----- Original Message -----
> Subject: Re: OOOPS! (Was Re: help with learning class
> From: Christopher Green <[EMAIL PROTECTED]>
> Date: Wed, 26 Oct 2005 13:44:50 -0400
> X-Message-Number: 4
>
> Mike Palij wrote:
> > On Wed, 26 Oct 2005 07:51:53 -0700, Christopher Green wrote:
> >
> >> The point is simply that what we, in the wake of behaviorism,
> >> see as the "normal" meaning for "learning" is a historical
> >> construction...
> >
> > I have a problem with calling it a "historical construction" in
> > contrast to a theoretical or paradigmatic development.
>
> As you wish, but I see no reason to privilege Kuhn's view of science in
> this way, especially with respect to psychology, whch isn't among the
> "sciences" he was writing about, though psychologists have, by and
> large, missed this point in their eagerness to adopt Kuhn (See my
> American Psychologist letter about this issue:
> http://www.yorku.ca/christo/papers/Kuhn.Driver-Linn.comment.htm
I think you miss my point. My reading of Danziger (though limited
to Chap 6 so far) is that a "historical construction" is akin to an event
that follows/flows from other "historical events" and not from the processes
involved in theory revision or the conduct of research or maintenance of
beliefs and attitudes toward phenomena (i.e., is is a "historical artifact"
in contrast to an an aspect of the social conduct of science).. Perhaps
my use of the term "paradigm" may be inconsistent with Kuhn's original
conception but I think that most people will understand that it as referring
to an organized worldview which attempts to guide scientific conduct as
well as defining the appropriate objects of such studies (and the boundary
conditions of such studies).
By the way, I read Driver-Linn's paper, your comments and those of other
tp her paper, and her response to the comments. I've just read them once
and I'll have to read them again to think more deeply about them but, it
seems
to me that Driver-Linn makes a number of good points (particularly on the
issue of psychology being muliparadigmatic) at least with respect to some
aspects of how psychologists have used Kuhn. I think that your comments
and references to materials outside of psychology (particularly the points
made by Friedman and Fuller) were helpful but didn't really alter one's
interpretation of the main points of Driver-Linn's article.
I did think, however, the Driver-Linn's responses were somewhat "thin"
in nature and, if it makes any difference, I did not perceive Driver-Linn's
article as being "too comforting". :-)
> > If we consider behaviorism to be a paradigm,
>
> We shouldn't. It has few of the standard features of a paradigm. For one
> thing, it never dominated the field in the way that, say, the periodical
> table of elements did.
I think that this is the point where I'm supposed to ask you to provide
a definition for the term "paradigm" but I have no reason to believe that
you will be anymore forthcoming to this request than my earlier one for
your definition of "learning". However, let me make a few points about
behaviorism. Behaviorism isn't a philosophy per se, it isn't a theory in
the
ordinary scientific sense of the word though it does guides research
practice,
identifies which "scientific problems" are appropriate or inappropriate for
scientific study, and has encouraged the development of many new research
and clinical techniques which have enriched our understanding of phenomena
though from a somewhat skewed perspective. I think that it is fair to
say the following:
(1) Self-identified behaviorists had certain metaphysical and theoretical
assmuptions about the nature of human behavior, cognition, and emotion
that has defined a framework of beliefs for viewing the world. Sometimes
these assumptions would be implicit in their research (i.e., those
unfamiliar
with behaviorism would not readily "see" why things "had" to be done in
a certain way) while at other times they would be explicit (perhaps
Skinner's
writing for the "popular" audience is the clearest example of this).
(2) The behaviorist worldview established "inclusion" and "exclusion"
criteria for (a) how to describe a phenomenon, (b) how to refer to
processes
that involved within a phenomenon, and (c) how to explain a phenomenon.
In my view, this defined a form of "normal science" which was meaningful
to people who subscribed the behaviorist perspective and provided a
"prescription" for how to conduct "good" science. Of course, other
theoretical viewpoints (paradigms?) such as the gestalt psychologists and
later the information processing theorists would promote their own
"worldview" (i.e., methaphysical assumptions + theoretical orientation +
social/political/cultural considerations + ... ).
The preceding may be taken as part of what I consider a paradigm to
be though it may not be in direct correspondence with Kuhn's or other's
definition of the term. However, if this poses a problem to some, I
would follow J.B. Conant's advice to Kuhn (as provided by Driver-Linn,
p274) about the overuse of the term paradigm as suggest "theory" in
its place. The problem with the term "theory" is that it has a much more
limited degree of "coverage". For example, consider the following quote
that I swiped from another mailing list which circulated an article on the
Intelligent Design (ID) trial in Pennsylvania:
***************
Rothschild told the court that the US National Academy of Sciences
supplies a definition for what constitutes a scientific theory: "Theory:
In science, a well-substantiated explanation of some aspect of the natural
world that can incorporate facts, laws, inferences, and tested hypotheses."
****************
Notice that factors such as metaphysical assumptions, etc., play no
direct role in defining what a theory even though such factors may affect
how a theory is constructed, tested, and promoted.
[Sidenote: In the article that the above swiped quote comes from,
the main story is that one of the ID proponents was arguing for a
"looser" definition of theory such that ID could be accomodated.
Another swiped quote directly relevant to this point:
*****************
Rothschild suggested that Behe's definition was so loose that astrology
would come under this definition as well. He also pointed out that Behe's
definition of theory was almost identical to the NAS's definition of a
hypothesis. Behe agreed with both assertions.
*****************
And so it goes.]
> >> In short, the meaning of the term "learning" had to change farily
> >> profoundly before behaviorism was possible. Danziger shows
> >> the course of that change, now forgotten but all by historians of
> >> the field.
> >
> > He provides one "story" of what happened. I reserve the right
> > to withhold judgment on its utility.
>
> How about its truth?
Sorry, I'm a scientist. I don't deal with the "truth" but with evidence and
judgments of the degree to which such evidence is consistent with one
of more groups of statements (i.e., hypotheses, theories). I think,
however,
one could ask about the "accuracy" of Danziger's account.
> How about it providing one with context (and
> therefore understanding) that one didn't have before.
Humor me while I provide a somewhat "eccentric" perspective on what
I see Danziger having done and, perhaps, on historical analysis in general.
With historical events we a likely to be dealing with events that occur
only once and have unique characteristics associated them (this is not to
say that there are no cyclic patterns, e.g., in the pre-1995 internet world
September was frequently thought of as "cruelest month" because that
was the time of year when college freshpersons would discover the "wide
wonderful of Internet/Usenet/IRC/etc" and promptly fall into the role of
"newbie" and disrupting the normal operations of such venues -- today we
are beyond that but the current situation has been described as "everyday
it's September somewhere in the [internet] world"). I assume that the
events that Danziger describes are "unique" and unlikely to re-occur. I
and no one else can independently replicate the events Danziger talks about
and examine them at our leisure -- consequently we have to ask "how complete
an account" does Danziger provide?
Imagine for argument's sake that there are perhaps 1000 relevant facts
pertinent to the topics that Danziger covers in his Chapt 6 on "behavior"
and
"learning". What guarantee do we have that all 1000 facts have been
reviewed
and considered by Danziger? Probably none. It may in fact be impossible
to examine all 1000 facts (some of them may be hidden away in written
materials
stored in secluded, dusty areas of libraries, attics, basements, storage
facilities, etc.).
If Danziger is presenting his argument on the basis of the subset of the
facts, then
there has to be some sort of inaccuracy in his presentation (a form of
"sampling
error", if you will). Perhaps Danziger used 100 facts or 950 facts -- the
number
is actually not important because the real issue is whether Danziger has
provided
a "complete" (however one wants to define that term) and accurate account
and
a "mere" reading of his writing is not going to reveal that.
Danziger provides one account and interpretation for what happened among
researchers in the first decades of the 20th century but I have no reason to
believe
that it is the "definitive" or "final" word on what happened back then.
As scientists, we recognize the tentativeness and fallibility of
knowledge --
I don't see why Danziger's account should be excluded from these
considerations.
> > I think it would have been more helpful if you had directly addressed
> > the questions I provided below. For example, it seems to me that
> > Danziger is dismissing a "trans-species" view of learning processes
> > when he reviews Thorndike's work -- the impression I got was that
> > Danziger Thorndike was "overreaching" in attempting to develop such
> > a broad theory of "learning". I am I wrong in this impression?
>
> Knowing Kurt, he probably does think that it was "overreaching." But, so
> what? You can make your own decision on whether it was a good bet or a
> bad one.
I am somewhat puzzle by your statements on this point, in part because
you haven't asnswered some questions I asked that are directly relevant.
For example, do you believe that there are "trans-species learning
processes"
such as classical and operant conditioning?
If not, then Thorndike was "overreaching". Perhaps you view human
cognitive processes somewhat like Chomsky views "language" as being
unique to human beings (I believe that Chomsky has recently described
language as "gift from God" but I don't have a source for that). From
this perspective it is easy to understand why one would want to keep
theories of "human learning" separate from theories of "animal learning"
(though if one is using a Chomskyan rationalist perspective, "learning"
is probably the wrong term to use -- "maturation" might be more
appropriate).
If yes, then Thorndike can be seen as furthering a program of research
and theory that derives from an evolutionary perspective. If evolution
and natural selection apply to all species and if adaptation to an
environment depends upon both (a) one genetic inheritance and
(b) one's ability to alter one's behavior in dynamic environments
(learning?), then Thorndike was not overreaching because the
fundamental question becomes "do animals and humans adapt to
changes in their environments in the same ways?"
Your aversion to answering relatively simple questions, such as
what is your definition of "learning", makes it very difficult to
understand what you are saying at times because it is not at all
clear what assumptions you are using or which frameworks you
are invoking.
> Second, wasn't it over-reaching?
I think you might want to ask evolutionary psychologists about this but
I think the short answer may well be "no".
> Didn't behaviorism ultimately founder
> (in the sense of ultaimtely failing to caputre all of psychology)
> because it couldn't cash in its promisory notes on things like language
> and complex decision making?
*sigh* I'm really not trying to engage you in a Wittgensteinian language
game by continually asking you "what do you mean by that" but the statement
above really does require some explicit definitions of re-formulation:
(1) Behaviorism (whether a paradigm or not, a community of like-minded
researchers, etc.) made a heavy reliance on particular forms of associative
learning (i.e., classical and operant conditioning). Contemporary neural
network models of cognition can incorporate conditioning as special cases
of more general forms of associative learning. This has led to the odd
realization that conditioning is no longer simply a "behavioral phenomena"
but a cognitive one. A relatively simple introduction to this view is
provided
by Alan Baddeley in his text "Human Memory" (1998 edition: Chap14.
Where next? Connectionism Rides Again -- Thorndyke originally coined
the term connectionism and it has been "rehabilitated" by contemporary
psychologists and cognitive scientists to refer to neural network models).
If one equates "associative learning" with "behaviorism", then, no, it
hasn't
foundered.
On a social note, I'm sure that there are still a number of psychologists
who self-identify as being a behaviorist, perhaps in the Skinnerian mode.
I have grad school buddy in California who does a lot behavior mod work
with autistic children (including language/communication skill development
and he makes a lot more money than me!). So, if by founder you mean
no one believes in it anymore, then no.
(2) Language: it would be immensely cruel of me to ask you to define
what you mean by language but I have the feeling that you might be
referring to the "syntactocentric" type of theoretical formulation that
Chomsky had back in the 1950s when he trashed Skinner's "Verbal
Behavior". I'm sure you'll correct me if I am wrong. Today, both
linguists and psychologists have gone far beyond the "traditional"
"rule and symbol" cognitive architectures that were implicit in Chomsky's
theory of syntax (and most of early information processing psychology).
In addition to connectionist modeling of language phenomena (where
associative learning reigns) new areas of research have developed on
the social use of language, how social context defines the meaning of
utterances, the nature of conversational interactions, etc. It should be
clear that associative learning/conditioning factors are going to be
relevant in a variety of these situations (the question becomes whether
one allows the behaviorists to claim all of associative learning or just
conditioning -- that is, if, by virtue of the ascenion of connestionism,
we're not all behaviorists now ;-). And if one needs a "traditional"
example of behavioristic treatment of language, behavior modification
in sign language training and other language skills training with autistic
children probably fills the bill.
(3) Complex-decision making: I don't know which research or theories
you're referring to but my first reaction is that there have been a lot of
"behaviorially informed" theories and models in this area. One of my
favorite papers in decision-making is Robyn Dawes American Psychologist
paper entitle "The Robust Beauty of Improper Linear Models" in which
he makes the simple but fundamental point that people may be good at
identifying "relevant" or "important" variables in the environment but
they're
bad at (literally) "adding up" this info (i.e., integrating it). Prior to
this
point Dawes and others have shown that "proper linear models" like
multiple regression equations ("proper" because the weights/coefficients
were "optimally" selected) were better at making predictions than human
judges using interviews (i.e.., predicting who will do well/fail at the end
of their first year of grad school). In this paper Dawes points out that
(a) allowing person to select which variables are relevant to a decision
(e.g., predicting grad school GPA at the end of their first year) and
(b) assign an "importance" rating or weight (an "improper" or nonoptimal
weight/coefficient), still outperforms predictions based interviews.
However, there has been and continues to be resistance to such "mechanical"
decision-making devices and some people in the decision-making area
thought the development of computer-based "expert systems", that is,
computer simulation of expert decision making (e.g., MYCIN was a program
that was developed to identify the "best" treatment for a medical
infection).
"Knowledge mining" of the rules that experts used served as the basis for
the development of these programs but though they could perform at
relatively
high levels, they disagreed with with human judgments (the final arbiter) to
a disturbing degree. Why? If one extracts all or most of the rules that an
expert uses in making a decision, why doesn't the program agree more closely
with the human expert? One answer to this question is that there may be
rules
that human experts develop with experience but for some reason either can't
articulate them or are not aware of them. If this is so, does it means that
expert
system performance has hit a ceiling and do no better?
Well, expert systems are usually implemented in programs called "production
systems", a classic "rule and symbol" architecture where the rule and
symbols
have to explicitly identified and programmed in. How can one extend the
capabilities
of such a program? By creating a hybrid program that has (a) a database of
expert knowledge and (b) a neural network/connectionist learning program
that
extracts implicit rules from dealing with numerous cases in the area.
You can see where I'm going, don't you?
Neural Network/Connectionist Modeling -> Associationist Learning ->
Behaviorism
And I'll leave the the topic of "behavioral economics" to one of the
hardcore
behaviorist types. :-)
So, what one means and doesn't mean by such terms as "behaviorism", etc.,
will have a significant impact on an argument, particularly whether one
thinks a
line of development is relevant or not.
-Mike Palij
New York University
[EMAIL PROTECTED]
> Regards,
> --
> Christopher D. Green
> Department of Psychology
> York University
> Toronto, ON M3J 1P3
>
> [EMAIL PROTECTED]
> http://www.yorku.ca/christo
---
You are currently subscribed to tips as: [email protected]
To unsubscribe send a blank email to [EMAIL PROTECTED]
