From LOJBAN%CUVMB.BITNET@UBVM.CC.BUFFALO.EDU Fri Dec 1 08:28:24 1995 Reply-To: "John E. Clifford" Date: Fri Dec 1 08:28:24 1995 Sender: Lojban list From: "John E. Clifford" Subject: fur muzzles To: lojban list Status: OR Message-ID: I remember the discussions about comparative-basic gismu somewhat differently from Lojbab, so my residual ill-will was directed not at him but at a group of people who, in the discussion, could not get beyond the English (etc.) surface forms to the deeper princples or who (shades of current topics) could not see the perceptual (better than "subjective") side of "average" (say) for the mathematical. However, if Lojbab's report of the argument that finally convinced him to give up the good stuff is correct, I'll be happy to include him. Reducing positive blanu to more blue than lo'e blanu is either infinitely regressive (bluer than the typical thing that is bluer than the typical thing that is bluer than...) or is viciously circular, specifying a new member of a class by a definition that assumes that the class is already complete. It is also, of course, irrelevant, since calling a broda blanu (or talking about a blanu broda) always referred not to lo'e blanu but to lo'e broda (and average with respect to color, at that). JCB had a nack for picking lousy examples and setting his case for these gismu on a color word may have been another instance. (But it does seem that all the talk about standard colors rather than comparisons is about answers to the question "What is this color?" rather than "What color is this object?" And even then, the process seems to me to be as plausibly reconstructed as a comparison -- more like standard blue than like standard aqua. As plausibly, at least, as any metrical or ordinal reconstruction of simple"tall" judgments.) He might have done better with "tall." But he had another problem, since he had set in stone that the missing arguments were all particularly bound variables, "something"s. So one good thing that did come out of the comparative- gismu discussion (and some others) was the Lojban (I think Loglan caught up eventually) rule that the missing arguments on a predicate are to be taken as conventionally or cooperatively defined for the context. In the same way, my comments about "fuzzifying proselytizers" or whatever was aimed more at some of the real pros, Kosko in particular. However, if he wants it, the shoe fits smoothly on stivn's foot. He does not, admittedly, say that the dictionary contradicts (or even is contrary to) pitr's claim, he just says that he uses the dictionary to "correct" pitr. A fairly fuzzy distinction, I think. But before you protest that that is not fuzzy, notice that, in stivn's usage, anything is or can be fuzzy. He keeps, for example, talking about various ordinal scales as fuzzy, without giving any evidence that they are and, indeed, even after defining the one for tallness (not directly related, notice, to "is tall") in metrical terms (between n and m centimetres high) as well as giving them common labels ("very tall," "rather tall," "quite tall," and so on). Then, to back this up (or because presupposing it to be true) he takes this ordinal scale and starts using it as though it were a ratio scale, taking 3/4 as .75 rather than in the 3 category of a (?)1-4 ordinal scale. In any case, nothing about ordinality (or any of the other Guttman types) is inherently fuzzy, though you can fuzzify any of them in some way -- differently for the different types, I would think. So back to the basic point. An awful lot of terms are not at all fuzzy, even if there are occasional hard cases about whether they apply. For some of these it may be useful in some discipline to introduce various scales corresponding to these concepts in that informal way that technical terms do correspond to natural ones and it may also prove to be useful for some of these scales to have fuzzy components or even be entirely (at the ratio level) fuzzy. None of this affects what the original term meant at home but is about some other term with a different home. It may even be that the new term tells us interesting things about the old term, that a fuzzy "tall" enables us to produce an intelligent system that reproduces people's judgments of tallness with high accuracy and allows us (as formal logic does for arguments) to have good reasons for decisions in hard cases. That still does not mean that the technical term (and its apparatus) have hit on the real meaning of the natural term; we can disagree with the machine's judgment -- even in systematic ways -- just as we do about formal logic's judgments about some arguments. Nice tool, but sometimes gets it wrong. And even if it were never wrong, that does not mean that it gets the right results for the right reasons: I don't measure heights and run calculations (and certainly not in centimetres), I just notice that someone is tall. As for the beauty of a fuzzy program for dealing with astronaut candidates, hey, that is what that language (and perhaps even that chip) was designed to do, so of course it does it apparently better than a general purpose langauge on a general purpose chip. But does the computer in the end do anything differently as a result of the instructions (assuming the program is run on general purpose chip)? Snobol programs for character manipulation are models of simplicity compared to, say, C programs to do the same thing. But the computer does basically the same thing when it executes (slightly -- but insignificantly - - different because of different optimizations; they just do useless things in different places). That does not mean that the concepts embodies in Snobol are better or nearer the mean of text than those in C. It is just a matter of convenience for a purpose (lost now, since there is no Snobol for anything after 286). On another point, "deep" should be used very cautiously (always good advice, of course, about key words) in this discussion, since it is not always a vertical dimension. It is, in fact , the front-to-back dimension in furniture (and the like, I'm unsure about 3-D matrices in math), along with "high/tall" (top-to bottom) and "wide" (left-to-right, all in "in use" orientation). I wonder what the difference is between the high sea, the deep sea and the wide ocean, in the English store of cliche's (or tropes anyhow). pc>|83