Re: Social meaning discussion 6th March

Comments below.  Generally I agree with DanBri's concerns.


Dan Brickley wrote:

> * Brian McBride <bwm@hplb.hpl.hp.com> [2003-02-27 10:34+0000]
> 
>>Sans chapeau:
>>
>>My bath time this morning was spent thinking about social meaning.  I came 
>>to the conclusion that 'meaning' is a difficult and slippery a concept that 
>>we should try to stay away from, sticking to things that are more 
>>concrete.  We should leave talk about 'meaning' to the 
>>philosophers.  Perhaps we can get all we need by describing intended use.
>>
>>And then I see this is exactly what Jeremy has done in his draft 
>>alternative text.
>>
>>  http://www.w3.org/2001/sw/meetings/tech-200303/social-meaning
>>
>>This is a significant shift in approach that may have considerable merit, 
>>and I wanted to draw attention to it.
>>
>>The mininalist approach looks good too, though we might go a bit further:
>>
>>  1 An RDF triple is intended to be used to represent an assertion
>>  2 A set of RDF triples is intended to be used to represent the 
>>conjunction of the assertions represented by each individual triple
>>  3 The assertion represented by an RDF triple is determined by the 
>>specifications of its subject, predicate and object.
>>
> 
> Happy up to here, though I'm not sure what 'specification' of a bNode or 
> literal is, and hence have growing sympathy for TimBL's predicate-centric
> view.


A couple of points:  First, this interrelates with this "statement" vs. 
"triple" terminology discussion we've had before.  Is "assertion" the 
same as a "statement"?

Second, I think what #3 is trying to say is something along the lines of 
"the *meaning* conveyed by an RDF triple is *partially-determined* by 
the *meanings* associated with its subject, predicate, and object" 
("partially-determined" because, at least when considering "social 
meaning", the context also helps determine the meaning).


> 
> 
>>  4 RDF triples representing false assertions SHOULD NOT be used to mislead.
>>
>>The above being non-normative.
>>
>>I'm not happy with 4.  RDF triples representing true assertions SHOULD NOT 
>>be used to mislead either, yet "RDF triples SHOULD NOT be used to mislead" 
>>is vacuous.  So maybe strike 4, though I'm inclined to leave it in.
>>


I'm not happy with #4 either, primarily because it seems to be giving 
advice in an area that's none of our business.  The point of this 
section, it seems to me, is to point out some of the potential 
consequences of RDF being used for its intended purpose of conveying 
"social meaning".  We can reasonably (it seems to me) point out that RDF 
will be used to convey such social meaning, and if it is, there are 
certain consequences of using it in that way, e.g., society may 
determine that there are legal ramifications to some such uses (we're 
not saying there will be such ramifications, let alone specifying them, 
we're just observing that, given this use, such things may follow). 
Similarly, we can reasonably say that people may use RDF descriptions to 
mislead either people (just as they use English descriptions to mislead) 
or software (just as they use inappropriate META tags to mislead search 
engines).  Again, this is not a specification on our part, it's simply a 
consciousness-raising observation on our part about some of the 
potential consequences that may ensue when RDF is used in its intended 
application area.  I believe we can reasonably make such comments, as 
long as it's clear that we're not *specifying* these things.

 
> 4. is so out of scope it's dissapeared over the horizon! This would be like
> telling the world that text/html SHOULD NOT be used to mislead, or that
> image/svg SHOULD NOT be used with nudie pictures. Modern capitalism, the 
> entire advertising industry and the business models of various companies not 
> a million miles from W3C... are all based on misleading people. That's just 
> the way the world works. If we want RDF to be part of that world, we'll have 
> to save the noble stuff for a follow-on working group. False, or False-ish RDF 
> statements _will_ be used to mislead people. That will be a metric for 
> measuring our success and the success of the Semantic Web. We can solve that 
> problem once we've finished creating it, but the solution won't look like (4).
> 
> In more practical terms, I think you'll have trouble defining 'mislead'
> without some model of intent; or 'used' without a story about deployment 
> models. If I republish some out of date RDF via my FOAF aggregator and 
> it confuses someone, did I break your 4th law? How about if all I did was 
> provide the HTTP cache they used to access it? What about if my cache 
> also embedded an RDF parser? ...and also computes RDFS closures? etc.
> 
> Backing off from 4., I'm v happy with the focus on 'assertions' rather than
> on the woolier notion of meaning. We should concentrate on the notion of 
> RDF documents being the sort of thing that can be true (or false) descriptions
> of the world, and from that naturally focus on the characteristics of those
> documents which determine the arrangements of the world that would satisfy 
> them. 
> 


I don't understand how you can talk about "assertions" (having its 
ordinary connotation) without getting into "meaning".  "Statement" seems 
more neutral.

--Frank




-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-875

Received on Thursday, 27 February 2003 09:33:56 UTC