* Monday, 2011-10-17 at 01:32 +0100 - And Rosta <and.rosta@gmail.com>:
> Martin Bays, On 16/10/2011 06:05:
> > * Sunday, 2011-10-16 at 02:56 +0100 - And Rosta<and.rosta@gmail.com>:
> >> Martin Bays, On 15/10/2011 21:04:
> >>> * Saturday, 2011-10-15 at 01:49 +0100 - And Rosta<and.rosta@gmail.com>:
> >>>> Martin Bays, On 14/10/2011 23:59:
> >>>>> * Friday, 2011-10-14 at 11:39 +0100 - And Rosta<and.rosta@gmail.com>:
> >>>>>> Martin Bays, On 13/10/2011 05:33:
> >>>>> But once we perform this resolution to the level of mundanes, we
> >>>>> find that different interpretations of {lo} resolve to different
> >>>>> logical forms. For example, {na ku lo cinfo cu zvati lo mi purdi}
> >>>>> has at least the two following meanings in terms of actual lions:
> >>>>> 1. {lo cinfo} is interpreted as a plurality of mundane lions, giving
> >>>>> roughly:
> >>>>> For L some (contextually relevant) lions: \not in(L, my garden)
> >>>>> (which probably means that there exists a lion among L which is not in
> >>>>> my garden)
> >>>>> 2. {lo cinfo} is interpreted as the kind Lions, giving
> >>>>> \not in(Lions, my garden)
> >>>>> which is then resolved existentially, giving
> >>>>> \not \exists l:lion(l). in(l, my garden) .
> >>>>>
> >>>>> So subtleties aside, we have a straightforward ambiguity between
> >>>>> \exists l:lion(l). \not in(l, my garden)
> >>>>> and
> >>>>> \not \exists l:lion(l). in(l, my garden) .
> >>>>>
> >>>>> This seems toljbo to me.
>
> Sorry, I was unclear. I meant that English seems to allow only reading (2), and that the same might go for Lojban.
Ah! Have {lo} *only* able to get kinds, you mean? That would indeed deal
with my main issue with kinds+xorlo. But presumably xorxes wouldn't like it.
> >> ); my only point was that lo + countable is not more ambiguous than lo + mass or le or la.
> >
> > (what does countability have to do with anything? Would anything change
> > if we were dealing with an uncountable set, say with {lo namcu}?)
>
> Countables have intrinsic boundaries, and that makes it relatively
> easy and natural to distinguish one mundane countable from another (of
> the same type). With uncountables, such as chlorine, it's relatively
> easy and natural to not distinguish one mundane from another, and
> hence the kind--mundane distinction too seems absent too.
I don't see any problem with the current approach - handling chlorine
with the same plural semantics as it does lions, where the atoms are
(rather arbitrary) quantities of chlorine (rather than, say, atoms!).
Anyway, I'm not sure this issue is important in the context of the
current discussion.
> >>> This seems to me a good reason not to have Obama-stages!
> >>
> >> Natural language (or english, at least) does
> >
> > Does it really? My impression from xorxes' explanation of them (and I've
> > never come across the concept outside of this mailing list) is that
> > they're an alternative way of handling tenses, eventually mostly
> > equivalent to the straightforward "possible worlds" approach (where
> > there's one obama, but many of his properties (including his existence)
> > vary from world to world). I don't see how english could force you to
> > use stages.
>
> I meant 'stages' not in the strict semantic sense but rather the
> looser sense that I'd understood it to have in this discussion, namely
> "subtype of a type that has intrinsic boundaries". "Obama" is a type
> that has intrinsic boundaries, but English allows us to speak of
> subtypes of Obama too, as in "the young Obama" or "(the) two Obamas"
> or "an unusually exuberant Obama".
>
> The main difference between 'Obama' and 'lion' -- as far as accounting
> for their differing grammatical behaviour goes -- is that Obama is
> naturally seen as being a singleton at any one point in time. But
> where that difference diminishes, as with 'Father Christmas' or
> 'Elvis' or 'Mickey Mouse', so too does the difference in grammatical
> behaviour, so that it is quite usual to speak of "two Father
> christmases, two Elvises, two Mickey Mouses, a Father Christmas,
> a Mickey Mouse".
Right... and "Elvis is not in my garden" could mean either "no Elvis is
in my garden" or "the Elvis is not in my garden". Yes, 'Elvis' in the
first sentence does seem to refer to a kind. But it isn't referring to
anyone called Elvis, so I don't see that this shows that Elvis has
stages.
The "temporal stages of Obama" example could be dealt with by
intepreting Obama as the kind 'Obama-stages', I agree, but it could also
be dealt with just by using tenses. I'm not sure how to deal with "an
unusually exuberant Obama"... but since it's a rare construction, a hack
like transforming it to "Obama, who was unusually exuberant" would seem
reasonable.
> >> , but the key point is that it's a metaphysical choice. You can choose
> >> to reject Obama-stages but accept lion-subtypes, but that must be your
> >> choice, not Lojban's. Lojban should be metaphysically neutral. Well,
> >> maybe you don't think it should be neutral,
> >
> > Probably not. I'm not entirely sure what you mean by metaphysics here,
> > but I'm taking it to refer to the question of what we put in the domain
> > of our universe when doing model-theoretic formal semantics.
>
> That's right.
>
> > I think
> > that the nature of lojban does impose some restrictions there - for
> > example, (roughly) there should be for each expressible unary predicate
> > precisely one object satisfying the corresponding ka. Lojban requires
> > this.
> >
> > Ignoring domain-switching technicalities, xorxes would want it to also
> > contain, for each (appropriate?) unary predicate, another entity,
> > a kind, which satisfies the predicate itself.
> >
> > In both cases, it's the language which is imposing this "metaphysical"
> > requirement. You can try to interpret expressions in the language
> > without following the requirement, but you're going to get bizarre
> > results which weren't those intended by the designers.
>
> I was intending to make two points, but not distinguishing them
> clearly. The first is that the metaphysics entailed by the semantics
> of lV is an especially permissive one, making the fewest possible
> distinctions and prejudgements.
But it requires a metaphysics which has kinds. That seems pretty
restrictive to me.
> If a language aspired to metaphysical neutrality and had to pick one
> metaphysics, it should pick that one. The other point is that this is
> just the metaphysics of lV; somebody wanting a different metaphysics
> could use different gadri. So Lojban could achieve metaphysical
> neutrality by offering a menu of different gadri.
But they're not isolated. If lV forces you to put kinds in your
universe, then (at least unless we introduce rules to prevent it)
quantifiers like {da poi broda} and {ro broda} will pick them up.
So if I choose to omit kinds from my universe but otherwise use the same
rules, I am likely to be misunderstood by a kind-using lojbanist, even
if I avoid using lV. Xorxes just gave a nice example, the other way
round: {mi zukte da poi do zukte} makes a sense with kinds that it
doesn't without them.
The metaphysics affects everything. I don't see that the language could
be considered well-specified if it didn't specify the metaphysics (on
the broad level we're talking about). Lojban with plural semantics is
very different from that without it; the same goes for kinds.
> >>>> And the exercising of that choice is metaphysical rather than
> >>>> linguistic. Lojban is metaphysically neutral.
> >>>
> >>> But the definition of xor{lo} is such that the existence of kinds is
> >>> required to make sense of many statements which are, in the final
> >>> analysis, about mundanes. So the metaphysics (if that's what it is) is
> >>> more-or-less hardwired into the language.
> >>
> >> Are they really, in the final analysis, about mundanes? Is there
> >> anything in xorlo that forces the kind--mundane distinction to be
> >> recognized? As far as lo goes, there is no kind--mundane distinction.
> >> For every individual X and Y there is an individual Z thatX and Y are
> >> subtypes of; for every individual Z, there are individuals X and
> >> Y that are subtypes of Z.
> >
> > Where an Obama-stage is a proper subtype of Obama? What's a proper
> > subtype of an Obama-stage?
>
> Does 'proper subtype' mean "X is a proper subtype of Y iff X is
> a subtype of Y and X is not a subtype of a subtype of Y"?
No, just "X is a subtype of Y and X is not equal to Y".
> If so, then I think the notion is not applicable; any putatively
> proper subtype can be reconstrued as an improper one.
>
> > Or do you not have discrete levels at all? Just the whole sort of
> > general mish-mash?
> >
> > If you do have discrete levels, replace "final analysis" with "analysis
> > at a particular level".
> >
> > If you don't... then I'm amazed that we can have a conversation in any
> > language!
>
> I probably don't understand your question. Imagine a biological
> taxonomy, of genera, phyla, and so forth. If you can't see the root or
> the leaves of the taxonomic tree, you can't really identify levels.
"levels" was too loose, then.
My question is whether you perceive a "jump" between individual lions
and the kind 'lions' of a different kind from that between the kinds
'fierce lions' and 'lions'. I don't think it's actually a precise
question about the structure of the partial order... it's rather that
I'd split "subtype" into two relations - "instance of" and "subclass
of".
> >>>>> Worse, we have no obvious way to disambiguate to case 1 (with its
> >>>>> subtleties included).
> >>>>
> >>>> If it's a problem, it's not a problem specific to kinds or to {lo}.
> >>>
> >>> Do you seriously not consider such undisambiguable ambiguity a problem?
> >>
> >> I think it's not an actual ambiguity. It's a kind of potential
> >> ambiguity, in that if Z is referred to as an individual, in any
> >> further inferentially derived propositions in which X is instead
> >> conceived of as a generalization over subtypes there may be a scope
> >> ambiguity.
> >
> > Yes, something like that. An ambiguity which it takes a few steps to get
> > to.
> >
> >> This is not a linguistic problem.
> >
> > It's a problem which could be fixed by changing the language. In that
> > sense at least, it is a linguistic problem.
>
> Translating one metaphysics into another will generally yield
> problems. That cannot be fixed by changing the language. I surmise
> that you would like just one metaphysics for the language, and you
> would like it to be much more restricted than the most permissive
> sort.
If the permissive one could be handled without introducing the
(effective) ambiguities we've been talking about, I'd be happy.
> The objections to that are that it is metaphysically biased,
Why is that a problem?
> that the metaphysics conflicts with the one that others might want,
> and that it is hard to implement as the basis of the semantics of
> default gadri.
>
> >> "Not every mammal gives birth to live young" -- false for kinds, true
> >> for mundanes; but that doesn't mean "mammal" is ambiguous.
> >
> > So you'd say the statement is simply false, with the kind 'porcupines'
> > as a witness?
>
> I don't understand the question.
Does every mammal give birth to live young?
Martin
Attachment:
pgp5k7d2tT7_j.pgp
Description: PGP signature