Subject: SUO: Abstractions and Embodiments (was Lifecycle Integration Schema From: "John F. Sowa" Date: Sun, 28 Sep 2003 12:11:22 -0400 To: Jon Awbrey CC: SUO , Matthew West , cg@cs.uah.edu, Nicola Guarino Jon, Matthew, and Chris, Yes, that is a good succinct summary: JA: There is a key insight in John Sowa's way of setting out the Abstract/Physical distinction that I would recommend for your consideration. If I remember right, he notices that each of the sorts of things on the Abstract side of the ledger serves to "characterize" the parallel sort of things on the Physical side of his account balance sheet. This is very important, as I see it, since it reinforces the struts that fix the categories together horizontally, calling long overdue attention to the functional and the relational architectures of the overall system, in other words, the functions, mappings, or transformations, plus the relations and dynamic structures that exist internal to the more static anatomy of the system, giving it life. It is also related to the type/token distinction. The abstract entities are the types, of which the physical embodiments are the tokens. Matthew used set terminology, but the set notion, by itself, is much too primitive because it fails to capture the complex structure that relates the members of the set: MW: Well I could try it the other way round. All the things that are abstract are set-like things, defined (in the end) by their extension, i.e. sets and relations. Possible_individuals are not sets of any sort and it would make no sense to talk about their members (in a set like sense). Yes, the type is abstract, and the tokens are physical. But you can define a set by grouping anything without any concern for any relationships among the entities. There is much more to be said, and just using set terminology does nothing to illuminate the essential characteristics. A very important application is related to everything we do with computer systems: all information is abstract, and computer systems are designed to change the physical embodiments while preserving the abstract correspondences. For example, a sequence of clicks on a keyboard is one physical embodiment, which gets transformed into a so-called "character string", which is an abstraction that is represented in numerous ways inside a computer system: 1. A string of "bits", which are abstractions that are embodied in magnetic spots on a disk, in currents flowing in transistors, or in marks that resemble "0" and "1" on paper. 2. One pattern of bits called ASCII or another pattern called Unicode (both of which are abstractions, which can be embodied in any or all the ways mentioned in point #1). 3. Many different kinds of marks, some of which are embodied in pencil, ink, chalk, or scratches drawn on some flat surface by humans. 4. Other marks on paper printed by little hammers hitting a ribbon that is pressed on the paper, by ink droplets sprayed on paper, or by light beams scanning across a charged cylinder that attracts dust that is baked on paper. 5. Still other marks by light spots caused by electrons hitting a screen, by reflecting "liquid crystals", or by little diodes that emit their own light. 6. And those marks themselves are not identical, but are organized in "fonts", which contain different patterns for the same abstract character, which are related only by some human conventions. 7. And by the way, I haven't even begun to talk about the "meaning" of that character string, which would open up more cans of worms than we could ever use on all our fishing expeditions. What is truly amazing is the number of different kinds of transformations that routinely take place under the covers of our computer systems. Even more amazing is that they somehow preserve the identity of the patterns of information from typewriter clicks to magnetic storage to computer displays to printed paper. These transformations are similar to the transformations that happen to Beethoven's 9th Symphony, which he imagined in his head because he had already gone deaf, which he transcribed to symbols on a manuscript, which publishers reproduced in different but related symbols on a score, which muscians repeatedly "perform" by causing their instruments to vibrate, which create sound patterns in the air, which are "recorded" as marks on wax, vinyl, magnetic tape, laser disks, or whatever. All those physical embodiments preserve the "same" abstract pattern, which we call "Beethoven's 9th". All of this is a bundle of semiotic processes that Peirce characterized over a century ago. And it is impossible to give an adequate ontology for any of it without using his categories (or some equivalent, but more often watered down, reinvention of them). And Chris, these are further examples of why the notion of "identity" is not primitive, but a very complex context-dependent feature. I agree that identity is important for ontology, but it is not a primitive. Just try to explain how to "identify" the letter "A" or Beethoven's 9th without getting into all the issues discussed above. As I said before, identity is much too complex and ill defined to be used as a foundation for ontology. Identity involves all of the issues of semiotics, and it must be treated as something to be explained, not as something that can be used to explain anything else. John Sowa