Re: Tim v. Bijan use case (was Re: Proposed issue:...)

>After banging about with this with Bijan for a while today (very 
>inconclusively I'm afraid) I think we may have created a case for 
>which Tim and Bijan would claim a difference - I'm not sure of this, 
>but let's try -- it's essentially a slight expansion of the same use 
>case I suggested the other day:
>
>CONSIDER:
>
>Document T defines Twin as a Class which has max cardinality 2 on 
>property childOf.
>   It defines Family01Twin as an instance of the class Twin.
>   It also states A, B and C are all different individuals (As defined in OWL)
>
>
>Document A defines T:A, T:B and T:C as all different individuals,
>  it states T:A is a T:childOf  T:Family01Twin
>
>Document B defines T:A, T:B and T:C as all different individuals,
>  it states T:B is a T:childOf T:Family01Twin
>
>Document C defines T:A, T:B and T:C as all different individuals,
>  it say C is a T:childOf T:Family01Twin
>
>The DISAGREEMENT:
>
>Consider document D which imports A, B, and C   (but not T)
>  Is document D consistent or inconsistent
>
>  I believe Bijan would say Document D is consistent - the statements 
>inherent on this document (all and only on the page created by the 
>imports) contain nothing that is logically inconsistent.  The 
>cardinality on T:Twin does not "influence" document D at all.

I agree. We have to say this, if 'consistent' is to have any coherent meaning.

>  I believe that Tim would say Document D is inconsistent (or at 
>least that something was clearly wrong with it).

Those are different claims; lets keep things straight. There is 
clearly something 'wrong' here, but that's not the same as saying 
that D is inconsistent.  One thing that is wrong is that (D union T) 
is inconsistent.

>  Since documents A, B and C clearly chose to use the T: definition 
>of Twin, it should be assumed they are in agreement with it in some 
>sense,

Well, they didn't import the ontology, but they used the URI. Maybe 
they are trying to tell us something?

>therefore they've inherently agreed with the cardinality statement 
>(otherwise they could have used some other URI to define their 
>notion of Twin and ChildOf)

Naaahhh, wait a minute. I don't think that is fair or reasonable. 
After all, they COULD have imported T. Why didn't they? Maybe they 
had a good reason to use T's terminology without importing T itself. 
Maybe, for example, they want to make a coherent claim (so, not be 
internally inconsistent) which explicitly disagrees with T (so they 
do contradict T, ie them+T is inconsistent.) They might be 
disagreeing about the facts about A,B and C but agreeing with T's 
intended meaning of "Twin".  If we had been able to give them a more 
fine-grained kind of importing, or if T had been more careful not to 
mix ontologies with data, maybe they could have said this more 
clearly: but we didn't and T wasn't.

Your comment above assumes that an inconsistency is a disagreement 
about intended meanings: but it need not be. It can be a disagreement 
about brute facts like the number of kids in his family. The semantic 
resources of logic don't enable us to distinguish these cases, which 
both appear as a mutual inconsistency (though one can make some 
useful distinctions based on the kinds of assertion involved in the 
contradiction, in many cases, eg if you get my phone number wrong, 
its usually thought of an error of fact rather than a conceptual 
re-interpretation of the meaning of the concept "phone number".)

Maybe the semantic resources of the SW are not yet sufficient, alone, 
to resolve the exact nature of the mutual problem that A,B,C and T 
have here: maybe it will require some other techniques to resolve. 
But the SW is sufficient to *discover* this kind of issue (or maybe 
problem), which is I think all that we can expect to be able to do 
right now.  If this kind of thing happens a lot, and if$$ depend on 
getting it sorted out, then people will find ways to sort it out 
automatically, or at least quickly. If it doesn't happen often, then 
maybe the SW will get along fine without 'solving' it at all, just as 
it gets along with 404 errors every now and again.

>and thus having three of these on the same document makes that 
>Document inconsistent.
>
>Note that if A, B, and C import T, then both Bijan and Tim would 
>agree that document D is inconsistent, because the import would 
>cause D to be logically "committed" (by the owl imports definition 
>in S&AS) to the max cardinality of 2.

Quite. So (assuming the ABC guys are SW-savvy) maybe they had a 
reason for not doing that.

>My personal dilemma is that I could go either way on this example - 
>document D makes me uncomfortable (since A-C explicitly picked T: to 
>use) but it doesn't seem like it should be considered as 
>inconsistent as it should be in the imports case -- problem is being 
>"a little inconsistent" is a lot like being "a little pregnant" it 
>doesn't allow for much of a middle ground...

Quite. Look, it seems obvious that inconsistencies in closely related 
documents (eg one uses a vocabulary originating in the other) are a 
Sign of Something being Wrong. What exactly is wrong, however,  will 
depend on the context, in ways we cannot possibly predict. Maybe 
people are confused; maybe they just disagree; maybe one of them is 
trying to screw over the other; whatever. We can't legislate for 
that; all we can do is to say when there are inconsistencies, and we 
do that already. In other words, I don't see this example as 
illustrating any real *problem*, or at any rate not one that we are 
in a position to solve (someone might think that the fact that people 
can disagree at all is a problem, I guess.)

>  -JH
>p.s. Note there are many complex variants on this (what if A imports 
>T, but B and C don't, etc.) but I was looking for the easiest 
>example I could find where there was an explicit problem, all in the 
>use of URIs with no need to invoke natural language...

There are also cases which are more symmetrical, where A uses B:foo 
and B uses A:baz but neither imports the other, but C imports them 
both and is immediately inconsistent. In fact one could  generate 
arbitrarily long 'loops' like this where any (n-1) are consistent but 
all n together are not, and for any i, document #i uses terminology 
belonging to #'s i-1 and i+1, all modulo n.

Pat


>
>--
>Professor James Hendler				  hendler@cs.umd.edu
>Director, Semantic Web and Agent Technologies	  301-405-2696
>Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
>Univ of Maryland, College Park, MD 20742	  *** 240-277-3388 (Cell)
>http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***


-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes

Received on Monday, 6 October 2003 15:38:13 UTC