says "Choose the simplest yet suffice (or the simplest out of all theories which are equal)".

I know, I merely point out how occam razor is (mis)used to defend the paradigm.

The discovery of A would render previous theories inadequate and therefore more must be added for the theory to suffice. However if A was unobserved, but thought of, and a theory was proposed which explained everything IF A exists but there is no observation of A then it will be cut down with Occam's Razor because until such A is observed you're unnecessarily complicating the theory.

Which makes it unclear if Occam Razor promotes or obstructs the end goal of approximation of the truth as closely as possible

If A is observed then the theory changes to be suffice but until then why complicate things?

Well there is a big wall of psychological complications (Thoman Kuhn) between reality and that ideal proces (Karl Popper).

My example of such an A -> B, A-> C would be order of magnitude,

A could be something on solar system level, B and C something on planetary level, D, E, F, on subsystems of the planets like Mars Venus, Saturnus, about tectonics, geology, climate, etc.

If B and C are on the same order of magnetude on planetary level, the chance would not be negliglible that both are caused by the next higher level but specialisations in those scientific branches discourages to look upwards and think of a possible A.

 

[Gelsamel Epsilon]

I don't think it discourages it, maybe slightly but science has already gone so far I doubt some discouragement is going to stop people from looking. There are people out there lookin' for ether.

It seems as though we discover things like A as we go along, if Occam's razor wasn't in place then ALOT more people would be wasting time trying to find out explanations for once off anomalies.

So far Occam's Razor has proven to be the best way to deal with things.

 

[Andre]

Well beyond the level of flat Earth's, ether, etc

Perhaps we are running around in circles, but how and when has Occam's Razor proven to be the best way to deal with things or has it proven to be the best defence of challenged paradigms and what would that say about it's role in the quest for the truth?

If there is one truth out there to explain a certain phenomenon and we have x possible explanations of different complexity, why should the simplest be the closest to the truth?

And then again what is simple and what is complex? Isn't it that a subjective label? For instance, is it the number of elements, the complexity of the elements, how about intuitive factors, finding something obvious or far fetched/counter intuitive? What if there is an A out there that obviously would have explained many B...N in a very simple way, whilst B only explains C but gets strongly challenged by D,E,F. However, what if A itself is highly counter intuitive and too complex to model?

I can be persuaded to give an example of that.

 

[Chaos' lil bro Order]

Andre,

How would you apply Occam's Razor to Feynman's notion that photons take all VIRTUAL path's in the Universe while traveling from point A to point B, but that the probability of the photon ACTUALLY taking one of these obscure virtual paths cancels out on aggregate and thus the photon takes the simplest geodesic path? This seems to fly directly in the face of Occam's Razor, yet its pretty well established physics.

I may be mistaken, but I always thought Occam's Razor was not a hard fast rule, but was more of a philosophical guideline to weed out crackpot theories that introduce more variables than necessary when addressing a question.

Personally, I do not rule out any theory, no matter how convoluted, but I do relegate them to the realm of incredibly small possibility. So small a possibility in fact, that they can be generally ignored, but I'd never be so conceited as to say they are impossible. To give one ridiculous example, the coffee cup on your computer table may spontaneously explode with the force of a 1 Megaton nuclear bomb for no apparent reason at all. But this would violate so many axioms of physics that have held true for such a long time, that the possibility of this actually occurring can be ignored for the length of yours or my lifetime.

Or maybe this explains why anytime I see someone drinking coffee I run into my bunker :)

 

[HallsofIvy]

The point of Okham's razor is not that there is any a-priori reason why "the simplest explanation is more likely to be the true explanation" but rather that the simplest explanation can be more quickly disproven than a complicated reason- and that's the whole point of science.

 

[selfAdjoint]

The point of Okham's razor is not that there is any a-priori reason why "the simplest explanation is more likely to be the true explanation" but rather that the simplest explanation can be more quickly disproven than a complicated reason- and that's the whole point of science.

Right. It's not a logical conclusion, it's a search strategy. Speculation on imaginary entirities without any reason is a good way to waste your life if you're a scientific researcher. Damn easy to get lured into, too.

 

[Chaos' lil bro Order]

The point of Okham's razor is not that there is any a-priori reason why "the simplest explanation is more likely to be the true explanation" but rather that the simplest explanation can be more quickly disproven than a complicated reason- and that's the whole point of science.

Very nicely said Ivy, I shall remember your definition for future reference.

 

[Chaos' lil bro Order]

Right. It's not a logical conclusion, it's a search strategy. Speculation on imaginary entirities without any reason is a good way to waste your life if you're a scientific researcher. Damn easy to get lured into, too.

Lol Self, you better not let us find you inventing a free-energy machine doohicky or its off to the asylum for you!

 

[selfAdjoint]

Lol Self, you better not let us find you inventing a free-energy machine doohicky or its off to the asylum for you!

Ever read the essay "Swimming through spacetime"? Apparently in curved spacetime you can move in any direction at a constant speed just by doing repititive body flexings. Crazy enough for you?

 

[jostpuur]

I've had bad experience with Occam's Razor. It sounds good in theory, but in practise its message is often: There's no reason to assume I'm wrong, since it is simpler to assume I'm right

Simplicity is so point of view depending. New things feel complicated, old things one has got use to with feel simple.

Everytime somebody uses Occam's Razor badly, you can say that it was not fault of the Occam's Razor, but fault of bad user of it. Still, Occam's Razor provides a good tool to skip logic.

 

[NeoDevin]

I always though Occams Razor tells us to use whichever consistent theory works best for us. As long as a theory correctly predicts all the observations, there is NO way to determine which mathematical model is 'correct'. However, as soon as a theory is shown to be inconsistant with observations, it is no longer consistent, and must be abandoned (so Occam's Razor no longer applies).

It has nothing to do with which theory is correct or incorrect, that can only come from comparison to observations.

 

[Old Smuggler]

I always though Occams Razor tells us to use whichever consistent theory works best for us. As long as a theory correctly predicts all the observations, there is NO way to determine which mathematical model is 'correct'. However, as soon as a theory is shown to be inconsistant with observations, it is no longer consistent, and must be abandoned (so Occam's Razor no longer applies).

That depends on what is meant by the phrase "inconsistent with observations".
Frequently one sees, at least in astrophysics, that the original prediction doesn't match
the data, so to recover observational consistency, one introduces fudge terms which
effectively act as free parameters. In this manner, "consistency" with observations is
often claimed even if the fudge factors have no independent constraints. However,
this comes with a price; namely that the predictive power of the theory is diminished.
In such cases Occam's razor should cut off the extra theoretic degrees of freedom if
there is a choice between theories.

It has nothing to do with which theory is correct or incorrect, that can only come from comparison to observations.

Yes, but as illustrated above, there are often ways to obtain "consistency" with observations by introducing free parameters. In such cases Occam's razor should
prefer alternatives which predict observations from first principles.

 

[Schrodinger's Dog]

I searched PF for "occam razor" and found 20 pages of threads. Hence it appears to be a very important argument. However, I think in reality that's is one of the most elaborate fallacies.

Suppose that A causes B and corrolates directly and that A causes C and corrolates directly, then is follows that B and C also corrolate.

Suppose that we discover B and C and its correlation but we do not know of not A. So logically we would like to explain the correlation between B and C and there is always a deducted hypothesis to explain why B causes C (or vice versa).

Searching for evidence we also may find D and E, which are all caused by interactions of the unknown A with B and C and we consider our hypothesis substantiated and we promote it into the theory of B, causing C,..E.

Even if F,G,H,I, which are discovered later, practically refute the theory, we consider that anomalies, measuring errors, contamination of evidence, or it's simply ignored.

Then somebody stumbles upon A and realizes that A causes B and A causes C and that B does not cause C and even that FGHI could be logically explained with A as ultimate cause.

Now the idea that our theory is wrong that B is not causing C appears to be hard to accept so what happens?: "Go away, we have already B and C, who needs A, Occam Razor".

So, here is the appeal to this very special rule, which sounds excellent but in reality is a fallacy since it's only recommending a course of action: "choose the simplest". However, the truth is completely independent of our choice and will not adjust to it. Choosing the more complex solution (if available - since A was not) would have had the advantage to help understanding the complexity of the matter and would have allowed for more flexibility fine tuning the hypothesis to the complex reality.

I think we should reconsider the logic of Occam razor.

Discussion?

Occam's Razor is not meant to be absolute, or a law in the sense of always being correct, or anything approaching an ultimate answer, it is just a principle, therefore it need not be 100% logically applicable in the same way "[URL law[/URL] isn't technically a law or applicable, and in fact it's an example of fallacious conclusion from poor understanding of probability, well it would be if anyone took it seriously.

But any way.

All things being equal, the simplest solution tends to be the best one

Note the use of the word tends, not is, not is always but tends.