[Top] [All Lists]

Re: [ontolog-forum] Ontology, Information Models and the 'Real World': C

To: Pat Hayes <phayes@xxxxxxx>, Ontolog Forum <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Waclaw Kusnierczyk <Waclaw.Marcin.Kusnierczyk@xxxxxxxxxxx>
Date: Sat, 02 Jun 2007 11:23:16 +0200
Message-id: <46613704.4050609@xxxxxxxxxxx>
Pat Hayes wrote:
>> Pat Hayes wrote:
>>>  This message, and indeed this sub-thread,
>>>  illustrate perfectly what seems to me to be the
>>>  key advantage of using a 'context ontology' (as
>>>  IKL does: that is, treating contexts as objects
>>>  and making non-contextual assertions about them)
>>>  as opposed to a 'context logic' (that is, a logic
>>>  in which assertions are understood as being made
>>>  in a context and interpreted there using
>>>  contextually local criteria). In a word, contexts
>>>  in a context logic make meaningful things
>>>  meaningless.
>> I have an uncomfortable feeling that IKL prevents me from making some
>> simple statements, or that it forces me to do it in an awkward way.
> Perhaps. But it is after all only a version of classical first-order 
> logic. Much of what you say below would apply to any classical logic.
>> The sentence
>> (sleeps pat)
>> means
>> (that (sleeps pat))
>> i.e., its meaning is the proposition that Pat sleeps, simply,
>> context-independently, eternally -- which is nonsense.
> Its not nonsense, but it is false.    (01)

Yes, that was what I meant -- sorry for the inappropriate expression.    (02)

> But there is something wrong with your exposition here. "simply, 
> context-independently" is not the same as "eternally", ie "in all 
> contexts". Simple logical truth is not contextual *at all*. It has 
> nothing to do with contexts, just as 2+2=4 has nothing to do with times 
> or tenses.  To be context-independent is not the same as being true in 
> all contexts. It puts the sentence or proposition into a completely 
> different category. The actual formal language is different. Classical 
> logic is not contextual in any way. It is not a schema for generalizing 
> over contexts.    (03)

This does not sound convincing to me.    (04)

>>  Let me want to
>> say that Pat sleeps during some specific interval, say int.  If I make
>> the statement
>> (ist int (that (sleeps pat)))
>> what I say is that the (simply and eternally false) proposition that Pat
>> sleeps, simply, context-independently, eternally is (the proposition) in
>> the relation ist with the context int.  But what does that mean?
> I have no idea. Im not sure I can even parse what you say here. But in 
> any case I would not recommend this way of representing it, myself. I 
> would follow a long tradition in KR and relativize the relation to the 
> time, ie treat it as a fluent:
> (sleeps Pat int)    (05)

But this, following IKL guide, is syntactic sugar for (ist int (sleeps 
pat)), and there we are again.    (06)

>> Instead of making a statement about Pat's sleeping during int, I make a
>> statement that some proposition -- which I already know is simply false
>> -- stands in a relation to an interval.  Pat suggests -- insists, in
>> fact -- that (ist c p) does not mean that p is true,     (07)

except for that he has just said, in the preceding mail, that (ist c p) 
does speak of the truth of p.    (08)

> it just means that
>> p ist-stands to c, which incidentally is read out as 'p is c-true', and
>> which has nothing to do with p's truthness.
> Right. Which is one reason why it is    (09)

If this is right, could we stick to one version only?    (010)

>> So perhaps I would need to say
>> (sleeps-during pat int)
> Yes. Or, an alternative formulation,
> (sleeps (pat int))
> or perhaps
> (sleeps (during pat int))
> where during is a function from spatiotemporal entities and times to 
> temporally restricted spatiotemporal entities ("pat during the interval 
> int"). Both versions can be made to work coherently, but it is probably 
> a mistake to get them confused with one another.    (011)

Yes, taking spatiotemporal slices is a solution.    (012)

>> for which I need to introduce the predicate sleeps-at, or maybe even
>> (sleeps-during-int pat)
> No, you want 'int' to be a genuine argument. But why change the relation 
> name?    (013)

No need, with the above.    (014)

>> ?
>> If we deny that (ist c p) means that p (somehow) is true when c is the
>> case (by which I mean *is true* in c, not is-true-in-c) -- given that it
>> is just false, immutably -- then (ist int (that (sleeps pat))) cannot
>> mean that Pat sleeps during int
> Sure it can. In fact, that is exactly what it is usually interpreted to 
> mean. But notice that the inner '(sleeps pat)' sentence is now being 
> interpreted in a context-relative way.     (015)

This is what is not obvious from IKL docs.  What I read there, is that 
the sentence (sleeps pat) expresses *the* proposition (that (sleeps 
pat)), which (the proposition) is taken to be true-in-context 
(irrespectively of whether it is true or not).    (016)

If (sleeps pat) has no indexicals, and ikl is transparent -- both sleeps 
and pat are the same within and out of any context -- then how (sleeps 
pat) can can be interpreted in a context-relative way?  And, if it were, 
I'd like to think that (sleeps pat) interpreted without a context and 
(sleeps pat) interpreted differently within a context -- so that one is 
true and the other is false, say -- refer, in fact to different 
propositions;  so that it is the sentence (sleeps pat) which has a 
context-dependent truth (sounds very reasonable to me) rather than *the* 
proposition (sounds somewhat unreasonable to me).    (017)

BTW, I note that in the IKL guide, which describes a language in which 
names are carefully distinguished from what they name, there is some 
dose of use-mention confusion (I am sure this is due to colloqualism 
rather than due to any underlying confusion).  For example,    (018)

" ... the indirect name (tnb 'The Prince') refers to "The Prince" ... "    (019)

is, strictly speaking, incorrect (wrt. the apparent intentions):  for 
consistency, the two strings    (020)

(tnb 'The Prince')    (021)

and    (022)

"The Prince"    (023)

should both be taken as names -- and then (tnb 'The Prince') refers not 
to "The Prince", but to what "The Prince" refers to, while "The Prince" 
is referred to by 'The Prince',    (024)

or should both be dereferred -- and then it is not (tnb 'The Prince'), 
but rather (tnb 'The Prince')'s name that refers to "The Prince".    (025)

I think there are more such examples in the text.
Call me what you want, but I think it would be useful to avoid this sort 
of loose talk, for the sake of clarity.    (026)

> It is not being treated in the 
> same way that it would be if it were simply a bare IKL sentence. 
> (Remember that IKL is not a context logic. IKL assertions are not made 
> "in" a context: they are just made.) It does not predicate a property of 
> pat: it predicates a property of pat *in a context*. This is a 
> *relation* between pat and the context, not merely a property of pat. 
> That is why its logical truth (as a simple predication) has no real 
> connection to its contextual truth: they are simply different meanings. 
> There is a whole topic of how to translate a contextual assertion into a 
> noncontextual one, eg by translating
> (ist C (that (P a b)))
> into
> (P a b C)
> and
> (ist C (forall (x)(P x)))
> into
> (forall ((x C))(P x C))
> where the C parameter name is being used also to refer to the universe 
> of discourse of the context (such multiple uses of a single name are 
> allowed in CL and IKL)
>> , because it is neither about Pat's
>> sleeping, nor about (that (sleeps pat))'s truth.
> Not about its logical truth, no. But surely you would expect that: you 
> are the one who wants pat to be sleeping in some intervals and not in 
> others. If this referred to logical truth, how can it *possibly* be both 
> true and false? In a logic which does not have tenses, and is not 
> indexical, and is not paraconsistent?    (027)

No, the same proposition can't, as argued and seemingly agreed earlier, 
be true *and* false.  But a sentence can;  the same sentence can be true 
in one context, and false in another, just because it refers to 
different propositions in different contexts, and not the same 
proposition which magically changes its truth value, or its 
truth-in-a-context value.    (028)

>> Can I say that Pat sleeps during int?
> Of course; in fact you can say it many different ways. (see for example 
> slide 14 of 
> ) But everything is clearer if you simply avoid contextual talk altogether.    (029)

thanks.    (030)

>> (I must be deeply wrong about IKL, I guess.)
> I think you are having two problems. First, you seem to be thinking 
> essentially contextually, but IKL isn't a contextual language. Second, 
> you seem to be being misled by the idea that truth-in-a-context must 
> refer to what the logic considers to be truth,     (031)

no, this is what has been clear enough;  this is precisely why i object 
to reading (ist int (sleeps pat)) as asserting anything about pat.    (032)

> perhaps because there is, 
> or should be, only one basic notion of truth. But this does not work for 
> IKL, since IKL provides not a context logic but (at best) a context 
> *ontology*, i.e. a logical *theory of* contexts. Since 
> truth-in-a-context is being modelled by this theory, we are free to 
> specify how it is realized: and the IKL strategy is to say that P is 
> true-in-C just when (ist C P) is logically true. I would add, this 
> corresponds very closely to what John McCarthy, who introduced the 'ist' 
> notion and the use of context logic, says (ist C P) should mean.
> Hope this helps.    (033)

I'll do my best to make it help me.  Thanks.    (034)

vQ    (035)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (036)

<Prev in Thread] Current Thread [Next in Thread>