Message-Id: <199602130616.BAA07937@mail1.access.digex.net>
Date:         Tue, 13 Feb 1996 00:18:38 -0600
Reply-To: "Steven M. Belknap" <sbelknap@UIC.EDU>
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: "Steven M. Belknap" <sbelknap@UIC.EDU>
Subject:      gismu for <lojrfuzi>
To: Bob LeChevalier <lojbab@ACCESS.DIGEX.NET>
Status: OR
Content-Length: 15737

coi

Yup. More fuzzy stuff. Hit N to skip over :-)

lojbab:
>>>Why must people insist on being so bloody metaphorical.  what is wrong
>>>with "nalsatci" as the critical modifer of the defining tanru.

Because <nalsatci> is a <malglico>.

>We also make conisderabel analysis for usage in tanru and lujvo making.
>Both words and their opposites are heavily used in natlangs for
>acompound building.  That  this does not apply to satci can be seen in
>English at minimum, by the fact that  we say
>light/heavy and not light/unlight
>long/short and not long/unshort
>thick/thin andnot thick/unthick
>
>But we use
>exact/inexact - a scalar negation form just like Lojban nalsatci would be.
(some deleted text)
>I really don't see in any event why you "rejected it", unless you see
>a meaning to nalsatci that is other than what you intend.

Sorry, <satci> - <nalsatci> is not the dichotomous pair we need. I would
translate this pair into English as something like "precise" - "vague".
They are both <malglico> as applied to logic.

"Fuzzy logic" really refers to "continuous-valued logic" rather than
"sloppy logic" or "vague logic" (Fuzzy also has "warm 'n fuzzy" overtones
in English, as opposed to "vague", which has a "clueless and confused"
overtone.) There is a pervasive misunderstanding as to what the "fuzzy"
means in "fuzzy logic", both on this list, and among the great unwashed.
(Although John Cowan did correctly distinguished fuzzy from probability.)
In my view, satci is a particularly inappropriate term for Zoroastrian
(true/false) logic. A better choice would be "discrete" Thus the English
antonymal buddy pair is:

"discrete" - "continuous"

I just reviewed my notes from 23 October 1995, when I was previously trying
to puzzle out the English to lojban translation of "fuzzy logic" and
"Zoroastrian logic":

$ 2 possibilities for "fuzzy"
$
$ <ranji> (refers to continuity of events and states ?English domain-range?)
$   &
$ <kuspe> (? implies a gap "bridged" by something.)
$ <kuspe> is better, since it is not restricted to events/states.

I couldn't come up with something good for discrete, eventually settled for
"satci" (precise) and then got disgusted and gave up and apparently tried
to learn about the lojban tense system, judging by my notes. (Which reminds
me of how I incensed I was/am by the existence of the "now" tense, but
that's another topic. (Yes, lojbab, I know lojban is tenseless by default.
Please.)

I actually have no objection to the lojban gismu antonymal buddy system,
but I am amused that the lojban cogniscenti first made a whole bunch of
antonymal buddy pairs, then groused about there being too many. I think the
objection about too many gismu is totally bogus. As long as they are
semantically and phonetically distinct, not clearly expressible as lujvo of
other gismu ( <na> constructions don't count), reasonably well spread out
in semantic space, and fairly universal across cultures, who cares how many
there are? Those who think there are too many words are welcome to limit
their vocabularies. Heck, it works for Pat Buchanan :-)

steven:
>>Besides, don't you mean <ka na satci kei>? (Has a rather doubleplusungood
>>quality to it.)

lojbab:
>Why?  What is the double plus ungood quality?
>
>Length?  No.

C'mon, lojbab. I realize that length of a selbri is not a big deal. Sheesh.
I am a relative newbie to lojban, but I knew the 17 meanings of <bilti
cmalo nirli ckela> years before you ever heard of Loglan. I am referring
here to the simplified English of the novel 1984, where compound predicates
like doubleplusungood were imposed on the hapless citizenry. This resulted
in great frustration because of the loss of the ability to express subtle &
complex ideas. In other words, <ka na satci kei> basically sucks.
(Humor-impaired insert laugh track here.)

>If the "ka" were necessary to the metaphor (and to me this
>might be the distinction between "fuzzy" and "fuzziness", then it would
>be kamselsatci, staying one syllable longer than the corresponding English
>word, which is probably at least average if not better than average for
>less-used concepts.

Right. And we are seeking the more abstract sense of the gismu here:

<satci>       -    "precise"
<ka satci>    -    "precision"

But <satci> doesn't cut it. "Precise" is a horrible metaphor for the logic
you Zoroastrians worship! "Accurate" isn't any better. What does precision
or accuracy have to do with it? It is quite precise but not very accurate
to say that "5 divided by 2 equals 3." We fuzzifiers call this "round-off
error." It is quite accurate but not very precise to say "It is either true
or false that lojbab can jump over his house." We call this "vague."

Accuracy and precision are certainly important, but using them to describe
a type of logic is absurd.  <satci> is just as good of a metaphor for
"fuzzy" as it is for "discrete"; at least as applied to logics; that is,
neither one is very good at all. "Two-valued" isn't good either; a
two-valued logic could be either fuzzy or discrete. There are multi-valued
logics (False, Maybe, True), but these could be either fuzzy or discrete.
(This is what "granularity" of a logic refers to.)

I think it was Goran who pointed out that people who claim that a political
system must be either capitalist or communist are committing a false
dichotomy. Couldn't it be *in-between* the two somewhere? In fact, as Goran
pointed out, most real political states *are* in-between. (See, Goran, *I*
read your posts!) This is the sort of logical fallacy I see committed
daily, even, I am sad to say, on this discussion group.

Part of the reason I feel so passionately about putting fuzzy logic in
lojban is that I believe that English and possibly other natlangs
"Sapir-Whorf" a person into false dichotomies, which contributes to
depression and thus much human misery. Interestingly, depression is
invariably characterized by what are called "cognitive distortions", which
are mostly highly personal logical fallacies.

(Technical note: depression is distinct from sadness. A patient of mine
once defined depression as "a dishonest sadness", which is about the best
definition I have heard.)

So lojban might be important, both to test this Sapir-Whorfian hypothesis
of mine, and possibly to decrease human misery. Healing with language; what
a concept. You can take the doctor out of the clinic, but you can't take
the clinic out of the doctor...

As lojbab (helpfully) corrected my spelling, a fuhivla for "fuzzy" is:

<lojrfuzi>

So we need a fuhivla for discrete logics. I want something long and hard to
say, and I like the overtone of "near religious devotion to an ancient,
out-dated, should-be-abandoned concept" So I suggest:

<lojnzorohastruni>

which scans beautifully, wouldn't you say?

But lojbab prefers we avoid importation (should there be a language
tariff?), so tonight I tried again to find a gismu for things
Zoroastrian/discrete/distinct, and this time hit paydirt (maybe). What
about <kantu>!!! But we need to use <ka>, thus suggesting "quantized" as
opposed to "quantum". So we have our antonymal buddy pair:

<kamkuspe>        -         <kamkantu>

          corresponding to

"continuous"      -         "discrete"

                 or

"fuzzy"           -         "Zoroastrian"


So lojbab was right about gismu. There *were* gismu for fuzzy and discrete;
it was my lack of lojban vocabulary (and the obtuseness of the rudimentary
dictionary) which was the problem.

Steven:
>>Perhaps sufficient resources are present. But every other question I've
>>asked about "how to say something" was readily answered almost immediately.

lojbab:
>And there is something wrong with this - that we can answer "how to say it"
>queations?  Or are you saying that fuzzy expressions alone are the
>exception.

I am saying that fuzzy expressions have been the exception. Usually, when I
ask a "How do you say it?" question, somebody, (usually Jorge) comes up
with an elegant, accurate expression. Compared with everything else in the
language, it is just too hard to say things fuzzily in lojban. A gismu does
not solve the problem, although at least now we can discuss the problem in
lojban.

djer:
>And has made the point that je'a has to mean "is to some degree .." if
>we are to use it as suggested. And that is a new idiomatic definition
>of je'a. At present we can only use the ranges (or are they points,
>thus not fuzzy?) defined by je'a, no'e, na'e and to'e.  Goran has
>raised the question, what do the rest of the NAhE_xiny mean.
>
>Subscripts are used in mathematics to indicate particular members of a
>class, such as the points on a line which may be designated X1, X2, X3,
>etc.  The "X" part of the symbols asserts that X_n is indeed a member of
>the line-set.  The _n designates a particular member.
>
>Now when we consider je'a, which asserts that some predicate has
>(scalar) truth, i.e. "da je'a melbi" ,[there exists something x which
>is (truly,scalar) beautiful], we expect that da will always be in the
>class, melbi.  However, when we consider the subscripted je'a---
>je'a_xiny, and we let the variable ny approach 0, we find that we have
>moved to the other end of the scale, and now we are talking about
>something ugly.  A subscript value has converted something beautiful
>into something ugly.  We don't expect this. We expect that da will
>always be something beautiful, as asserted in the original sentence.
>So we have created an idiom. Because an idiom creates a new meaning
>different from what one would expect from the meaning of the component
>words.
>
>Futrthermore, a subscript on je'a would normally be expected to
>describe some member of a set of je'as.  We have a similar case
>documented in the grammar where back-counting sumti is accomplished
>with "ru_xiny." Applying a similar convention to je'a means that
>je'a_xiny refers to some previously used je'a in the sentence or
>discourse.  Here is an example of such a usage:
>
>i.le do du'u je'a jetnu ije
>le mi du'u je'a jetnu ije
>le je'axipa du le je'a xire
>
>Your predications are (scalar fuzzy) true and
>my predications are (scalar fuzzy) true and
>the first scalar fuzziness is equal to the second scalar fuzziness.
>
>My point is that the construct "je'a_xiny" has the above plausible
>interpretation, John's proposal supplies another plausible
>interpretation which I believe is more idiomatic, and requires more
>explication to make any sense of.

[ political points, with which I have some sympathy, deleted :-) ]

djer:
>A dictionary translation of "je'a jetnu"  would be "certainly true" but in
>a relative sense. It is not an absolute claim for truth as is seen in a
>binary truth table. It just means that the truth value is between je'a
>and no'e, and closer to je'a. Or does it mean that the truth value is
>anywhere left of to'e? Or very close to je'a but not touching it?
>
>Right there is a problem for this idiom because "mi je'a
>xipinonononopa" melbi" would mean I was very close to no'e melbi, (or
>is it to'e melbi as above?)  the next value on the scale, and I don't
>think this is what the idiom is trying to say. But we can't know what
>the idiom is supposed to say from the dictionary.

lojbab:
>>But subscripts have no COMMON "literal meaning" in such a context - all
>>we have is some suggestion of a 'different variety of je'a', since that is
>>what subscripts usually mean in variable contexts..
>
>However they have been used, they have not been used or defined as an
>operator to slice up a continuum. Je'a alone does not do this. Je'a is
>merely a vague range on the continuum.  That is why the proposed usage
>is idiomatic.  You have to know the idiom up front to get the meaning,
>it could not be inferred from the atomic meaning of the words "je'a"
>and "xiny" where xi means subscript and ny are numbers. Possible
>exception would be the use of "puxipa" and "puxire" to refer to points
>in the past, earlier or later, the usage was never clarified.  I
>haven't seen "vaxire" used, but why not if this is the way we want to go
>to identify points on a continuum?

djer argues more eloquently than I that John Cowan's <ja'a...xi> proposal
for <kamkuspe logji> is too idiomatic. The meaning of expressions using
these constructs are not apparent from the <cmavo> definitions alone. This
strongly suggests they are too idiomatic. As djer points out, there are
other problems as well.

These piecemeal catch-as-catch-can schemes are awkward in the extreme. I
have no objection to overloading operators, where such overloading follows
clearly from the meaning of the root primitive. For example:

a+b

is nicely overloaded with:

{a1,a2,a3}+{b1,b2,b3}

but nastily overloaded with "OR" Sure there's a conceptual basis for this
nasty overloading, but it becomes badly idiomatic when the overload is too
far removed from the primitive.

<kamkuspe logji> (fuzzy logic) is a superset of discrete logic. It is
distinct from the meaning and operation of the other available cmavo. We
need <cmavo> and possibly <selmaho> to be properly fuzzy. These should
allow one to express continuous properties with arbitrary granularity and
Guttman scale as applied to *any* construct in the language. I would like
to know more about lojbab's SEI ideas. Perhaps he's got something there.

djer:
>>At present we can only use the ranges (or are they points,
>>thus not fuzzy?) defined by je'a, no'e, na'e and to'e.

lojbab:
>You have presented the strongest case for And's proposal yet, though not
>to my mind convincing.

>Fuzzy logic is a rapidly evolving very new field and I am quite sure that
>what is understood about fuzzy logic in 10 years will be unrecognziable
>today.

This is patently false. Lotfi Zadeh's initial paper was written in the
fifties. Fuzzy logic *predates* lojban. It was well established and
published *before* the 1960 Scientific American article. If JCB had spoken
to the right person, he could have included it in his first version of
Loglan, and there would have been no need for revision since then. I reread
Zadeh's article recently, and it is perfectly clear in view of current
understandings of things fuzzy. All the basic principles of fuzzy logic
were done by the early sixties. Fuzzy logic isn't drifting about. It's as
solid as arithmetic.

>We cannot hit a rapidly movcing target, and that target will no
>doubt like the rest of logic and matematics consist of idiomatic and
>highly unnaturalistic  conventions that no one outside the fields could
>hipe to understand.

Surely, you jest. Is lojbab unaware that he used a fuzzifier in the
previous sentence? The word "quite" in "I am quite sure" is a fuzzy
operator. So much for your "highly unnaturalistic conventions that no one
outside the fields could hope to understand." You sound like the guy who
was surprised to find he had been "ambulating" all his life. Technical
words are useful only so far as they are understood. Need I say it? "Quod
erat demonstradum."

>I don't intend to try to do so.

If what you say is true, I would agree with your position. But what you
claim about the novelty of fuzzy logic is quite false. Sure there is active
research in the area of fuzziness, just as there is active research in the
area of algebra. But the essentials of fuzzy logic are rock solid, and
aren't going anywhere. Are you also afraid that someone will redefine
"zero" out from under you? Are you suggesting we delete the numbers from
lojban because people are still doing mathematics research? Give me a
break.


cohomihe kamkuspystivn


Steven M. Belknap, M.D.
Assistant Professor of Clinical Pharmacology and Medicine
University of Illinois College of Medicine at Peoria

email: sbelknap@uic.edu
Voice: 309/671-3403
Fax:   309/671-8413