Re: principles of collective organization
Rob Freeman wrote:
>On Tuesday 10 January 2006 18:12, Frederick B. Kintanar wrote:
>
>
>>>Within a consensus on subjective knowledge representation would it be fair
>>>to say we have at least three suggested solutions:
>>>
>>>
>>If you are trying to present these three solutions as alternatives
>>or a spectrum to choose from, please elaborate because they
>>seem mutually compatible to me. In fact, they may identify three
>>somewhat orthogonal dimensions that should all be taken into
>>account.
>>
>>
>
>I think all these approaches are appropriate as far as they go. Most
>importantly they are all based on the same insight that... we must have
>different categories for different purposes.
>
>This consensus I'm talking about is by no means well established. I would like
>us first to understand that what we have is a consensus. If a group of people
>can agree on some new principles, even a new hypothesis, that is an important
>advance.
>
>That said I don't think the different approaches are completely
>interchangeable.
>
>In particular while it is important to acknowledge categories are subjective,
>I think it is more important to understand why.
>
>
>
>>>3) I'm suggesting that we think about knowledge representation directly in
>>>terms of "principles of collective organization". In particular I'm
>>>suggesting we should think about it in terms of "principles of collective
>>>organization" which also underly language; that they are the same, and
>>>that we can model knowledge this way through language, that this indeed
>>>is what language does.
>>>
>>>
>>I am sympathetic to this suggestion, and I agree that language is
>>quite important and likely to dominate practical initiatives. However,
>>I am not sure that it is the most basic level of organization that we
>>need to worry about.
>>
>>
>
>Sorry, Fred, I'm not clear what you mean. You are not sure it is the most
>basic level, or you are not sure the basic level is the one we need to worry
>about?
>
>
I mean "I don't think the level of 'the principles of collective
organization that
underly language' is basic enough." I propose going one "level" lower to
the the social-conceptual schemes mentioned below, and then characterizing
language based on that.
>>In particular, I think that humans have mostly
>>innate (but realized through epigenetic ontogenic development,
>>and some prelingual experience) social-conceptual schemes which
>>underly linguistic semantics.
>>
>>
>I guess so. I mean lots of behavior is innate, even in humans. That will
>parameterize what we do, and what we do will parameterize language.
>
>Whether instincts are in themselves collections of individual behaviors or
>can be modelled as primitives is another question.
>
>Probably we are innately tuned to some aspects of experience, like the
>perception of time, and this will have an impact on the nature of our
>experiences.
>
>But I'm not sure language is directly parameterized on this level, or if what
>is parameterized are only the experiences language is exposed to.
>
>Anyway I think the collective aspect of language is the one we have missed,
>and which has prevented us adequately modeling linguistic and cognitive
>behavior.
>
>
When you say collective aspect of language, do you mean "all parts of
the brain"
or do you mean "social" or what?
>I'm sure there is a big win to be had here just by turning our models around
>and allowing grammar and concepts to emerge ad-hoc from examples, rather than
>trying to parameterize examples using posited primitives for grammar and
>concepts.
>
>
I agree that the more empirical direction is a good idea. If by concepts
you refer to not just verbal concepts but also some concepts (and related
principles) that may be more basic, then I think we may be in close
agreement. Some of the underlying principles I think need to be
considered are subjective qualia, intentionality, mental causation,
and function of biological systems and artifacts. None of these
seem to be specifically underlying language, although they may do that
as well.
>>I often reflect about the nature of the mind of pre-linguistic deaf
>>children. These are kids, typically in rural areas, who are born to
>>hearing parents who grow up with no contact with deaf adults who
>>know a developed natural sign language. These kids may develop
>>a small set of "home signs" depending on the nature of their
>>relationship with their mother and immediate relatives, but will
>>demonstrate practically no syntax or structured communication.
>>And yet they can function in a social world, with sophisticated
>>concepts and mental states that go beyond other mammal
>>individuals. They can understand how light switches work, and
>>respond to requests to turn a light off that you make with your
>>eyes or by pointing. They can understand how money works at
>>the neighborhood store. They categorize the world and
>>remember how it is organized, they can make sense of the
>>behavior of others and reason about causes. And when they
>>do meet adults who know sign language, they acquire vocabulary
>>and linguistic competence rapidly as a layer on top of existing
>>concepts.
>>
>>
>
>I thought there were a couple of famous "wolf-children" who missed a crucial
>stage and never could be taught to talk adequately.
>
>
If there really are cases of "feral children" without language or the
capability to acquire it, that would be interesting data. However,
a truly feral child would lack the socialization to human behavior
that a language-isolated deaf child would have, so might not
contradict the framework I advocate.
>In general though, I wouldn't imagine isolation from language itself would
>mean you would not have cognitive development. I'm not making an identity
>between language and cognition, only speculating they have the same form. I
>think language has the same form as thought, and is useful for that reason
>(form), but it is not the same thing. Even without language, concepts could
>and must emerge based on collections of experiences.
>
>
I agree about form. A specific theme I am interested in is applying Typed
Feature Structures (from grammar frameworks like HPSG) to visual object
recognition.
>Lots of behaviors, social and even not social, are like language to a
>greater or lesser extent, anyway. Dance, art, perhaps... Even without
>language our minds would do the same kind of thing.
>
>I just think that language, if it is available, provides a kind of mirror and
>inverted index for those collections of experiences.
>
>
>
>>... The details
>>of anisotropic properties of crystalline solids are
>>irrelevant though for many purposes, and a common
>>sense scheme of individuation of material objects
>>and their motion is adequate to many purposes.
>>
>>
>
>The thing is, a similar "common sense scheme" does not seem to be adequate for
>language and cognition.
>
>
I agree that we need something more basic than common sense and
folk psychology, I just don't think it is necessary in general to get
down to neurophysics or a reductionist theory of everything.
>I'm sure it is a fundamental character of collective organization which has
>prevented us codifying grammar and meaning adequately up to now.
>
>
Some candidate principles with such a fundamental character that I have
been thinking about are: Typed Feature Structures (as used in grammatical
frameworks like HPSG), dynamic semantics, and Arbib's functional brain
schemas. All of them are mathematically rigorous and accommodate
the discrete world of logic and language. I think their insights can be
reformulated in terms of channel theory, using the category of Chu
spaces.
>This is simply a thing which is not being done. Even so-called distributional
>methods are universally, to my knowledge, limited by the expectation that
>there are underlying primitives (or at least primitives by consensus, c.f.
>Luc Steels at Sony, etc.)
>
I am not familiar with distributional methods, a pointer would be helpful.
>Perhaps it is a lack of popular knowledge about how
>wacky these collective phenomena can be which is responsible: uncertainty
>principles, incompleteness, and the like. Most people seem to understand
>emergence as evolution of primitives, not something which replaces
>primitives. Without personal experience they cannot assess the implications
>of collective behavior.
>
>I thought this comment in your citation of Laughlin's PNAS essay was pertinent
>(http://large.stanford.edu/rbl/articles/p01apr99.htm):
>"To solid state physicists and chemists, who are schooled in quantum mechanics
>and deal with it every day ... the existence of these principles is so
>obvious that it is a cliche not discussed in polite company. However, to
>other kinds of scientist the idea is considered dangerous and ludicrous, for
>it is fundamentally at odds with the reductionist beliefs central to much of
>physics."
>
>Anyway, to my knowledge, a true re-evaluation of cognition and language in
>terms of ad-hoc structure according to "principles of collective
>organization" is just not being done.
>
>
I'd like to suggest that Channel Theory and some other compatible principles
provide a good starting point. Working with IFF also gives a clearer reason
why such a program of work should be done within the SUO group.
>We should try it.
>
>
>
I look forward to such an effort.
Cheers,
Fred
>-Rob
>
>