RE: Proposed issue: What does using an URI require of me and my s oftware?

>  > -----Original Message-----
>>  From: pat hayes [mailto:phayes@ihmc.us]
>>  Sent: Friday, October 03, 2003 1:48 PM
>>  To: LYNN,JAMES (HP-USA,ex1)
>>  Cc: public-sw-meaning@w3.org
>  > Subject: RE: Proposed issue: What does using an URI require
>>  of me and my
>>  s oftware?
>  >

<<SNIP>>

>  >
>  > >
>>  >The reason I digress to these somewhat simple examples is
>>  that it seems to
>>  >me that once the tools evolve, these judgement calls will be
>>  made not by
>>  >computer scientists, but by businessmen, supervisors, etc. I
>>  suppose we
>>  >could build some kind of semantic consistency checker into
>>  the tools; MS
>>  >Word RDF/OWL Check?
>>  >
>>  >Naively curious,
>>  >James
>>
>>  Not naive at all, right on the button. Like, what
>>  problem are we setting out to solve here? What
>>  might go wrong that our declarations of Policy
>>  and Correct Architecture and so on are aiming to
>>  prevent? I for one am completely unclear what the
>>  issues are supposed to be that so concern us
>>  here, and I am extremely worried that we will
>>  make declarations based on mistaken ideas about
>>  meaning rather than on any actual problems.
>
>Ok. ACorp creates a acorp:uri123 which is a serial
>number of one of its acorp:StandardWidget, which
>is the product ID of its standard widget and has property
>listPrice = $2.00 according to its ontology acorp:catalogue.
>BCorp, thru their sw-agent, buys a batch of these including
>acorp:uri123.  Now BCorp turns around and sends the batch to
>CCorp's sw-agent with an RDF invoice that states that
>acorp:uri123 a ACorp:DeluxeWidget.  CCorp can verify that
>the list price of a ACorp:DeluxeWidget is $10.00 and happily
>pays BCorp their asking price of $5.00.
>
>Now the RDF invoice used two of ACorps URIs to
>commit fraud.  Those URIs belong to ACorp and it was never
>ACorps intention that acorp:uri123 be called anything other
>than a acorp:StandardWidget.  How could ACorp make this
>clear to CCorp?  One solution would be to publish at
>acorp:uri123 the statement, this is <> a acorp:StandardWidget.
>
>Note that this is a boring, trivial example.  There is no
>inference, semantic search, or other sw-interesting ideas
>in it.  I'm using it to point out that URIs have
>social meanings that will become represented and
>communicated by the Semantic Web.

OK, several observations.

First, as you say, this example has got nothing to do with the SW as 
such: it is about error or maybe fraud, pure and simple. In fact it 
has nothing to do with the Web at all. You could run it unchanged 
without mentioning URIs; the only point is that ACorp used a name 
with a declared meaning which was then misused by BCorp.  What makes 
this example have teeth is not that the name was a URI, but that 
ACorp said it had a specific meaning in a commercial context, and 
then the rest follows from general considerations of legal use, 
publication, misrepresentation, etc. (and maybe copyright, for all I 
know) which apply throughout human society. So I don't see what 
relevance this has for our topic or why it is our business 
particularly to say anything about it: in fact, since the relevant 
social conventions here are ultimately determined by the courts, it 
is kind of silly for us to even set out to make new law in this kind 
of an area.

Second, if you were to elaborate the example a little so that a piece 
of SW reasoning intervened in the connection: say, that BCorp had 
published an ontology which asserted that  ACorp:StandardWidget 
rdfs:subClassOf ACorp:DeLuxeWidget, and then C had used this to come 
to his own conclusions about prices, then it gets a bit more 
interesting.  We have discussed examples like this in the working 
groups. I think that it is reasonable to say that in cases like this, 
the SW reasoning processes are understood to *preserve* any intended 
'meaning' that might otherwise, on general social/legal/whatever 
grounds,  be presumed to be associated with the terms used. So in 
this case, C is not making any kind of SW *semantic* error in 
presuming that his conclusion about the price of standard widgets 
really does mean what it seems to mean, ie it really is talking about 
those widgets and about real dollars. Now, of course, that is not to 
say that it is *true*, only that the process of running an SW 
reasoning engine doesn't as it were strip off any meaning that the 
terms might be said to have had on other grounds.  Or, to say the 
same thing less obscurely and without using the M-word, any process 
which uses the URI after it has been processed by a valid SW reasoner 
should be able to treat the URI in the same way that it could have 
been treated before it was processed by the reasoner. In this case, 
of course, C has now been misled, and I think that the general 
presumption about SW inference and the same general commercial/legal 
considerations about publishing would have little trouble pinning the 
blame for this misapprehension on B rather than on C or A; but that 
isn't our concern. Our concern is only to ensure that BCorp's lawyers 
can't mount an ingenious defense along the lines: an OWL reasoning 
engine was used, so the conclusion only has a formal meaning, because 
the OWL MT doesn't mention widgets or US dollars. I agree that would 
be wrong; but the W3C Amicus Curae brief could point out that the MT 
refers to entities, and widgets and dollars are both entities, so....

BTW, it would be reasonable, even in cases like this, to say that if 
ACorp had been so sloppy as to publish an ontology that allowed 
people like B to say things like this *without thereby causing an 
inconsistency* then maybe they were partly liable for this mess, by 
virtue of carelessness. They could have put out an OWL ontology which 
explicitly said that the DeLuxes weren't Standard, and then any SW 
agent worth its salt could have immediately spotted the problem with 
B's ontology.

Third, a general point: in order to support all of the above, we need 
to say very little. Most of it is implicit in the model theories 
already published (not the details, just that general approach to 
semantics.)  We don't need to say anything weird or controversial, 
such as that URIs have unique meanings, or that the meanings must be 
determined by a unique ontology published by the owner, or that 
properties mean more than names. All we need to do is to say that URI 
meanings, in any sense of 'meaning' you care to use, are understood 
to not be *changed* by drawing valid conclusions from ontologies, and 
that if two ontologies contradict one another then normal 
considerations should be used to try to determine which of them 
should be believed more than the other, or is more suitably used than 
the other for a given purpose (such as ordering widgets). (I don't 
even think we need to say this, myself, but it would do no harm to 
state the obvious.)

Fourth, your example is actually quite subtle when examined in 
detail.  If the service B is offering is, say, a retail outlet 
serving a different clientele than the wholesale outlet provided by 
A, then there may be no fraud or misrepresentation involved here, 
only a different pricing structure. The world provides many such 
examples. It is perhaps a little careless of B to use A's terms in 
stating its own pricing, but I see no reason why we have to get 
involved in meddling with social/commercial details like this, and I 
certainly don't think that this is anything remotely like a matter of 
SW *architecture*.

Pat Hayes


-- 
---------------------------------------------------------------------
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 Friday, 3 October 2003 20:06:48 UTC