W3C home > Mailing lists > Public > public-sw-meaning@w3.org > June 2004

RE: No Standard Semantic Web Pragmatics?

From: John Black <JohnBlack@deltek.com>
Date: Wed, 16 Jun 2004 09:43:48 -0400
Message-ID: <CBEA695878CA104ABC6E74C6B176927507BD3F@DLTKVMX2.ads.deltek.com>
To: "Jim Hendler" <hendler@cs.umd.edu>, "Pat Hayes" <phayes@ihmc.us>
Cc: "Sandro Hawke" <sandro@w3.org>, "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>, <public-sw-meaning@w3.org>

> From: Jim Hendler
> Sent: Monday, June 14, 2004 7:07 PM
> 
> >>However, another key idea grew about the same time -- as long as we 
> >>are using URIs, we could make those URIs dereferencable -- that is, 
> >>we could look and see if there is a document there, and if so, 
> >>perhaps that document could describe the link -- RDFS and OWL 
> >>provide vocabularies that live at those links and provide 
> >>information about the "intent" of those relationships.
> >
> >No, that is exactly what they do not provide. That is John's point: 
> >there is a gap here precisely because the SW notations only express 
> >CONTENT, they do not express INTENTION.  The stuff about 
> >performatives in the paper I helped write was intended to be a step 
> >towards bridging this gap, since performatives in natural language 
> >are exactly where an intention is expressed unambiguously by stating 
> >- describing - the intention. If enough people say that Jack and 
> >Jill are married, in the right way and under the right 
> >circumstances, then they are married. If I say "I promise to buy you 
> >lunch" then an actual promise got created: I performed a social act 
> >by saying that I was performing it. Very handy, that is: it gets you 
> >from mere descriptions (which we indubitably have in RDF and OWL) to 
> >actual intentional actions: it gets assertings (denials, explicit 
> >non-assertings, endorsements, whatever) actually done, and in a 
> >publicly checkable way rather than being left implicit.
> >
> >>
> 
> Ah, right right right -- I knew you logician types have this bug up 
> your butts about "intent" -- and that is why I put it in scare 
> quotes, but I should have made it even clear i was using the term in 
> an informal and not a technical sense -- however, that said,I think 
> it is this hangup with "intent" somehow being a mysterious thing that 
> is largely to blame for our lack of progress on this social meaning 
> stuff

Could you please elaborate on this attribution of blame to "this hang-up 
with 'intent'"? I'm really curious about this. Since it is my intention 
to achieve progress on this issue, I don't want to stumble into a 
position where I am to blame for the lack of it. (You could leave out 
the scatological ad hominem attacks and slurs against professional 
groups unless you think that is critical to your point)
 
> Consider, if you go to my HTML web page you will see a link to a page 
> that has pictures of my daughter.  You have no way to know what my 
> intent was in putting them there -- you can guess at some 
> possibilities (wrong, I was not trying to raise the price that would 
> be offered for her on the black market) but you can't know what I had 
> in mind.  If I had labeled those photos in RDF, OWL, or KIF there is 
> no reason why I would have told you one whit more about my intent -- 
> I might have made it unambiguous that I was averring that the person 
> depicted in the photograph was one with whom I had the familiar 
> relationship of type daughter -- but you still wouldn't have any 
> intent behind it.

Well to the extent that I don't know what your intent was you didn't 
communicate. There is nothing wrong with that. Its sounds like what 
you intended. I can still view the pictures. 

But I disagree with you that I cannot recognize your intent if you 
take pains to communicate it to me. Nor am I talking about any 
mysterious supernatural mind-melding, telepathy, or otherwise 
peering into the interiority of your consciousness. No, I am  
talking about one of the most extraordinary properties of human 
communication, "...If I am trying to tell someone something, then 
(assuming certain conditions are satisfied) as soon as he recognizes 
that I am trying to tell him something and exactly what it is I am 
trying to tell him, I have succeeded in telling it to him. Furthermore, 
unless he recognizes that I am trying to tell him something and what 
I am trying to tell him, I do not fully succeed in telling it to him." 
-John R. Searle, Speech Acts, p.47

And again using "intention" explicitly, "In speaking I attempt to 
communicate certain things to my hearer by getting him to recognize 
my intention to communicate just those things. I achieve the intended 
effect on the hearer by getting him to recognize my intention to 
achieve that effect, and as soon as the hearer recognizes what it is 
my intention to achieve, it is in general achieved. He understands 
what I am saying as soon as he recognizes my intention in uttering 
what I utter as an intention to say that thing."
-John R. Searle, Speech Acts, p.43

> So if we simply can argue w/respect to what the links state (in the 
> factual sense) and attribute their ownership based on where they are 
> asserted (oops, I mean where the bit stream defining them is found - 
> since assertion is another bug up the arse) - then perhaps we could 
> have made some useful progress on stating "what's in a link"
> 
> 
> >>And technically, that is the heart of the Semantic Web architecture 
> >>- links that can be named and described more formally.
> >>
> >>IMO, the social meaning issue arises from the fact that we have 
> >>both referencing and dereferecing going on.  When links share a 
> >>URI, and there's no document at that URI to dereference, then it is 
> >>clear that any meaning of that term is in some sort of off-line 
> >>"Social" conventions between the users thereof.  However, when we 
> >>add the dereferencing it becomes trickier -- because now we have to 
> >>ask if use of the term in some way "commits" to what is in the 
> >>dereferencing document, if the owner of that document controls the 
> >>use of the term, etc.
> >>
> >>There's lots of other "social meaning" issues on the Semantic Web, 
> >>and the threads on this list talk about many of them, but in my 
> >>mind the key ones are those that arise from the issue of the 
> >>relation between the named terms and the documents that describe, 
> >>in some formal way, the use of those terms
> >
> >Well, yes, but (IMO) only because such dereferencing is the only way 
> >to establish any kind of link between a URI and anything that can be 
> >plausibly attributed agency. In order for a promise (etc) to be 
> >done, there has to be an agent doing the promising. Similarly for 
> >asserting, denying, etc. . Without agents (and I really do mean 
> >SOCIAL agents, not software agents) in the picture, all we have is 
> >sentences being looked at: nothing is asserted at all.

And I would add, defining and naming. Unless we also have an agent 
acting with the intent and authority to define or name, all we have 
is even more sentences. Without it, we are stuck in a room with a 
Chinese/Chinese dictionary, trying to ground our symbols in an 
endless series of more symbols. I really think this may be the 
Grand Synthesis I predicted in our last telecon. We have been thinking 
of URI dereferencing as being a means of adding in more formal 
propositions and descriptions to our groundless lists. Instead, 
URI dereferencing could be the source of AGENCY, INTENTION, AND 
AUTHORITY that really could be the basis of an automated communication 
system with reasoning.

However, I don't want to alarm anyone with this word, AUTHORITY, with 
its connotations of condign power and all. So I came up with a 
replacement, UNDERWRITER. To say your terms are underwritten by 
an impressive agent would be to say that agent provides a warranty 
promising to compensate you if you use a term they have underwritten 
to mean one thing and you lose money because someone misunderstands 
you to mean something else. It gives a whole new meaning to Humpty 
Dumpty's strategy to have his words mean what he pays them to mean. 
That said, I am completely serious about this.

> well, we humans seem to be social agents who handle this assertional 
> stuff just fine - we know how to differentiate (at least in 
> principle) between "what Pat said" and "What I think Pat meant" - and 
> we usually conditionalize the latter in civil discourse -- seems to 
> me our Sem Web agents could do the same and we could move on to 
> actually looking at this use of dereferencable URIs as something that 
> could add a lot of power to the SW if we had some social 
> conventions/expectations.
> 
> (i.e. I claim I am legitimized to believe that Peter believes that he 
> is a perfect being by dint of stating it on his page -- I may be 
> wrong, but then I'm usually wrong about what Peter believes, so 
> what's different about this?)
> -- 
> Professor James Hendler			  
http://www.cs.umd.edu/users/hendler
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)
Received on Wednesday, 16 June 2004 09:44:02 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:42:16 GMT