- From: pat hayes <phayes@ai.uwf.edu>
- Date: Wed, 5 Feb 2003 18:49:48 -0600
- To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
- Cc: www-rdf-comments@w3.org
>This is a continuation of my comments on what I consider to be a fatal flaw >in the RDF specification. I had submitted my views on this flaw to the W3C >RDF Core Working Group before the beginning of the Last Call period in the >message archived at > > http://lists.w3.org/Archives/Public/www-rdf-comments/2002OctDec/0297.html >but the working group chose to go into last call without addressing my >comments on this issue. Well, they were considered, but the WG came to different conclusions than you did. (However, in the rest of this message I speak for myself, not for the WG itself, which has not yet had a chance to discuss your email.) >What is the ``social meaning'' (Section 4.2 of RDF Concepts) of RDF? In brief, anything at all that someone or something might consider to be part of the meaning intended to be conveyed by the publication of some RDF. By its very nature, this does not admit of a tight formal definition. > Does >it have any relationship to how an RDF application should act? If by 'RDF application' you mean something which draws valid RDF conclusions, the answer is no. > If so, what >is this relationship and how can it be conveyed to an application? If not, >what business does this have in a document about RDF? See my earlier message. It is important to convey the point that the use of formal methods of inference does not automatically cancel or nullify any other aspects of meaning that might be conveyed by the RDF. >How does an RDF expression get to be asserted? Good question, I wish there was a good answer to that. > What syntax can I use to >assert RDF expressions There is no specific syntax. You just kind of put the RDF on a web page somewhere, and that amounts to publishing or asserting it, I guess. >, or to prevent their assertion? Can I use this >notion in OWL? Well, how does OWL get asserted or published? Or DAML? Jim tells us that there are a million lines of DAML code out there (or some other incredible figure). I presume it is all published or asserted or something like that, and isn't just electronic wallpaper. This issue goes beyond RDF. There are no agreed-on Web protocols for distinguishing different propositional attitudes, like assert or deny or disagree with or cast doubt on. The only thing we have that comes close is the 'imports' idea, interpreted as a kind of endorsement, but we havnt been able to formalize even that properly. Never mind agreed-on notations to express them. We are at ground zero on this issue. I am sure that they are needed, and will rapidly become so much needed that ad-hoc conventions will emerge spontaneously, but its not an matter that the SW working groups have paid very much attention to, as you know. > If not, then what good is it? Without any method given for >asserting an RDF expression or graph, what good is a paragraph that starts >``When an RDF graph is asserted in the Web''? > >How is social meaning determined? By society :-) But that is in fact a serious answer. >Does it have to be part of the RDF model >theory? No > Does it have to be part of an RDF graph? No > Does it have to be >accessible on the Web? Accessible in what sense? It is accessible in some sense, more or less by definition: if you put up something on a website that I read, and some content or meaning is thereby conveyed from you to me, then of course that meaning, whatever it was, was 'accessible' in some sense, at least to you and me. >Must it be common knowledge, and for what >community? Yes, no, it doesn't matter to the discussion. Presumably it is common knowledge to some community, or at any rate depends on some shared knowledge or beliefs. But we are straying into semiotics here; the point is only however you define it, it ought to be preserved through a formal entailment. > Must it be written down somewhere? No. > Can it exist only in >someone's mind? Hopefully more than one, but yes. >The idea that RDF graphs contain ``defining information'' that is opaque to >logical reasoners is ludicrous. Of course its not ludicrous. You just have to relax over the idea that there might be other information than that described by the formal model theory. In fact, RDF itself is so weak as a logical formalism, that it strikes me as ludicrous to assert the opposite. If all I can convey is what can be completely axiomatized in the positive binary fragment of existential conjunctive logic, then I really can't say very much at all about anything. > An RDF graph is simply a set of RDF >triples. But a published RDF graph is a set of RDF triples which has been created and made publicly available by some agency at a certain time and probably for a certain purpose, and all these (and other) aspects of the surrounding circumstances may be relevant to someone or some community or even some (non-RDF) software. > It is certainly possible that there can be communities that have >intended meanings for these RDF graphs, but these intended meanings are >external to the RDF graph External in what sense? They are not conveyed by the RDF MT, but they are certainly associated with the graph (or more accurately, the document which contains a token of the graph in some concrete syntax) in some sense. They are external in the sense that they are not defined by the RDF specs, indeed, but then a lot of stuff is like that. >, and, indeed, external to RDF as a whole, and >thus have no place in a normative part of a document about RDF. I disagree. RDF is not a pure formalism: it is a formalism intended for use in a public domain; its function is to convey and transmit meaning. "Transmit" in this sense assumes a transmitter and a transmittee, and who knows what kind of surrounding context. Surely our goal is to have RDF (and OWL, etc) deeply integrated into the social fabric, being of actual use. And we ought - I think we have a duty to - say as much as we can about the intended conventions that are supposed to be used when handling RDF in such rich contexts. Which is what the stuff in the Primer and other documents that you are objecting to tries to do. >What social conventions surround the use of RDF? Even if there were some, >why should they make their way into a normative section of an RDF document? >The idea that some owner of a URI reference can control the use of that URI >reference goes counter to the bedrock goal that RDF allows one to say >anything about anything. I think the discussion has gone into the stratosphere at this point, but I don't see how this follows. And in any case, nobody is saying anything about 'controlling the use' of urirefs, only of using them to convey social/informal meaning. > The RDF model theory contains no hint that any >of these sorts of restrictions are possible. Their connection to the model theory is only that they should respect it, so that formal consequences retain any informal meanings or intentions which might be associated with any vocabulary which gets used. >The example in Section 4.5 of RDF Concepts brings forward these problems. >The document at http://skunk.example.org/ does not entail anything >derogatory about C:JohnSmith, It depends on what you mean by 'entail'. It does formally entail something which might be interpreted as derogatory, although that interpretation itself is of course not a formal entailment. >which is reinforced in the section just >above. This being the case, there is no reason for any notion related to >RDF to bring this forward. > >If, however, the opposite was the case then there would be no way for any >organization to deploy any RDF-based application. On the contrary, I think that some such assumption is essential, and that without it there would be no point in deploying any RDF-based application. Without this, the only possible use of RDF would be to be as input to RDF engines which would produce more RDF. Surely you want it to be the case the the use of RDF inference does not erase or cancel any meaning which the RDF was being used to convey? Why should it? What purpose is served by claiming that it does? > Such applications would >not be able to understand the social meaning of the RDF they created or >manipulated That is probably true: so what? A database server doesn't understand what 'payroll' or 'employee' mean, but that doesn't prevent it doing something useful with such labels. Most useful software manipulates data without understanding its meaning; certainly without fully understanding all its possible meanings. My web browser has no idea what I am reading when I look at its display. Peter, you and I both have a background in AI/KR, so I think I know where you are coming from. We both have been steeped in the need to avoid the gensym fallacy and the concomitant dangers of thinking there is more in one's KR than there really is there, and the use of an MT to provide the needed rigor to resist such errors. But that is all to do with modelling belief: representing the private mental state of a believing agent. The SW really is a different situation. RDF isn't just going to be used by agents to think private thoughts with, it's not a Fodorian Language of Thought; if anything, its more like a language for agents to talk to one another with. You know the classic 'grounding problem' for formal KR semantic theories? Well, RDF in use is grounded by its surrounding context of use, and it may be only a small part of something much larger, which is representing other information in other ways. Think of RDF as more like a simple, formalized, sharply defined "natural language" for software agents, something whose chief function is for communication, not for thinking with; and then observe that the software agents are also working in a context which involves human and social 'agents'. We really do not know what aspects of meaning might arise in the uses of RDF in such contexts, and we don't really need to know: but we DO need to say, normatively, that whatever they are, they ought to at least *respect* the minimal constraints on meaning described by the formal MT, so that the use of inference processes which depend on these constraints does not destroy or distort these social or contextual aspects of meaning. And this is a real constraint, not just a form of words: for example, RDF really is monotonic, and that imposes some nontrivial conditions on *any* notion of RDF meaning, social or otherwise. >, and thus could easily create documents holding the >organization liable for just about any imaginable consequence. The liability would be determined by the same social/commercial/legal rules and conventions that govern normal human intercourse already. The point at issue is only that the use of RDF inference somewhere in the overall process should not be seen as cancelling or nullifying the normal machinery of human communication (including communication via the Web.) So RDF entailment can't possibly create new liabilities out of a vacuum, but it can transmit liabilities which would have been present anyway. You can't hide from your liabilities by saying: the formal RDF inferences cancelled all that social stuff. Seems fair enough to me. >In this >case I would have no choice but to tell Lucent Technologies not to deploy >any RDF applications. Well, that would be Lucent's loss, but I think you would be over-reacting. Pat -- --------------------------------------------------------------------- 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@ai.uwf.edu http://www.coginst.uwf.edu/~phayes s.pam@ai.uwf.edu for spam
Received on Wednesday, 5 February 2003 19:49:54 UTC