Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!samsung!zaphod.mps.ohio-state.edu!mips!prls!pyramid!amdahl!kp From: kp@uts.amdahl.com (Ken Presting) Newsgroups: comp.ai Subject: Re: Emergence and Static Vs. Dynamic properties Summary: Interpretation and Normativity explain better than emergence Keywords: emergence, interpretation, normativity Message-ID: Date: 22 Mar 90 23:59:42 GMT References: <12502@venera.UUCP> Reply-To: kp@amdahl.uts.amdahl.com (Ken Presting) Organization: Amdahl Corporation, Sunnyvale CA Lines: 125 In article <12502@venera.UUCP> smoliar@vaxa.isi.edu.UUCP (Stephen Smoliar) writes: >In article kp@amdahl.uts.amdahl.com (Ken >Presting) writes: >> >>I would suggest that we defer all discussion of the relation of perception >>to description, until we have decided on the logical status of the >>concept of emergence. I believe that we will have little success in any >>attempt to understand language until we have replaced "emergent" with >>"normative". >> >>Does your concept of "emergence" essentially depend on perception? Or >>could you define it in some other terms? >> >I'm not sure that I (or anyone else) can effectively DEFINE "emergence," which If this is true, it will be difficult to use the concept of emergence in an argument, or as a design objective. Nagel's _The Structure of Science_ includes an extended discussion of emergence, as does the _Encyclopedia of Philosophy_, article "Emergent Evolutionism". The usual definition is relational, ie: 1) P is emergent wrt Q,R,S... if Px & Qx & Rx & Sx ... holds, and it is not the case that Qx & Rx & Sx & ... entails Px. This definition reduces to a straightforward notion of "logical independence" unless some additional claim is made about Q,R,S... such as: 2) x=y if and only if Qx<=>Qy and Rx<=>Ry and Sx<=>Sy ... ie, Q,R,S,... form a "complete individual description". Note that a property can be emergent under this definition only in a formal system which is *incomplete*. For example, if arithmetic is given the usual Goedel interpretation into its own sentences, and P is extensionally equivalent to truth, and Q,R,S... are recursive syntactic predicates, then (1) and (2) can both hold. Therefore truth is an ... whoops ... Truth is a *normative* property - it only looks emergent. This definition of emergence is effective for MY purposes. :-) >. . . >may mean that emergence, itself, is a normative property of dynamic systems. > Bite your tongue! er, well, then bite your fingers. I don't want to let one of my favorite concepts get tarred with a vague brush. Truth, Beauty and Goodness are normative concepts - do you want to put (yechh) emergence (ptooie) in that class? Some normative properties are in fact not usefully definable, but so are a large number of descriptive concepts, notoriously "game" and "chair". The practical vocabulary of everyday speech needs little in the way of definitions to advance its purposes, and suffers from most attempts to supply precise analyses. When normative concepts are part of everyday speech, they generally resist definition. As long as a predicate is restricted to use in communication, this is only a small problem. But when arguments or designs are stated with undefined terms, confusion or failure is likely to result. "Truth" is a significant exception to the undefinability of normative properties which have everyday utility. Tarski's theory of truth is widely applicable, although some work does continue among logicians who feel that Tarski's approach may not sove every problem related to truth. In Tarski's theory ("The Semantic Conception of Truth", in a collection by Leonard Linsky whose title escapes me), truth is defined in a metalanguage. A metalanguage is a descriptive abstraction whose terms refer to sentences and formulas of another descriptive abstraction, called the "object langauge". For example, in the metalanguage one might make assertions about the sentences in the object language, such as "It is true that 'Snow is white'". In this example, the object language includes the terms "Snow" and "white", whereas the term "true" is NOT a part of the object language. (I am omitting many fascinating details) Note that the truth of "Snow is white" depends on the *interpretation* given to the words. If "Snow" is taken to denote slush as well as new, drifted flakes, then "Snow is white" will be false. Normative properties always have this characteristic of dependence on an interpretation. Thus, normativity is a strong candidate for inclusion in a theory of perception. Discriminating and identifying and object within a scene seems to require interpreting a region of high contrast as a boundary, for example. Perception also functions as a sort of bridge between structured systems - the outside environment and the representational concepts of the mind. Concepts such as truth are descriptive in the context of a metalanguage, but are normative when applied "across" formal systems, ie, from the metalanguage to the object language. > . . . A glider in LIFE is >the same way. We can only talk about a glider as something we can observe; I don't think this is true. Stability of cellular structures, as well as stable translational motion of structures, is not difficult to formally define. The definition of emergence I gave above does not explicitly mention relational properties or dynamic properties, but that is easily remedied, and the conclusion stands. The closest I have been able to come to a useful conception of emergence is in relation to chaotic systems. In a chaotic system, it may be impossible to know the "complete description" of the initial state. But this does not show that subsequent states are emergent, it only shows that they may be surprising. To adopt emergence in this form, and claim that consciousness or intelligence is emergent is to admit that AI cannot succeed on the basis of a deliberate design. The concept of emergence in a weakend form such as this (where the appearance of emergence is attributed to limitations of the starting description) does have a genuine use, I think. Similarly, when a system evolves probabilistically, the appearance of emergence may be attributed to "imperfect knowledge" of the rules for inferring future states from past states. I wouldn't go so far as to say that emergence never occurs, ie, that unexpected things happen. I just want to avoid any claim that certain properties can be exhibited ONLY by means of emergence. >so I think it DOES make sense to say that, whatever emergence may be, it >depends on perception. Stephen, why don't we invert this and substitute, viz, "whatever perception may be, it depends on a process which is formally similar to interpretation"? Ken Presting