Occam's Razor hypocritical? Discussion links request.

February 13, 2007 12:36 AM

Deceptively unchatfilter =P I am looking for good discussions on Occam's Razor -- which I think is obscenely absurd. Looking for links to discussions, not necessarily petitioning comments (although welcome).

Mostly I'm looking for links to discussions that have already taken place (as in, not currently taking place) or perhaps chat logs, or something of that nature -- discussing the validity of Occam's Razor. I've already seen the wiki.

I just think that trying to whittle down as few assumptions as possible when trying to find an answer, may in fact be the most absurdly irrational assumptions one could possibly make. It's like throwing out all possibilities that involve doing a lot of research (perhaps out of one's field) for the sake of not having to go look stuff up. Is it not largely a way to get out of doing a lot of work?

Extra points if the discussion became so confounded to be ended by Godwin's Law (but not necessary, I just like those) =D

posted by vanoakenfold to grab bag (28 comments total)

The term "Occam's razor" is bandied about all over the internet in all sorts of inconsistent ways. I'm not exactly sure what you mean by "trying to whittle down as few assumptions as possible", but I don't think that captures what is best thought of as Occam's razor. Here's a good place to start reading on it. (That's a Stanford Encyclopedia of Philosophy link.)
posted by painquale at 1:09 AM on February 13, 2007

Oops, that's not a discussion of Occam's Razor. Unless you consider a summary of arguments that have gone back and forth in philosophy a discussion. Apologies.
posted by painquale at 1:13 AM on February 13, 2007

I just think that trying to whittle down as few assumptions as possible when trying to find an answer, may in fact be the most absurdly irrational assumptions one could possibly make. It's like throwing out all possibilities that involve doing a lot of research (perhaps out of one's field) for the sake of not having to go look stuff up. Is it not largely a way to get out of doing a lot of work?

From what I understand, OR isn't applied prior to having an answer (i.e. "when trying to find an answer"). It's used when you already have more than one answer that all explain a phenomena equally well. In that regard, you don't get out of doing any work as far as I can see.
posted by juv3nal at 1:26 AM on February 13, 2007

yes... it is also known as the principle of parsimony and has more to do with shaving away the unnecessary: "plurality should not be posited without necessity."

... perhaps if you searched on the principle of parsimony.
posted by de at 2:43 AM on February 13, 2007

Here's Carl Sagan talking about Occam's Razor.
posted by joannemerriam at 2:49 AM on February 13, 2007

Gotta love Sagan.

juv3nal has an important piece of this: explaining phenomena equally well. Think about basic science. If a theory explains most but not all data, and the outliers appear to not be spurious, the theory is not likely complete, yet the alternatives may predict it less well so for the moment it stands as the best theory.

But indulging alternative theories when the data is completely explained by a simpler (or more "elegant") theory is an endless, meritless task. Consider that there is an infinite number of explanations (just by virtue of recursive explanations) to explain any observation, any event.

I drop a ball and it hits the ground.
Maybe a "force" which I will call "gravity" is responsible.
Maybe this force is fictitious, and the event occurs because space is bent in some unseen fourth dimension, but let's still call it gravity.
Maybe a space gremlin did it.
Maybe a red space gremlin did it.
No, no, a leprechaun did it...

I'm being intentionally silly here, but these alternative theories could explain the data equally well in terms of accounting for all data points. In particle physics, that gremlin or leprechaun isn't so far-fetched -- it's just a new particle, theorized to account for some observations not yet explained. And where you can explain an observation with one particle, you can certainly with two. Or four. Or forty-four. Which should you prefer? There really is no limit.
posted by dreamsign at 3:06 AM on February 13, 2007

Also, science is an interminable process by definition, so you don't really abandon explanations that you shave off with Occam's razor. You eliminate them for the time being in favor of the simplest possible explanation, which then guides your future research. If the simplest explanation subsequently meets obstacles it can't account for, it stops being an explanation, and it may be one of the previously discarded explanations that takes its place. Occam's razor in empirical science, at least, is really only an efficient way of steering research (and there are competing ways). You've got to steer somehow.
posted by game warden to the events rhino at 3:20 AM on February 13, 2007 [3 favorites]

The utility of occam's razor is self-evident. It is like reducing an equation: the simpler form is easier to work with. Science is essentially about the ability to predict. When you have a theory that explains the known facts, you don't gain anything by complicating it with unnecessary assumptions or entities that do not add to its explanatory power, and you lose by making it less wieldy.

This doesn't mean that the simpler theory is more "true" than the complex one, just that it is more useful. It also doesn't mean you stop testing or questioning it, or looking for new facts which require a more complex theory.
posted by Manjusri at 3:58 AM on February 13, 2007

Are you aware that Occam's "Least Hypothesis Theorem" is a heuristic? Like all heuristics, sometimes it's wrong. Heuristics are guidelines that you apply because they're usually right, and this one is usually right, too. But as the old joke goes, if a heuristic was never wrong, it would be an algorithm.
posted by Steven C. Den Beste at 4:03 AM on February 13, 2007 [1 favorite]

Hypocritical? It's simply useful. Suppose I leave my car keys on the kitchen counter. Later I return and they are gone. There is no sign of forced entry to the house. Let me think up some reasons for what could have happened:

a) My spouse, who lives in the house with me and was home, moved them for some reason.
b) Jesus was reincarnated in the house and took them. He walked through the wall when he left.
c) An alien spacecraft crashed into the house, destroyed my keys, invaded the local city, took prisoners, gave them anal probes, took off again, repaired my house such that I couldn't tell it was damaged, but forgot to replace my keys. They used a mind-field so I didn't notice.
d) The computer simulation of reality that I actually exist in dropped a bit somewhere and the key object was deleted as bogus data.
e) The neighbors next door used their telekinesis ray to snatch the keys, transporting them out through a dryer vent.

Every one of those explanations fits the observations I made in the original paragraph, and therefore is a valid hypothesis for explaining what happened to my keys. Occam's Razor:

"This is often paraphrased as "All things being equal, the simplest solution tends to be the best one." In other words, when multiple competing theories are equal in other respects, the principle recommends selecting the theory that introduces the fewest assumptions and postulates the fewest hypothetical entities. It is in this sense that Occam's razor is usually understood."

suggests that a) is the best or most likely explanation. It is essentially "common sense" applied to scientific theories. It is useful because it works; from experience with exactly this problem, I will tell you that theory a) is extremely likely to be borne out as the correct explanation.
posted by jellicle at 5:51 AM on February 13, 2007 [1 favorite]

It's worth also looking at some of the more mathematical expressions of the principle which balance how well a model fits the data and how many free parameters it requires you to put in.

Wikilinks: 1 2 3
posted by edd at 5:59 AM on February 13, 2007

Manjusri: "This doesn't mean that the simpler theory is more "true" than the complex one, just that it is more useful."

It would be extremely convenient if truth and utility were separate and unrelated realms. The painful reality, however, is that the whole point of seeking truth is to live better lives; and there is no convenient yet false theory that isn't likely to blow up in our faces tomorrow. More to the point, it's impossible to know if our "convenient theories" really are useful; they may seem so to us, but with what can we compare them if we don't know the truth of the matter? And even if they have been convenient in the past, how can we know that they will continue to be?

jellicle: "It is useful because it works; from experience with exactly this problem, I will tell you that theory a) is extremely likely to be borne out as the correct explanation."

Once again, unfortunately, nobody can count on past experience when trying to understand what will happen in the future. And, once again, even if you could, how exactly do you know that choosing the simplest explanations has always served you best? What do you compare that with, and how? Things get very complex very quickly, unfortunately.

I should say that I'm a very big fan of our friend William of Occam, a man whose name has been dragged through the dirt and misused far too often. His position was meant as a denial of the multiplicity of beings-- you can get this from the wiki, so I imagine you understand-- but not a denial of complexity. Anyone who's read Occam can tell you that his arguments are very, very complex-- the world is a difficult place. It's also quite useful to note that Occam was one of the great theologians; which is to say that he was a highly rational and religious man. It's a little silly when one sees his name bandied about in support of the rank denial of all metaphysical existence.

I think that, in the end, what people call "Occam's Razor" could be very, very useful if it were applied correctly. As it is, however, modern minds accept a plethora of irrational and unproven assumptions: that there is a special and particular entity called "material existence" which constitutes everything that is; that there is nothing which doesn't follow specific 'rules' inherent in said existence; that all processes are somehow reversible and testable; that experimentation constitutes the deepest knowledge of the observable world, and perhaps the world in general; et cetera. But tell anyone nowadays that these things are worthy of some doubt and skepticism, and they'll laugh you out of town. Modern minds, much unlike medieval minds, are closed to other possibilities.
posted by koeselitz at 8:35 AM on February 13, 2007

In the non-scientific world, I've seen/heard it used as jellicle describes it: "the simplest explanation is probably right." And like others have said, it's typically used after two (or more) competing theories have been raised.

In my job (lawyer), we use that concept all the time, though we may not call it Occam's Razor. For example, in a case that we presently have, there is an e-mail that was purportedly written by the opposing side on the day of the event in question. It corresponds with our client's testimony about what happened and it basically beats their case. The opposing side says "I didn't write that e-mail, it's either a fake or someone else wrote it." The e-mail came from his computer, the metadata shows it was sent on the date in question, and he works alone from home. So the three scenarios are (1) the e-mail is legit, (2) it is a very elaborate fake, or (3) somebody else wrote it, unbenownst to the guy, from his home computer on the day in question. The argument to the jury is a simplified version of Occam's Razor: you have to decide which of the three is the truth. The simplest one is the most likely.

I think the term "Occam's Razor" is often used by non-scientist types because it sounds cooler than the regular old explanation. And it sounds a bit more official.
posted by AgentRocket at 8:37 AM on February 13, 2007

One thing to add: usually, when people invoke Occam's razor, they're being lazy. They'd like to think that they can do away with messy possibilities without actually facing them. People will say: "well, somebody stepped on my cookie. Maybe a purple elephant just walked by and stomped on it, eh? I suppose that's what you'd like to think! But Occam's Razor!" It is, however, usually possible to show evidence that there wasn't a purple elephant. You have to go through it, though. That's hard-- and not possible in some cases (see religion)-- which pisses people off. Hence the dramatic proliferation of the misuse of Occam's Razor.
posted by koeselitz at 8:40 AM on February 13, 2007

koeselitz's point, I think, is about the inapplicability of Occam's Razor to the supernatural.

I think that it's fair to everyone to say that koeselitz's point applies to discussions about cosmic issues, but not to the day-to-day situations that jellicle and AgentRocket describe.
posted by ibmcginty at 9:04 AM on February 13, 2007

In science, Occam's razor is at best a tentative way to select among competing theories which explain the evidence at hand equally well. If theories A, B, and C all explain the evidence you have so far equally well, but A is the simplest explanation, the best Occam's razor can give you is to say that for now, A seems like the most likely explanation.

It most definitely does not absolve the scientist from doing subsequent experiments which would distinguish between A, B, and C. If those experiments are not feasible at the time, whether due to financial considerations, current technological limitations, other experiments having higher priority, or whatever, you put the experiment on the back burner until it becomes possible, but you don't forget about it.

If there is no conceivable experiment which could distinguish between explanations A, B, and C, then you are out of the realm of science and into philosophy. And it's not at all clear to me that it's correct to apply Occam's razor in cases such as those.
posted by DevilsAdvocate at 9:16 AM on February 13, 2007 [1 favorite]

So the three scenarios are (1) the e-mail is legit, (2) it is a very elaborate fake, or (3) somebody else wrote it, unbenownst to the guy, from his home computer on the day in question.The argument to the jury is a simplified version of Occam's Razor: you have to decide which of the three is the truth. The simplest one is the most likely.

Well, that's absurd. First of all, faking an email is very, very easy, and so is faking the meta data. I would hardly call any faked email 'elaborate'.

Secondly, occam's razor never applies in multi-agent situations where deception is possible.

Finally, to answer the initial question: philosophical arguments are often bandied about by people who don't understand what they mean. Occam's razor should be used only when two (or more) explanations fit all the data.

In AgentRocket's example, his explanation does not fit all the data, because the sender denies sending it. The data is in conflict, which means some of the information is incorrect. Occam's razor doesn't really apply.
posted by delmoi at 9:23 AM on February 13, 2007

In AgentRocket's example, his explanation does not fit all the data, because the sender denies sending it. The data is in conflict, which means some of the information is incorrect.

Well, but that's easily solved by rephrasing AgentRocket's explanation (1) as "the email is legit and the sender is lying."

I agree that this is not necessarily "simpler" than the competing theories though. It's not clear how you would measure simplicity in this case.
posted by game warden to the events rhino at 10:23 AM on February 13, 2007

Well, perhaps now we are unintentionally getting into the type of discussion vanoakenfold was looking for.

In my case, I used the term "Occam's Razor" in a purely descriptive, advocative sense; the jury has to choose between competing theories of this e-mail's authenticity, and I am using the concept as one more reason why they should find my way.

The concept of "the simplest explanation is typically the most realistic one" is, as others have pointed out above, a variety of common sense. It seems from the more informed comments here that I am (blatantly, but unintentionally) misusing "Occam's Razor" to dress up my argument and give it greater weight. I think that's probably what a lot of people do.

And the guy totally wrote that e-mail.
posted by AgentRocket at 11:30 AM on February 13, 2007

Good discussion people! The OR tells us, roughly, not to postulate entities unnecessarily. So when considering an effect, one shouldn't postulate any causes beyond what is already included in the set of all accepted theories. If we have an effect, X, and a set of accept theories S, to add a postulate is to change the whole set of accepted theories - making S', which is fine in the face of empirical evidence, but problematic without. A new theory can only be meaningful in the context of the field of theories it is contained within. S', however, being made without empirical evidence, is outside this context, and is thusly meaningless.

In terms of truth, well, let's just leave Truth aside for awhile. We don't have access to such a thing as Truth. What we have is our sense data, our minds, and our theories. (Even these three stand in some dispute.) What we are trying to do in science, and perhaps with knowledge in general, is predict future events, or bring past events into a coherent field with the present. To do so we need to behave as if certain entities exist and others do not. Deciding whether particular entities should or should not be considered when attempting to further scientific, or in some cases general, knowledge is the essence of Occam's razor.

koeselitz: Interesting post. I would hold that philosophers, more than scientists, are quite prepared to be skeptical about material existence etc. Scientists are too busy using their base postulates to create predictive systems. Philosophers, on the other hand, are given the time to think beyond our current epistemological systems and question such assumptions. Perhaps, science and philosophy are on a continuum - with science doing the heavy lifting and philosophy doing the heavy thinking.
posted by elwoodwiles at 11:31 AM on February 13, 2007

ibmcginty: "I think that it's fair to everyone to say that koeselitz's point applies to discussions about cosmic issues, but not to the day-to-day situations that jellicle and AgentRocket describe."

I'd be very interested to know where precisely you draw the line between 'day-to-day' and 'cosmic.' Besides, I wasn't merely speaking of 'supernatural' situations. In fact, I was speaking precisely about 'natural' ones.

elwoodwiles: "What we are trying to do in science, and perhaps with knowledge in general, is predict future events, or bring past events into a coherent field with the present."

This seems to be the consensus among scientists now, who get very annoyed when one questions them regarding the 'truth' of their conclusions. But it's fairly important to point out that this is an extremely recent development; until at least fifty years ago, scientists aimed at knowing the truth about the world. People tend to have short memories nowadays.

"I would hold that philosophers, more than scientists, are quite prepared to be skeptical about material existence etc. Scientists are too busy using their base postulates to create predictive systems."

I would argue that many of those base postulates are wrong or at least unwarranted. Doesn't that impact science a great deal? Note that I'm not saying that "material existence isn't all there is!" but rather "material existence is fundamentally different than scientists assume it is." If the postulates are simply wrong, then, while it may seem useful, science might not be useful at all.
posted by koeselitz at 3:49 PM on February 13, 2007

Here's one answer to roughly this question at askphilosophers.org. And here's my quickie answer -

The bottom line is that you are working with a distorted idea of what Occam's Razor recommends.

Roughly, it recommends that when we are choosing among equally good possible explanations for some observation, we should choose the explanation that doesn't posit the existence of a bunch of extra things.


Simplest example:
Observation: an apple fell from the tree.
Explanation 1: gravity
Explanation 2: gravity plus invisible ghosts

Either of these would explain our observations. Occam's Razor says that we should choose Explanation 1, because it doesn't posit the existence of extra entities - ghosts.


Another example:
Observation: an apple fell from the tree.
Explanation 1: gravity
Explanation 2: a heretofore undiscovered force, totally unlike the forces we already believe in.

In this case, we want to say that Explanation 2 is a less likely explanation, because it would force us to believe in a totally new force, when a force we already believe in offers a perfectly good explanation. But it's not clear whether thisi s because of Occam's Razor, or some other principle of simplicity in scientific explanations. (Occam's Razor tells us not to "multiply entities" if we don't need to; it's not clear whether forces count as entities. There is a different principle that says between equally good explanations, we should favor explanations that fit in well with the rest of the science we already believe in, over ones that require a radical departure. If you're interested in this question, get an introductory book about philosophy of science, and maybe also read Thomas Kuhn's Structure of Scientific Revolutions.)


Example where Occam's Razor doesn't help:
Observation: this airplane can fly
Explanation 1: the Bernoulli effect
Explanation 2: the Newtonian explanation

Occam's Razor won't help us to decide between these two theores -- we need to do more tests, gather more observations that will help us see if either is the correct explanation. So, if you're concerned that Occam wants us to stop doing experiments, set your mind at ease.



There are lots of questions about the principle -- for example, what's the best way to formulate it exactly (when we're comparing more-complex theories, how should we judge which posits fewer entities? is it about particular, individual things, or kinds of things?). And of course even the best formulation of it will sometimes lead us to a wrong answer -- since it's a "rule of thumb" (a heuristic) rather than a guarantee of a right answer. But it's not vulnerable to the objection you make, and there's nothing "hypocritical" about it.
posted by LobsterMitten at 4:18 PM on February 13, 2007 [1 favorite]

If the postulates are simply wrong, then, while it may seem useful, science might not be useful at all.

If science lets us build bridges, make vaccines, and so on, then in what sense is it not useful? If you just mean that it's not useful for finding out the truth about the world, then you're collapsing the distinction that the truth-vs-usefulness discussion was based on.

(Honestly, it sounds like you are interested enough in this to benefit from reading up on philosophy of science. These issues have been exhaustively debated by some very smart folks over the last 100 years.)
posted by LobsterMitten at 4:22 PM on February 13, 2007

I apologize for my derail. Here's a brief response to LobsterMitten's points:

LobsterMitten: "If science lets us build bridges, make vaccines, and so on, then in what sense is it not useful? If you just mean that it's not useful for finding out the truth about the world, then you're collapsing the distinction that the truth-vs-usefulness discussion was based on."

False conclusions cannot be useful in a consistent or demonstrable way. How do we know that those things will continue to work, and how is it possible to extend the our powers, if we cannot generalize?

The purpose of science has, for centuries, been to understand the truth about nature in a way that makes nature useful to humans. The truth about nature has always been emphasized, from Bacon to Newton to Faraday to Einstein. For a variety of reasons, the basic formulation of science has now been popularly denigrated by those who'd like to do away with the messiness of the real world; but plenty of people (I'm thinking particularly of Max Planck) have seen that it's not so easy as saying that "science is simply about utility."

"Honestly, it sounds like you are interested enough in this to benefit from reading up on philosophy of science. These issues have been exhaustively debated by some very smart folks over the last 100 years."

I know. I have. But they've been significantly less smart over the last fifty. Hence the parroting of a bastardized version of positivism that goes on so often.

And I have a feeling we'd all benefit if scientists read a bit of the history of science. But reading anything from prior to fifty years ago seems to be out of vogue at the moment.
posted by koeselitz at 10:58 AM on February 14, 2007

I wonder what you mean by "true"?

Anyway, one problem is that we have no way of determining whether or not our scientific knowledge is true, but we can know if it is repeatable, regular, predictive, etc. Thus, science sets truth aside and works on what can be directly observed. The world, as you say, is messy. Science, however, works hard to simplify the world so we can navigate it's weirdness, hopefully to our benefit. Science isn't out to make dogma; it is always and only provisional.

anyway, it's an interesting issue. Maybe you should think of a question to post to askme that would spark an informative discussion.
posted by elwoodwiles at 11:56 AM on February 14, 2007

Koeselitz: I did not imply that "truth and utility were separate and unrelated realms", but that they are not identical. It is certainly possible to know if a theory is useful, you simply test it. If it allows me to predict the future with an increased degree of accuracy, it is. Occam's razor is not opposed to complexity, but unnecessary complexity that does nothing to improve the predictive power of the theory.

I put "true" in quotes as a shorthand nod to the unknowable nature of ultimate reality. Sure, maybe tomorrow suddenly things will stop behaving in a predictable manner, or maybe it's all an illusion anyway. Science doesn't have much to say about this any more than arithmetic can describe how to enjoy a painting. I apply the razor and scientific method the same way I do math or a pocketknife, as a tool for manipulating the world where appropriate. There are plenty of areas where they do not apply, but when they do, they are an improvement over the alternatives.

The canonical example of applying the razor to the existence of diety(s) is perfectly appropriate, regarless of Occam's original beliefs. If subtracting this entity from one's worldview does nothing to impair one's ability to predict things, then it is beneficial in this context to slice it away. It doesn't really matter whether the deity exists or not, removing it results in a more elegant model.
posted by Manjusri at 2:32 PM on February 14, 2007

If the postulates are simply wrong, then, while it may seem useful, science might not be useful at all.

False conclusions cannot be useful in a consistent or demonstrable way. How do we know that those things will continue to work, and how is it possible to extend the our powers, if we cannot generalize?

Well, science works its way through successively better approximations to the truth. It's improving, we think, all the time. That's no reason to think that our current, merely-partial understanding of the world isn't useful to us!

It would be silly to say "the Newtonian model of physics is not true, so it could never have been useful". It was, and is, useful for all sorts of things. It's close enough to the truth to be useful. If we want to predict the forces acting on passengers in a crashing car, Newtonian mechanics will do nicely for that purpose, and it doesn't matter that in the limit of inquiry Newtonian mechanics turns out to be false. The usefulness of a theory has to do with (1) whether it's "true enough", that is, whether it matches the world enough for our purposes, and (2) what our purposes are. For many of our purposes, current scientific theories are plenty useful -- even though they will probably be modified in the future as we learn more.
posted by LobsterMitten at 11:06 PM on February 14, 2007 [1 favorite]

Notice that for a theory to be useful, it doesn't matter whether it will eventually turn out to be false. It doesn't matter if we come up with a better theory next week. If our current theory allows us to accomplish some goal, then it's useful, even if next week's theory will be even better/more useful.
posted by LobsterMitten at 11:09 PM on February 14, 2007