Return-Path: Message-Id: From: eric (Eric S. Raymond) Subject: Re: names as predicates To: cowan (John Cowan) Date: Thu, 13 Jun 91 12:20:43 EDT Cc: lojban-list X-Mailer: ELM [version 2.2 PL13] Status: RO X-From-Space-Date: Thu Jun 13 12:40:43 1991 X-From-Space-Address: eric > > > (My skepticism about the dichotomy of names and predicates > > is related to my distrust of equality as a primitive notion > > in predicate logic. I suspect that the abstract notion of > > equality misleads us concerning the nature of perception; > > in this view, equality is properly applied--if at all--only > > to abstractions and not to physical objects.) > > I don't understand this. (BTW, by "equality" I assume you mean "identity"; > what is expressed by the Lojban word "du" or the Old Loglan equivalent "bi".) > Why shouldn't identity be applied to physical objects? It is simply that > relationship which holds only between a thing and itself: the smallest > reflexive relation, in Kripke's definition. The previous poster was correct (damn, I wish more people on this list new some General Semantics). Identity cannot even obtain between two observations of an object made at different times, because every `object' changes constantly (in GS terms, "There are no things, only processes."). You may if you wish assert identity between two observations of an object at the *same* time, but even this is shaky; two different observations will yield two different sets of data, and the a-prioristic assumption of "identity" sweeps the issue of how you justify connecting one set to the other under the rug. A sane language should only allow "identity" to be asserted between abstractions in a formal system (what philosophers call "formal identity"). The strongest assertion a sane language should allow you to express about other things is "x is indistinguishable from y for purpose z" Natural languages, among other things, promote the assumptions that (a) there is a valid physical identity relation with properties like formal identity, and (b) that this is what we express with natural-language 'is' or equivalents. I won't even bother deconstructing the second; anybody on this list should be able to see the problems with it, and (except for the existence of `du') lojban does a fair job of addressing them. I've indicated above why (a) is false. It is dangerous for a number of reasons. The most important is that it fools people into thinking you can apply straight 2-valued Aristotelian logic with "hard" categories to the physical universe. For a formal demonstration that this is false, study quantum mechanics. For a less formal but more convincing one, consider carefully what it means to say that something is "not red" or "cold" (the problem is easier to see in negated statements). In fact, you cannot apply logic to "reality" at all. You can only apply it to abstractions captured from reality. Thankfully, it happens that the capturing process often preserves enough of what's important that your logic manipulations can be used to predict the future of "reality" -- but there is *no a-priori guarantee* that this need be the case. There is no identity anywhere; there's just "close enough for predictive work". I'd forgotten that lojban supported a "material identity" assertion. This is a serious flaw. Is there any chance we can fix it, say by adding to the place structure of "du"? -- >>eric>>