sandyz@ntpdvp1.UUCP (Sandra Zinn) (06/01/90)
> >> (Aaron Sloman) writes: > >>> > >>>A lot of our education is concerned with developing such reflexes > >>>(e.g. reading, arithmetic, language comprehension and production, > >>>learning patterns that enable decisions to be taken without > >>>analysis, etc). . . . > >> > >(Ken Presting) writes: > >>Much of this learning is dependent on pre-existing language > >>understanding. > > > (Stephen Smoliar) writes: > >This may be less important than you might suppose at first blush. I may be > >misinterpreting Aaron, but I think the sort of thing he has in mind are > >situations which may involve rote memorization with any necessary underlying > >UNDERSTANDING of what is being memorized. ... > > (Ken Presting) writes: > > My concern is whether Aaron has outlined a system of abilities to learn > which can explain language learning. There is no question that rote > repetition can facilitate ritual behavior, and that the ability to > associate complex responses to simple stimuli is adaptive. The question > is (a) whether intelligence can be said to exist in the absence of these > capacities, and (b) how much of human behavior can be reproduced by > iterative application of this type of learning. I'll jump in with my interpretation of Aaron's description, which is I think extremely suggestive and important. I think Stephen is right that language *dependency* is irrelevant, but missing the boat by talking about rote memori- zation. Ken gets *very warm* when he says the ability to associate complex responses to simple stimuli is adaptive. Why does Aaron call this phenomenon a reflex? Because it is so similar in function to more physiological reflexes. Why did they develop? Because function calls to generic routines are more economical than writing the durned things from scratch every time -- but that's looking at it wrong- end-to, and we might miss some subtleties. The evolutionary development of biological systems proceeds, according to theory, from simple organisms to complex ones. The complex ones, however, are not huge lateral masses of details, but hierarchically structured, with layers of complexity built on successful previous layers of complexity. We tend to think of reflexes as being simple, or at least rudimentary. That's only when we look at them in terms of the "more complex" system which invokes them. But the reflexes themselves are elaborate, and the "packaged responses" which Aaron speaks of are similarly complex skills. It's this "packaged response" characteristic which is so important here. Imagine walking down the aisle in the grocery store, with your arms full of items, and you remember your wife said to get tomato sauce, but you couldn't even balance it on your current load, much less pick it up. So you go on and check out, and explain to her what happened. But the next five times you go to the grocery store, the same thing happens. She's getting awful tired of your redundant explanations, and you're under a lot of stress to get that tomato sauce. You suddenly realize you can use the little two-handled baskets stacked up front. You make your rounds, but now, when you come to the tomato sauce, you've got a hand free to pick it up, 'cause you've stuck everything else in the basket. You still have the same number of items on one level (in the basket), but at the "top" level, you have only two items: a basket-full and a can of tomato sauce. In communications theory, if your channel is full, you recode your data, compressing it, and thereby gain capacity to say something additional (maybe something new). I'd suggest that reflexes (whether physiological or cognitive) become reflexes only when a code is developed which reflects them more simply "at a higher level" (it becomes a higher level only by the development of this ability to represent, *and evoke by representation*, the earlier response). Logically, a complete description has been translated into a single symbol in a new descriptive system. All the previous complexity still exists in the first system, and its operative capacity is *not necessarily* diminished. Since most of us don't put exactly the same things in our cart at the grocery store each week, plasticity at the lower level has clear value. The interesting part is the relationship between the two. If the "reflex" and its representation remain static, then you have a very mechanical system. However, if the "reflex" remains dynamic, then your invoking system has a problem. Simply, which should it invoke? The old pattern of behavior, or the new derivative? It could use trial and error, but that's costly in adaptation terms. Its best bet is to have some knowledge of the *functional content* of the reflex -- but this can be specified only WRT the full descriptive system for that reflex! So if the reflex system remains dynamic, the invoking system MUST have *both local and global* correlations in its encoding of the reflex in its own descriptive system. My guess is that this powerful encoding capacity exists (within single descriptive levels) only in the nervous system. Most behaviorist models keep the reflexes as static constellations. However, you don't get enough integrative power that way -- merely boxes stacked with memorized relationships, but without strong logical relationships. Reading a book under strong light and reading a book under weak light could then only be discrete phenomena. But a dynamic learning model preserves functional identity across varying conditions. > When verbal instructions are used to identify the behavior to be learned, > questions arise as to the "reflex" quality of the response. What questions, exactly? > Understanding of the *behavior* is not at issue, rather, understanding of the > *instruction* is the problem. > In the knee-jerk case, the subject's > preparation has nothing to do with the evoked response. Not so in the > multiplication table case. According to my treatment above, the *preparation* is the same in both cases. Understanding the instruction is not really germane, either. You're trying to smuggle in intentionality by creating arbitrary divisions, again! It's not apparent where it logically comes in from looking only at the "still frames" -- again, process is the thing. @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ Sandra Zinn | "The squirming facts (yep these are my ideas | exceed the squamous mind" they only own my kybd) | -- Wallace Stevens
smoliar@vaxa.isi.edu (Stephen Smoliar) (06/12/90)
Sandra Zinn is back "on the air" with a fascinating piece on reflexes (and a pleasant relief from all those bits expended over John Searle). However, I cannot resist the urge to question one of her foundations. In article <583@ntpdvp1.UUCP> sandyz@ntpdvp1.UUCP (Sandra Zinn) writes: > >The evolutionary development of biological systems proceeds, according to >theory, from simple organisms to complex ones. The complex ones, however, >are not huge lateral masses of details, but hierarchically structured, with >layers of complexity built on successful previous layers of complexity. > My current .signature is taken from Lewontin's recent review of Stephen Jay Gould's new book, WONDERFUL LIFE: THE BURGESS SHALE AND THE NATURE OF HISTORY. As I understand Gould's story, the fossil record of the Burgess shale essentially contradicts the traditional view of evolutionary development which Sandra has articulated. Nature is apparently much richer in the diversity it provides and perhaps more arbitrary in which forms actually survive than we might wish to think. ========================================================================= USPS: Stephen Smoliar USC Information Sciences Institute 4676 Admiralty Way Suite 1001 Marina del Rey, California 90292-6695 Internet: smoliar@vaxa.isi.edu "So, philosophers of science have been fascinated with the fact that elephants and mice would fall at the same rate if dropped from the Tower of Pisa, but not much interested in how elephants and mice got to be such different sizes in the first place." R. C. Lewontin