- From: Pat Hayes <phayes@ai.uwf.edu>
- Date: Thu, 4 Oct 2001 15:32:47 -0500
- To: "Seth Russell" <seth@robustai.net>
- Cc: www-rdf-logic@w3.org
> >It seems to me, and you seem to have confirmed this, that this document will >tell us, in no uncertain terms, what circumstances must obtain for a RDF >graph to be true or false. Is that a fair summary ? ... and may I >proceed as if it were ... Yes. that is pretty much exactly what the MT does, with admittedly a rather abstract notion of 'circumstances', ie an interpretation. >But it seems to me that whether a RDF graph is true or false Wait. That doesn't make sense. True or false in what interpretation? The MT does not simply assign truthvalues to graphs simplicter; it gives the rules for determining the truthvalue of a graph *in a given set of circumstances*, as you just said. The most interesting use of this is precisely to turn your point upside down; the issue is not whether a graph "is" true or false, but what constraints are put on the world - the "circumstances" being described by the graph - in order that the graph shall come out to have the value 'true'. Because to assert a piece of RDF, just as to assert a piece of FOL or of English, is to *claim that it is true*. So in making such an assertion, one is making a claim about the ways that the world can possibly be; restricting the collection of possible obtaining circumstances to those that make one's assertion true. Of course, most realistic circumstances are far more complex than could be completely described in RDF, so other things will be true as well, and what your RDF graph asserts about reality is only one rather weak constraint on the entire cosmos; nevertheless, it might be a very useful constraint to someone or something that shares with you an interest in those particular aspects of the circumstances that obtain. >may not have >much bearing on how a well formed graph affects its author or its audience. Well, to repeat an earlier point, there is indeed no way to determine how any datastructure might affect every piece of possible code; but I would submit, that if you or your program is reading RDF and being affected by it in some way that has no bearing on its truth, then you or your agent are very likely to misunderstand the intentions of whoever put that RDF onto a web page in the first place. (To see why, change "RDF" to "English" in the above and imagine applying it to, say, CNN.) >As you well know, there are some of us who seek to adopted radically >different epistemologies. In my own words let me briefly attempt to sketch >mine: which, as far as I can see, holds the truth of a RDF graph to be >somewhat irrelevant .. or if not irrelevant, at least relegated to the >status of a propositional attitude. Suppose that each agent (human or >cybernetic) carries around a grab bag of tools ... and that these tools in >these bags create an intricate network of possible interactions. Agents >transmitting messages from bag to bag cause changes that ripple around the >global network .. and in this environment: which tools work, will survive >in the bags. Analyzing what works in terms of two values, {true or false}, >becomes too complex, too inexact, and ends up being unnecessary. Well, you are free to pursue this philosophy, of course, but it seems to have nothing to do with either the WW web or the semantic web. It sounds more like a kind of Darwinian web. I have to say, that I would think very hard before putting any money in a bank run on these principles. We humans have come quite a long way since being paramecia, and it seems to me to be more productive to extend our cognitive apparatus into the semantic web as far as possible, rather than return to a primordial soup and re-run evolution in simulation >I don't expect you or the W3C to adopt that philosophy ... but by the same >token I don't expect you to rule it out by what you write in our >specifications. Tough luck. But in any case, if you are uninterested in truthvalues or interpretations, and consider them unimportant, the existence of an MT does not restrain your activities in any way. Just ignore it. >Now let me ask a more practical question which, hopefully, will help me >focus my understanding of your document. Before the Model Theory [1], we >had RDF graphs and detailed specifications of how to form and communicate >them [2 - 6]. Can you provide an actual example of a RDF graph that adheres >to those specifications, yet is invalid according to the Model Theory ? No, but the question isn't meaningful. Validity is a property of a(n inference) process that transforms graphs into other graphs, not a property of a graph. Pat Hayes -- --------------------------------------------------------------------- IHMC (850)434 8903 home 40 South Alcaniz St. (850)202 4416 office Pensacola, FL 32501 (850)202 4440 fax phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Received on Thursday, 4 October 2001 16:32:54 UTC