If You Dont Like It You Dont Understand It

Try it! You'll like it!

One of the FallaciousArguments, wherein it is asserted that dislike of the topic or position at hand (which is supported by the speaker) implies a lack of thorough understanding of that topic or position. In other words, a type of AdHominem argument - it is asserted that disagreeing with the speaker implies a lack of qualifications in the subject.

Often, this fallacy is supported by a demonstrable correlation between knowledge of the subject and agreement with the speaker's position! However, CorrelationImpliesCausation itself is a fallacy; there are several different possibilities which may be explain the correlation:

  1. Familiarity with the subject does indeed lead to agreement with the speaker's position. The conclusion the speaker wants you to draw.

  2. Those who agree with the speaker's position are more than likely to study the position in detail than are the general population. This is common when the speaker's position represents a minority viewpoint, and a large number of those agreeing with the speaker see it as their mission to evangelize to the rest of us.

  3. The literature, coursework, etc., which one must use to become familiar with the topic at hand is itself biased to the speaker's position; or the speaker (and his clique) have become experts by reading only a subset of the literature available (that which supports their position).

  4. Literature which disagrees with the speaker's position is dismissed as lies, propaganda, cluelessness, etc. - anyone who is well-read in that can similarly be dismissed as an ignoramus.

This fallacy also contains elements of NoTrueScotsman; as the definition of "knowledgeable person" is limited in scope to "knowledgeable person who agrees with me".

And it seems to be a special case of the DogmaticFallacy.

This (alleged) fallacy is often used by those who don't support general SelfStandingEvidence. They will argue or imply that the topic is too complex or give a rebuttal on Wiki, and that the answer is found in volumes of references or books. In other words, a BookStop. Related: ScienceShouldBeEasy. My belief is if you cannot explain your viewpoint on Wiki, then it is likely an invalid or ill-formed view. -- top

{I wrote a response here that I decided belonged on TopMind instead --ChrisMellon?]

Re: "alleged fallacy" -- the start of the page clearly describes circumstances when it would indeed be a fallacy. I assume that your objection is to the cases when someone claims that is the case incorrectly. That is, just because I claim you dislike it merely because you don't understand it, that doesn't mean it's true, and if in fact it is not true, then I would be the one making a fallacious claim.

That's always the case, though; for every logical fallacy, someone can incorrectly claim that it has occurred. Conversely, someone can correctly claim it has occurred, and yet it could be incorrectly denied. That's not unique to the topic of this page.

Anyway, top, you're not entirely wrong when you say if you can't explain, there's a problem, but you're on thin ice, because you don't always accept explanations that other people consider clear, so with that history, you're not going to get very far saying that, you know?

Regardless, I think that your phrasing is too strong anyway. It is not "likely an invalid or illformed view", it's more like a Smell. Not the same as "likely". Gotta soft-peddle that more.

Some things are just hard to explain, truly. Here's an example I ran across yesterday, where DanBernstein is describing the future prospects for proving numbers prime...a small quote: "For these cohomology groups, one can take a p-adic approach using a de Rham-type cohomology, lifting X to a p-adic ring R and take the hypercohomology of the de Rham sequence, quite explicit and computable but the complexity is worse than linear in p. Instead, <equation>, the H^i derive from injective resolutions on the etale topology."

Some technical subjects can't really be adequately explained to anyone but another specialist, so one should be careful of blanket statements about explanations.

-- DougMerritt

Well, that is the crux of the problem. Somebody can claim that their pet position is like that even though it might not be true. In software it often comes down to, "If you were properly informed, you would then understand how elegant my solution is". In your example it is not that difficult for a non-mathematician to verify that a claimed prime is a prime. Testing a prime is far simpler than deriving one. The above appears to be speculation about future techniques, not something objectively measurable. If software solves (or creates) real problems, then there should be real metrics. Elegance is not a metric. If you consider "elegant" math to be a short equation, then that is a code-size issue more or less, which is measurable to some extent. Software is not to please programmers, but to please managers and customers. If X is better than Y, then ways must be found to relate it to what managers and customers want. If the above people made claims such as "The X algorithm can find more primes than the Y algorithm per hour", that can be tested. The manager or customer who wants primes, say for security purposes, can count that even if they are not mathematicians. The above are like to engineers arguing about which car engine design is the fastest. The ultimate test is on the race course where even non-engineers can measure the results.

Umm...no, some problems there. "is not that difficult for a non-mathematician to verify that a claimed prime is a prime" -- that's actually false, because we're talking about arbitrarily large primes. If I claim that 1 billion factorial plus one is prime, you can't verify that (but in some cases DanBernstein and his colleagues might be able to).

Well, okay a "usable" prime.

Nor does it matter that he's talking about future techniques, because that lecture is years old and is now current technique. :-) Also you're now talking about "objectively measurable", which is always interesting to talk about, but before it was just about "explanation", so that's a topic change.

I am quite sure what you mean. However, the topic is kind of weird in that it talks about "liking" instead of being better or worse.

Also "Software is not to please programmers" -- not quite true either; the behavior of software is to please managers and customers, but the interior form of software is irrelevant to them, and is indeed crafted to please programmers (and of course micromanaging first level managers, but then they are a kind of programmer when they do that). One needs to keep claims about the behavior of software separate from claims about the interior form of software. Primes per hour can be verified objectively, yes; that's behavior. Easier for future maintenance programmers to understand, is about the form, but is frequently subjective, and is almost never purely objective.

The manager cares about productivity. If maintenance slows down productivity, then it is an issue. It appears to me that people are most productive when they work in like-minded teams. Twins tend to be highly productive because they think so much alike that they don't have to spend a lot of time communicating and living with code style that they are uncomfortable with. It is the similarity, not absolute value of techniques that seem to matter the most.

To bring up a specific example, Barbara Liskov's LiskovSubstitutionPrinciple is purely and wholly from mathematics, but is reinterpreted by non-mathematician programmers as a rule of thumb about elegance - which works to some extent, but misunderstands. Certain aspects of programming such as that actually do need appropriate mathematical background to truly understand, without which even the best of explanations will be at least partially misunderstood.

So I wasn't just making a completely unrelated analogy. Some aspects of software need math. -- DougMerritt

But that is generally a domain-specific issue.

Not in the case of this particular example. It is one of the strongest non-objective criteria regarding creating OO software of any and all sorts. (It actually applies to non-OO, also, but I don't think that's super widely appreciated/practiced.)

I did not mean type substitution but the first example. I did not make that clear. Besides, the substitution rule does not say whether types are good. It only provides a framework to keep them predictable IF you decide to rely on types.

Almost forgot: note the rather interesting point above, that the number in question is definitely objectively either prime or composite, but only a specialist has a hope of verifying which. Everyone else is forced to accept the judgement of the specialists, even though it's an objective matter.

Aren't there accepted algorithms for those kinds of things?

In cognitive science/anthropology this is well-known and has a name, which I've forgotten... the usual example is that everyone knows that a birch and an ash are both trees, but a large percentage of the population has no idea how to identify them. We rely on specialists (which in this case could be simply nature lovers in parts of the country where both grow) to tell the difference for us. Or less dramatically, most people could probably do pretty well at pointing out bushes versus trees, but would not be able to formally define the difference (unless you accept Peanut's "a tree is a bush that made the bigtime").

The same thing absolutely happens in software. You're competent, you can tell a bush from a tree. You may even be able to tell an ash from a birch. But there are always going to be some areas where you're not a specialist, and must rely on the opinion of specialists. -- DougMerritt

Classifications usually don't make very good examples. What something is labelled as and whether it is useful are two different things. Labels may allow us to make short-cut judgements, but they are not a substitution for science. For example, a paper-maker may know from experience that certain kinds of trees make better paper. However, "better paper" is the real metric. It is just that it is not practical to test every tree. Related: LimitsOfHierarchiesInBiology

Yes, I know, we've talked about that page before, and I just wrote something on Intertwingled just yesterday that roughly agrees with your favorite philosophy here trees and hierarchies have limits, yes, yes.

But no, that doesn't really apply here, because nothing you said contradicts the point that you must rely on specialists' judgements. If an expert paper maker says that it's best to use birch wood, then you can't verify his argument - it may sound convincing or it may not, but you really don't know, since you lack the various multiple related background specialties to verify. Conversely, you're agreeing that the classification of trees does matter to that paper making specialist. So that whole paragraph really went nowhere.

Either you become a specialist and you learn things like topology to understand Bernstein's argument about primes above, or else you just have to accept his judgement. Either you become a specialist and learn about mathematical type theories, else you have to accept the judgement of specialists like Liskov. You haven't made a counter-point against this. -- DougMerritt

IdontUnderstandLispButIlikeIt?

I really like Wiki, but I only think I understand it because I know how to use it.

Exactly like everything else in life! :-)

FallaciousArgument

EditText of this page (last edited July 8, 2007) or FindPage with title or text search