| View previous topic :: View next topic |
| Author |
Message |
baylor
Guest
|
 Posted: Wed Nov 05, 2003 2:11 am Post subject: Matching Law and context |
 |
|
Another matching law question, although really it>s much more generic
than that (it has to do with generalization). And it>s probably an
FAQ. What is context?
The (few) matching law experiments i read about talked about choice in
a single context. One experiment did look at a season of basketball,
and i could believe that playing a specific team constituted a change
in context, but the matching law held over the entire season
In trying to find a way to implement this in software (and test in
video games), i>ve run into the question of whether the matching law
holds for some set of options regardless of context or if the matching
law holds for a set of options within a specific context. Also, the
question in comp.ai.games is how changing environments are handled
And there>s the strong possibility that i>m misusing the matching law
and should be looking at something else
Here>s a contrived example. i have 3 weapons, a sword, a wand of fire
and a wand of ice. At the highest level, i have a rate of
reinforcement (and assumedly response) of 33% each, so the matching
law says when i get into a fight, i have a 1-in-3 chance of using any
of the options
i fight several different monsters. Many, such as dwarves, elves and
goblins, are equally affected by each weapon. But i have three
specific monsters, a fire monster, snow monster and stone monster,
that are special. The fire monster can>t be hurt by fire, the snow
monster can>t be hurt by cold and the rock monster can>t be hurt by
fire or cold
So at a general level, where my relative rate of reinforcement for
each weapon =33%, i would choose weapons basically at random. With no
other changes, that means i have a 1-in-3 chance of using my flame
wand against a fire monster, which i know will fail 100% of the time
So it makes sense to me that i don>t track rate of reinforcement at a
general level but at the monster level. My rate of reinforcement of a
fire wand against a fire monster is 0%. That would solve the problem
Which leads to the question - is that how the matching law works? Does
it rely upon context?
If so, this leads to some questions:
1. Generalization. The rate of reinforcement of a fire wand against a
fire monster, snow monster and rock monster is 0%, 100% and 50%
respectively. Assume for now that the rate of reinforcement for the
other two weapons is 50% in each case. i now fight a brand new
monster, a giant bear or radioactive lobster or whatever. Can i use
the matching law to decide the probability of using a given weapon
given that the context is new?
i did read last night that positive associations generalize much more
than extinguishment, so a 100% rate of reinforcement might apply to
anything that didn>t match the context of the extinguishing case (i
might be using really incorrect terminology here), which is the
equivalent of broad based generalization with the memorization of
exception exemplars, but i>m not sure what you>d do where the rate of
reinforcement is a probability rather than a sure thing
2. Feature extraction. What constitutes the context? Let>s say i fight
a two-headed goat monster with my wand of fire three times and each
time i get hurt and the goat monster doesn>t. My rate of reinforcement
for using the wand of fire is 0%. In each of these cases, i fought the
same goat monster (we>ll call him Bubba) in the same room during full
moons. What would i likely learn as the context? That two-headed goat
monsters are flame retardant? That fire wands don>t work against Bubba
specifically? That fire wands don>t work in Bubba>s cave? That fire
wands don>t work during a full moon? That fire wands don>t work if
it>s Bubba AND he>s in his cave AND it>s a full moon? Do i learn all
of these associations with some given level of strength or
probablility? And how would it reconcile with the matching law?
As background, in case anyone cares, i have no reason for asking, just
curious. i like making software that acts like humans. i>ve taken some
AI and psych grad classes but am not formally enrolled. My plan is to
enroll in an AI and/or Psych PhD program Fall 2005 (it was going to be
2004, but plans change). i>d like to pursue questions similar to the
above (mostly on learning, decision making, recognition and
similarity) when i start school full time and am giving myself crash
courses on all this stuff right now
Thanks
-baylor, the curious |
|
| |
|
 Back to top |
Michael Olea
Guest
|
| Posted: Wed Nov 05, 2003 4:38 am Post subject: Re: Matching Law and context |
|
|
in article 27bc3110.0311041211.38d22c84@posting.google.com, baylor at
baylor@ihatebaylor.com wrote on 11/4/03 12:11 PM:
[quote]Another matching law question, although really it>s much more generic
than that (it has to do with generalization). And it>s probably an
FAQ. What is context?
The (few) matching law experiments i read about talked about choice in
a single context. One experiment did look at a season of basketball,
and i could believe that playing a specific team constituted a change
in context, but the matching law held over the entire season
In trying to find a way to implement this in software (and test in
video games), i>ve run into the question of whether the matching law
holds for some set of options regardless of context or if the matching
law holds for a set of options within a specific context. Also, the
question in comp.ai.games is how changing environments are handled
And there>s the strong possibility that i>m misusing the matching law
and should be looking at something else
Here>s a contrived example. i have 3 weapons, a sword, a wand of fire
and a wand of ice. At the highest level, i have a rate of
reinforcement (and assumedly response) of 33% each, so the matching
law says when i get into a fight, i have a 1-in-3 chance of using any
of the options
i fight several different monsters. Many, such as dwarves, elves and
goblins, are equally affected by each weapon. But i have three
specific monsters, a fire monster, snow monster and stone monster,
that are special. The fire monster can>t be hurt by fire, the snow
monster can>t be hurt by cold and the rock monster can>t be hurt by
fire or cold
So at a general level, where my relative rate of reinforcement for
each weapon =33%, i would choose weapons basically at random. With no
other changes, that means i have a 1-in-3 chance of using my flame
wand against a fire monster, which i know will fail 100% of the time
So it makes sense to me that i don>t track rate of reinforcement at a
general level but at the monster level. My rate of reinforcement of a
fire wand against a fire monster is 0%. That would solve the problem
Which leads to the question - is that how the matching law works? Does
it rely upon context?
If so, this leads to some questions:
1. Generalization. The rate of reinforcement of a fire wand against a
fire monster, snow monster and rock monster is 0%, 100% and 50%
respectively. Assume for now that the rate of reinforcement for the
other two weapons is 50% in each case. i now fight a brand new
monster, a giant bear or radioactive lobster or whatever. Can i use
the matching law to decide the probability of using a given weapon
given that the context is new?
i did read last night that positive associations generalize much more
than extinguishment, so a 100% rate of reinforcement might apply to
anything that didn>t match the context of the extinguishing case (i
might be using really incorrect terminology here), which is the
equivalent of broad based generalization with the memorization of
exception exemplars, but i>m not sure what you>d do where the rate of
reinforcement is a probability rather than a sure thing
2. Feature extraction. What constitutes the context? Let>s say i fight
a two-headed goat monster with my wand of fire three times and each
time i get hurt and the goat monster doesn>t. My rate of reinforcement
for using the wand of fire is 0%. In each of these cases, i fought the
same goat monster (we>ll call him Bubba) in the same room during full
moons. What would i likely learn as the context? That two-headed goat
monsters are flame retardant? That fire wands don>t work against Bubba
specifically? That fire wands don>t work in Bubba>s cave? That fire
wands don>t work during a full moon? That fire wands don>t work if
it>s Bubba AND he>s in his cave AND it>s a full moon? Do i learn all
of these associations with some given level of strength or
probablility? And how would it reconcile with the matching law?
As background, in case anyone cares, i have no reason for asking, just
curious. i like making software that acts like humans. i>ve taken some
AI and psych grad classes but am not formally enrolled. My plan is to
enroll in an AI and/or Psych PhD program Fall 2005 (it was going to be
2004, but plans change). i>d like to pursue questions similar to the
above (mostly on learning, decision making, recognition and
similarity) when i start school full time and am giving myself crash
courses on all this stuff right now
Thanks
-baylor, the curious
[/quote]
You put your finger on the heart of a deep issue when you wrote:
[quote]
2. Feature extraction. What constitutes the context? Let>s say i fight
a two-headed goat monster with my wand of fire three times and each
time i get hurt and the goat monster doesn>t. My rate of reinforcement
for using the wand of fire is 0%. In each of these cases, i fought the
same goat monster (we>ll call him Bubba) in the same room during full
moons. What would i likely learn as the context? That two-headed goat
monsters are flame retardant? That fire wands don>t work against Bubba
specifically? That fire wands don>t work in Bubba>s cave? That fire
wands don>t work during a full moon? That fire wands don>t work if
it>s Bubba AND he>s in his cave AND it>s a full moon? Do i learn all
of these associations with some given level of strength or
probablility? And how would it reconcile with the matching law?
[/quote]
A short, if not very helpful, response is that people and other biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time), but are equiped
with a genetic endowment - in effect learning much of what is the relevant
set of features constituting a "context" has been done by natural selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
The broader point of view, one that has been hotly debated, is that the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain. A fine elucidation
of this point of view is Pinker>s "How the Mind Works" - a must read for
anyone contemplating a Psych PhD program. And while I>m plugging Pinker,
I might as well recommend his other superb book: "The Language Instinct:
How The Mind Creates Language".
So, as for learning to deal with Bubba, you will want to constrain the class
of hypotheses that are considered - I guess you could try to emulate
evolution (sounds like a big project), or simply build in (the hand of god)
specialized context feature detectors... |
|
| |
|
| Back to top |
Glen M. Sizemore
Guest
|
| Posted: Wed Nov 05, 2003 7:55 am Post subject: Re: Matching Law and context |
|
|
MO: A short, if not very helpful, response is that people and other
biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time),[...]
GS: It was Locke, not Skinner, that used the term "blank slate." Even Watson
argued that it was not correct. BTW, anybody who thinks that we are, in
general, "following rules" when we speak our native language is a moron.
Pinker is a moron.
MO: but are equiped
with a genetic endowment - in effect learning much of what is the relevant
set of features constituting a "context" has been done by natural selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
GS: What is your evidence for this? You are essentially talking about the
truth or falsity of the Lashley-Wade hypothesis. The Lashley-Wade hypothesis
(discriminative behavior requires exposure to discriminative contingencies)
is probably not completely true but there certainly is evidence that
discriminative behavior frequently "waits" on exposure to presence vs.
absence training. Further, I>m not completely convinced that the
"monochromatic light rearing experiments" are not confounded - apparent
distance of the light source is a function of wavelength, and discriminative
contingencies naturally arranged by living in a three dimensional world (no
matter what color it is) could produce such discriminations. But even if one
can get peaked generalization gradients without discriminative training
sometimes, so what? This is a long way from your wildly extravagant "Even
something as basic as carving up the visual scene into objects, like the
moon, or the concept of 'moon' as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA."
MO: The broader point of view, one that has been hotly debated, is that the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain.
GS: No, it isn>t "hotly debated." Evolutionary psychologists have merely
claimed that this is so. The real issue is "How general are the 'learning
mechanisms?'" This has been discussed within behavior analysis since the
Brelands and Garcia were current. Behaviorists continue to investigate
operant conditioning mechanisms because they have, for about 70 years now,
directly demonstrated the reliability and generality of many facts uncovered
in the experimental analysis of behavior, and they have noted circumstances
where generality is, for example, limited by species membership. Usually, of
course, the limits on generality have to do with the stimuli used and
topography of the required response, and this is consistent with what
Skinner wrote in the Phylogeny and Ontogeny of Behavior. Yes, it is harder
to train a rat to press a lever to avoid shock than it is to get it to
"shuttle" but they can still be trained to lever-press in avoidance
experiments.
MO: A fine elucidation
of this point of view is Pinker>s "How the Mind Works" - a must read for
anyone contemplating a Psych PhD program.
GS: Yeah......it is so rare to find that sort of indoctrination given the
"evolutionary psychology revolution" which is really just a subset of the
so-called "cognitive revolution."
MO: And while I>m plugging Pinker,
GS: Good idea.....Pinker should be plugged.
MO: I might as well recommend his other superb book: "The Language Instinct:
How The Mind Creates Language".
GS: Yes....just what we need, more unabashed mentalism, and further
propagation of the misrepresentations of behaviorism spawned by Chomsky and
other fools.
MO: So, as for learning to deal with Bubba, you will want to constrain the
class
of hypotheses that are considered - I guess you could try to emulate
evolution (sounds like a big project), or simply build in (the hand of god)
specialized context feature detectors...
GS: Yeah, that>ll go a long way towards AI. Oh, incidentally, I>m being
facetious.
Glen
"Michael Olea" <oleaj@sbcglobal.net> wrote in message
news:BBCD6A54.F09%oleaj@sbcglobal.net...
[quote]in article 27bc3110.0311041211.38d22c84@posting.google.com, baylor at
baylor@ihatebaylor.com wrote on 11/4/03 12:11 PM:
Another matching law question, although really it>s much more generic
than that (it has to do with generalization). And it>s probably an
FAQ. What is context?
The (few) matching law experiments i read about talked about choice in
a single context. One experiment did look at a season of basketball,
and i could believe that playing a specific team constituted a change
in context, but the matching law held over the entire season
In trying to find a way to implement this in software (and test in
video games), i>ve run into the question of whether the matching law
holds for some set of options regardless of context or if the matching
law holds for a set of options within a specific context. Also, the
question in comp.ai.games is how changing environments are handled
And there>s the strong possibility that i>m misusing the matching law
and should be looking at something else
Here>s a contrived example. i have 3 weapons, a sword, a wand of fire
and a wand of ice. At the highest level, i have a rate of
reinforcement (and assumedly response) of 33% each, so the matching
law says when i get into a fight, i have a 1-in-3 chance of using any
of the options
i fight several different monsters. Many, such as dwarves, elves and
goblins, are equally affected by each weapon. But i have three
specific monsters, a fire monster, snow monster and stone monster,
that are special. The fire monster can>t be hurt by fire, the snow
monster can>t be hurt by cold and the rock monster can>t be hurt by
fire or cold
So at a general level, where my relative rate of reinforcement for
each weapon =33%, i would choose weapons basically at random. With no
other changes, that means i have a 1-in-3 chance of using my flame
wand against a fire monster, which i know will fail 100% of the time
So it makes sense to me that i don>t track rate of reinforcement at a
general level but at the monster level. My rate of reinforcement of a
fire wand against a fire monster is 0%. That would solve the problem
Which leads to the question - is that how the matching law works? Does
it rely upon context?
If so, this leads to some questions:
1. Generalization. The rate of reinforcement of a fire wand against a
fire monster, snow monster and rock monster is 0%, 100% and 50%
respectively. Assume for now that the rate of reinforcement for the
other two weapons is 50% in each case. i now fight a brand new
monster, a giant bear or radioactive lobster or whatever. Can i use
the matching law to decide the probability of using a given weapon
given that the context is new?
i did read last night that positive associations generalize much more
than extinguishment, so a 100% rate of reinforcement might apply to
anything that didn>t match the context of the extinguishing case (i
might be using really incorrect terminology here), which is the
equivalent of broad based generalization with the memorization of
exception exemplars, but i>m not sure what you>d do where the rate of
reinforcement is a probability rather than a sure thing
2. Feature extraction. What constitutes the context? Let>s say i fight
a two-headed goat monster with my wand of fire three times and each
time i get hurt and the goat monster doesn>t. My rate of reinforcement
for using the wand of fire is 0%. In each of these cases, i fought the
same goat monster (we>ll call him Bubba) in the same room during full
moons. What would i likely learn as the context? That two-headed goat
monsters are flame retardant? That fire wands don>t work against Bubba
specifically? That fire wands don>t work in Bubba>s cave? That fire
wands don>t work during a full moon? That fire wands don>t work if
it>s Bubba AND he>s in his cave AND it>s a full moon? Do i learn all
of these associations with some given level of strength or
probablility? And how would it reconcile with the matching law?
As background, in case anyone cares, i have no reason for asking, just
curious. i like making software that acts like humans. i>ve taken some
AI and psych grad classes but am not formally enrolled. My plan is to
enroll in an AI and/or Psych PhD program Fall 2005 (it was going to be
2004, but plans change). i>d like to pursue questions similar to the
above (mostly on learning, decision making, recognition and
similarity) when i start school full time and am giving myself crash
courses on all this stuff right now
Thanks
-baylor, the curious
You put your finger on the heart of a deep issue when you wrote:
2. Feature extraction. What constitutes the context? Let>s say i fight
a two-headed goat monster with my wand of fire three times and each
time i get hurt and the goat monster doesn>t. My rate of reinforcement
for using the wand of fire is 0%. In each of these cases, i fought the
same goat monster (we>ll call him Bubba) in the same room during full
moons. What would i likely learn as the context? That two-headed goat
monsters are flame retardant? That fire wands don>t work against Bubba
specifically? That fire wands don>t work in Bubba>s cave? That fire
wands don>t work during a full moon? That fire wands don>t work if
it>s Bubba AND he>s in his cave AND it>s a full moon? Do i learn all
of these associations with some given level of strength or
probablility? And how would it reconcile with the matching law?
A short, if not very helpful, response is that people and other biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time), but are
equiped
with a genetic endowment - in effect learning much of what is the relevant
set of features constituting a "context" has been done by natural
selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
The broader point of view, one that has been hotly debated, is that the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain. A fine elucidation
of this point of view is Pinker>s "How the Mind Works" - a must read for
anyone contemplating a Psych PhD program. And while I>m plugging Pinker,
I might as well recommend his other superb book: "The Language Instinct:
How The Mind Creates Language".
So, as for learning to deal with Bubba, you will want to constrain the
class
of hypotheses that are considered - I guess you could try to emulate
evolution (sounds like a big project), or simply build in (the hand of
god)
specialized context feature detectors...
[/quote] |
|
| |
|
| Back to top |
Michael Olea
Guest
|
| Posted: Wed Nov 05, 2003 10:16 am Post subject: Re: Matching Law and context |
|
|
in article 30d633d6ee0594e2ed2a971bfbbd17eb@news.teranews.com, Glen M.
Sizemore at gmsizemore2@yahoo.com wrote on 11/4/03 5:55 PM:
[quote]MO: A short, if not very helpful, response is that people and other
biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time),[...]
GS: It was Locke, not Skinner, that used the term "blank slate." Even Watson
argued that it was not correct. BTW, anybody who thinks that we are, in
general, "following rules" when we speak our native language is a moron.
Pinker is a moron.
[/quote]
Ha, ha, ha! I love it. It would be incendiary of me to note that such
passion is highly correlated with the rear-guard actions of a dying, or long
dead, school, so I won>t mention it (jury, disregard this statement). But,
Glen, I sense you are holding back, please tell us how you really feel.
(Who mentioned Skinner - why so defensive? Muffled snickers.)
How does a moron link Pinker become director of the Center for Neuroscience
at the Massachusettes Institute of Technology? One grant at a time, I
guess.
[quote]
MO: but are equiped
with a genetic endowment - in effect learning much of what is the relevant
set of features constituting a "context" has been done by natural selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
GS: What is your evidence for this?
[/quote]
Primarily two sources: first of all, I have a twin brother, we were
separated at birth, he just walked in, and LOOK - WE>RE BOTH WEARING
SHIRTS!!! Second, an article in last week>s National Enquirer: BABY GROWS
BEARD, BORROWS CAR FOR DATE, FEEDING SCHEDULE RESPONSIBLE.
[... snip ...]
[quote]
MO: The broader point of view, one that has been hotly debated, is that the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain.
GS: No, it isn>t "hotly debated."
[/quote]
I don>t know, bro, you sound a little hot to me.
[quote]Evolutionary psychologists have merely claimed that this is so.
[/quote]
Why those scoundrels - plug 'em all! They are in violation of
the territorial imperative, invading the turf of the only true
bearers of the flame, the pellet counting observers of razor-back fish; give
'em what they deserve - you can>t help yourself: it>s in your genes. |
|
| |
|
| Back to top |
David Longley
Guest
|
| Posted: Wed Nov 05, 2003 7:29 pm Post subject: Re: Matching Law and context |
|
|
In message <BBCDB9B8.F12%oleaj@sbcglobal.net>, Michael Olea
<oleaj@sbcglobal.net> writes
[quote]in article 30d633d6ee0594e2ed2a971bfbbd17eb@news.teranews.com, Glen M.
Sizemore at gmsizemore2@yahoo.com wrote on 11/4/03 5:55 PM:
MO: A short, if not very helpful, response is that people and other
biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time),[...]
GS: It was Locke, not Skinner, that used the term "blank slate." Even Watson
argued that it was not correct. BTW, anybody who thinks that we are, in
general, "following rules" when we speak our native language is a moron.
Pinker is a moron.
Ha, ha, ha! I love it. It would be incendiary of me to note that such
passion is highly correlated with the rear-guard actions of a dying, or long
dead, school, so I won>t mention it (jury, disregard this statement). But,
Glen, I sense you are holding back, please tell us how you really feel.
(Who mentioned Skinner - why so defensive? Muffled snickers.)
How does a moron link Pinker become director of the Center for Neuroscience
at the Massachusettes Institute of Technology? One grant at a time, I
guess.
MO: but are equiped
with a genetic endowment - in effect learning much of what is the relevant
set of features constituting a "context" has been done by natural selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
GS: What is your evidence for this?
Primarily two sources: first of all, I have a twin brother, we were
separated at birth, he just walked in, and LOOK - WE>RE BOTH WEARING
SHIRTS!!! Second, an article in last week>s National Enquirer: BABY GROWS
BEARD, BORROWS CAR FOR DATE, FEEDING SCHEDULE RESPONSIBLE.
[... snip ...]
MO: The broader point of view, one that has been hotly debated, is that the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain.
GS: No, it isn>t "hotly debated."
I don>t know, bro, you sound a little hot to me.
Evolutionary psychologists have merely claimed that this is so.
Why those scoundrels - plug 'em all! They are in violation of
the territorial imperative, invading the turf of the only true
bearers of the flame, the pellet counting observers of razor-back fish; give
'em what they deserve - you can>t help yourself: it>s in your genes.
[/quote]
Tell me Michael - how do *you* ascertain the difference between science
fiction (or other creative writing/literature) and science?
--
David Longley |
|
| |
|
| Back to top |
Glen M. Sizemore
Guest
|
| Posted: Thu Nov 06, 2003 3:54 am Post subject: Re: Matching Law and context |
|
|
[quote]MO: A short, if not very helpful, response is that people and other
biological
systems are not born "blank slates" (see Steven Pinker>s "The Blank Slate:
the Modern Denial of Human Nature" for a bracing good time),[...]
GS: It was Locke, not Skinner, that used the term "blank slate." Even
Watson
argued that it was not correct. BTW, anybody who thinks that we are, in
general, "following rules" when we speak our native language is a moron.
Pinker is a moron.
[/quote]
MO: Ha, ha, ha! I love it. It would be incendiary of me to note that such
passion is highly correlated with the rear-guard actions of a dying, or long
dead, school,[...]
GS: You mean sort of like the way some evolutionary biologists talk about
Creationists? Ironic, though, that the sub-field of cognitive "science"
pushes a sort of creationism under the guise that the "processes" are
selected.
MO: [...]so I won>t mention it (jury, disregard this statement). But,
Glen, I sense you are holding back, please tell us how you really feel.
(Who mentioned Skinner - why so defensive? Muffled snickers.)
GS: Of course I knew you>d mention it. What else have you got? You certainly
are unable to make any substantive rebuttal.
MO: How does a moron link Pinker become director of the Center for
Neuroscience
at the Massachusettes Institute of Technology? One grant at a time, I
guess.
GS: Same way a whole generation of psychologists can "grow up" thinking that
Chomsky>s "review" of Verbal Behavior had something to do with what is
actually in the book. It is a long story. But I will just mention that it
has to do with the conceptual mess in which mainstream psychology finds
itself (yes, I am including EP - remember I make the assertion that EP is
really just a sub-field of cognitive psychology). In short, they make the
mistake of thinking that the facts of cognitive psychology (such that they
are) support the existence of "cognitive processes." But the latter are
assumptions, not hypotheses. To conflate theories and concepts is
devastating and has all but wrecked mainstream psychology. The position
cannot be rejected by any group of experiments because the position is
philosophy. And it is a comfortable philosophy. I>m sure that there was, at
one time, a top alchemist. You>d have to think about it for awhile, and that
is unlikely. But don>t think that I am writing for your benefit.
[quote]
MO: but are equiped> with a genetic endowment - in effect learning much of
what is the relevant
set of features constituting a "context" has been done by natural
selection.
Even something as basic as carving up the visual scene into objects, like
the moon, or the concept of "moon" as a coherent entity, is a formidable
learning problem, largely solved not through learning via experience, but
inherited in DNA.
GS: What is your evidence for this?
[/quote]
MO: Primarily two sources: first of all, I have a twin brother, we were
separated at birth, he just walked in, and LOOK - WE>RE BOTH WEARING
SHIRTS!!! Second, an article in last week>s National Enquirer: BABY GROWS
BEARD, BORROWS CAR FOR DATE, FEEDING SCHEDULE RESPONSIBLE.
GS: That>s what I thought. By the way, I figured that you would cut the
stuff on the Lashley-Wade hypothesis. After all, you made a statement that
actually can be put to experimental test, no wonder you wish to cut a
description of data that suggest that maybe all "concepts" are not the
product of natural selection.
[... snip ...]
[quote]
MO: The broader point of view, one that has been hotly debated, is that
the
brain is not a uniform substance operated on by general purpose learning
algorithms, but a dense collection of a vast number of highly specialized
gadgets, each with rules tuned to its specific domain.
GS: No, it isn>t "hotly debated."
[/quote]
MO: I don>t know, bro, you sound a little hot to me.
GS: Oh yeah......but that misses the point doesn>t it? This isn>t a
debate.....I>m just chastising your "scholarship." Were you to make a
substantive argument, rather than the usual ad populum nonsense that
"scholars" like you spew it would be a debate.
[quote]Evolutionary psychologists have merely claimed that this is so.
[/quote]
MO: Why those scoundrels - plug 'em all! They are in violation of
the territorial imperative, invading the turf of the only true
bearers of the flame, the pellet counting observers of razor-back fish; give
'em what they deserve - you can>t help yourself: it>s in your genes.
GS: What is sad is that behavior analysts - except for a few iconoclasts -
does represent the near totality of the NATURAL science of behavior. EP
could be a part (to some extent ethology was, and to some extent so was
sociobiology) but it has, instead, either adopted no particular position on
what constitutes legitimate behavioral processes, or it adopts the standard
cognitive metaphors. Yeah.....kids say "He goed across the street" because
they>re following a self-constructed rule; what nonsense. Oh.....but
hey.....it must be because natural selection makes it that way. It is just
the characteristics of the LAD. What a vacuous, silly position.
G.
"Michael Olea" <oleaj@sbcglobal.net> wrote in message news:BBCDB9B8.F12% |
|
| |
|
| Back to top |
|
