Subject: Re: SUO: Re: One Stone Fits All From: "John F. Sowa" Date: Fri, 18 Jul 2003 13:09:38 -0400 To: Stefano Borgo CC: standard-upper-ontology@ieee.org, nicola.guarino@ladseb.pd.cnr.it Dear Stefano, I agree that my formula for similarity was oversimplified, and corrections along the lines you suggest are important. My only excuse is that I was writing a short reply late at night for which the primary point was about identity. The main thrust of the argument (which can be made with any definition of similarity you choose) is that similarity is much closer to perception than identity, which can only be inferred, never perceived. You may criticize that point by saying that you are doing "pure" ontology, not to be confused with "epistemology". But I would like to emphasize that ontology has its roots in Aristotle's Metaphysics, which begins "Man by nature desires to know." Aristotle wisely presented both aspects in the same book, and modern attempts to separate them have been singularly unsuccessful. Instead of making progress, they have created more philosophical "puzzles" which are inherently unsolvable, as Wittgenstein demonstrated again and again. Several years ago, I was visiting Nicola at his office in Pisa, when he brought out one of his books with a philosophical discussion of problems with identity. He became angry when I dismissed the book as "garbage." I was a bit hasty because the book was a thoughtful, well-written, well-researched analysis of the various problems of identity. So the book did have some merit. But that merit was destroyed by its failure to admit that identity is less fundamental than similarity. You might add a qualifier "epistemologically", but I would insist that the foundations of ontology must be grounded in primitives whose links to both perception and action can be specified by testable criteria. SB> However, you are aware that this "definition" does not work unless you > state very carefully which predicates you allow in the system (for instance, you don't want "being self-identical" to be a predicate, right?). Othw, everything turns out to be similar to everything else and the whole argument leads nowhere. Formal logic is of no help here. Right. There is a great deal of qualification that must be added, and again, I apologize for writing hastily late at night. The most important correction is to limit the predicates to ones that can be defined by perceptible qualities or positive conjunctions of perceptible qualities. When you do that, you have much simpler and more testable criteria for similarity than for identity. JS>> What I recommend is that you delete every occurrence of those words >> [universals and particulars]. Instead, you should replace that >> discussion with talk about how the symbols of logic relate to >> the world. > SB> You keep saying this but you carefully avoid mentioning how one can do it. > ... > Please, give me an example that works for you. How can one do (to your satisfaction) what you are asking for? Just to make sure this is actually feasible... That is essentially what I've been doing for years. In both my Conceptual Structures book (CS 1984) and my Knowledge Representation book (KR 2000), I addressed a wide range of philosophical issues without using the terms "universals" or "particulars". Instead of the word "particular", I refer to the value of a variable in predicate calculus or to the referent of a concept in CGs. Instead of the word "universals", I refer to the predicates in predicate calculus or to the concept and relation types in CGs. I don't claim that the value of a variable is a synonym for "particular" or that a predicate is a synonym for "universal". But what I claim is that any discussion that uses the U and P words can be restated in a way that relates directly to the logical notation. That restatement has several advantages: 1. It avoids an additional level of abstraction between the natural language descriptions and their formalizations in logic. 2. It avoids the distracting and confusing issues about how universals are related to predicates. Perhaps some philsophers might like to see such a discussion -- that's OK if you want to publish a paper in a philosophical journal, but it does not belong in the DOLCE documentation for users and developers of the ontology. 3. It avoids dubious and confusing discussions about "abstract particulars". Mathematics has been formalized since the time of Euclid without using such terms, and there is no need for DOLCE to introduce them. JS>> To summarize, I'd be happy to recommend the inclusion of DOLCE in >> the collection of modules for an SUO registry. I'm sure that there >> may be many useful applications for it. But as I said about Cyc, >> I have some serious concerns about what directions the ontology >> work should take. > > SB> I appreciate that and, to some degree, I share your concerns about work > in ontology. I'm happy to end on a note of agreement. I've known Nicola for many years, and we've had many useful and stimulating discussions. Even though we disagree about important issues, there are also a lot of points of similarity (not identity) on which we can collaborate. So I hope that we can continue. John