Reorganization

Bruce_Gregory4 · December 13, 2003, 3:14pm

[From Bruce Gregory (2003.12.13.1013)]

Martin Taylor 2003.12.13.09.24

I think the Fox and Geese game also involves learning a new
perception, but at a rather more abstract level. I'm not sure what
kind of reorganization is involved in the new perception in Tanner's
study. But the results do seem to point in the same direction--not
perceiving and then suddenly being able to perceive the new structure.

This seems to me to be the way pattern recognition works. The brain
seems to always be looking for patterns, and when it finds one, we
acquire a new "object" of perception.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Richard_Marken2 · December 13, 2003, 11:50pm

[From Rick Marken (2003.12.13.1550)]

Dick Robertson (2003.12.12.2330 CST) --

Well, that might look a little like our results, but the graph isn't
long enough to see where it would go.

I agree. I wish he had let the kids keep solving the problem until they
were able to consistently make it through with no error. The data, by
the way, are from a dissertation done in 1980.

Their approach seems to make the research unduly complex.

I still don't really know what their method was. But I think it was
pretty simple. They just had kids (1st and 2nd graders) solve the Fox
'n Geese problem, which, if it's like Missionaries and Cannibals, is a
pretty hard problem. What they seem to be plotting is the number of
moves until the kid gets the problem into a failure state, which
probably occurs when the foxes outnumber the geese on one side of a
divide. It looks like each game is a trial that can last from 1 to 20
moves. I think the problem is solved if the kid can make 20 moves in a
row without getting into a failure state. What seems to be plotted is
individual data showing the number of moves on each trial until a
failure state occurs.

The blurb says this experiment was conducted over a three week period
so I don't know how much time there was between each trial. It looks
like each kid did 27 trials. If these trials were done over a three
week period there could have been days between some trials. I'd like to
know whether this is the case or not. If there is a lot of time between
some trails, that may account for the fact that the kids who seem to
solve the problem (indicated by a run of 20 move trials) always have
some trials during this run where they fail.

It looks like the data is relevant to reorganization because the number
of trials until failure seems to vary randomly until suddenly you start
getting 20 move (solution) trials. One kid (the last one, ID =15) even
seems to show a distinct plateau before final solution. From a PCT
perspective (and consistent the results of your study with Glines) this
would suggest that the kid is able to control the problem at one level
(relationship, perhaps, where the kid can control the relative number
of foxes and geese) but not the level that allows solution (program,
perhaps, where the kid can control the move contingencies that produce
the goal result, which is probably to have all foxes and geese on the
other side of a divide).

Martin Taylor (2003.12.13.09.24) --

I think the Fox and Geese game also involves learning a new
perception, but at a rather more abstract level.

I agree. If Fox a'n Geese is like Missionaries and Cannibals, then the
perception the kids have to learn to control is a rather complex
algorithm.

I'm not sure what kind of reorganization is involved in the new
perception in Tanner's study.

useless skill I learned while doing my doctoral research -- I would say
that it's not so much a new perception that is learned. It's learning
_which_ existing perceptual variable(s) to control in order to
accomplish the task.

But the results do seem to point in the same direction--not
perceiving and then suddenly being able to perceive the new structure.

Yes. That's the interesting aspect of the data to me. It suggests that
kids are basically behaving randomly -- reorganizing in an unsystematic
way -- until they suddenly hit on the perceptual algorithm which, if
controlled, lets them solve the puzzle _regularly_.

I think there may be ways to design these kinds of studies (using
puzzles like Fox 'n Geese) so that the transition from random
reorganization to systematic control can be seen more clearly.

Best regards

Rick

···

From my experience with learning to detect signals in noise -- a
---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Martin_Taylor2 · December 14, 2003, 12:16am

[Martin Taylor 2003.12.13.1910 EST]

[From Rick Marken (2003.12.13.1550)]

I'm not sure what kind of reorganization is involved in the new
perception in Tanner's study.

From my experience with learning to detect signals in noise -- a
useless skill I learned while doing my doctoral research -- I would say
that it's not so much a new perception that is learned. It's learning
_which_ existing perceptual variable(s) to control in order to
accomplish the task.

I don't think that's the case in Spike Tanner's study, though I could
be wrong. He explained to me that he went to considerable lengths to
make sure that nobody had ever heard sounds behave like the sounds he
was given the subjects. Nobody ever had binaural signals combine that
way in their normal life, or in previous experiments, and it was only
at the binaural level that the full signal could be heard. I don't
think that noise was an issue, at least not in the sense that signal
detection problems use noise (and I've done a lot of that, too).

I think there may be ways to design these kinds of studies (using
puzzles like Fox 'n Geese) so that the transition from random
reorganization to systematic control can be seen more clearly.

Nice if true. Nicer if one could fill the gap in levels of
abstraction between the Tanner and the Fox and Geese perceptions. Go
for it! But don't ask me to:-)

Martin

Bill_Williams1 · December 14, 2003, 6:19am

[From Rick Marken (2003.12.13.1550)]

Dick Robertson (2003.12.12.2330 CST) --

Well, that might look a little like our results, but the graph isn't
long enough to see where it would go.

I agree. I wish he had let the kids keep solving the problem until they
were able to consistently make it through with no error. The data, by
the way, are from a dissertation done in 1980.

Their approach seems to make the research unduly complex.

I still don't really know what their method was. But I think it was
pretty simple. They just had kids (1st and 2nd graders) solve the Fox
'n Geese problem, which, if it's like Missionaries and Cannibals, is a
pretty hard problem. What they seem to be plotting is the number of
moves until the kid gets the problem into a failure state, which
probably occurs when the foxes outnumber the geese on one side of a
divide. It looks like each game is a trial that can last from 1 to 20
moves. I think the problem is solved if the kid can make 20 moves in a
row without getting into a failure state. What seems to be plotted is
individual data showing the number of moves on each trial until a
failure state occurs.

The blurb says this experiment was conducted over a three week period
so I don't know how much time there was between each trial. It looks
like each kid did 27 trials. If these trials were done over a three
week period there could have been days between some trials. I'd like to
know whether this is the case or not. If there is a lot of time between
some trails, that may account for the fact that the kids who seem to
solve the problem (indicated by a run of 20 move trials) always have
some trials during this run where they fail.

It looks like the data is relevant to reorganization because the number
of trials until failure seems to vary randomly until suddenly you start
getting 20 move (solution) trials. One kid (the last one, ID =15) even
seems to show a distinct plateau before final solution. From a PCT
perspective (and consistent the results of your study with Glines) this
would suggest that the kid is able to control the problem at one level
(relationship, perhaps, where the kid can control the relative number
of foxes and geese) but not the level that allows solution (program,
perhaps, where the kid can control the move contingencies that produce
the goal result, which is probably to have all foxes and geese on the
other side of a divide).

Martin Taylor (2003.12.13.09.24) --

I think the Fox and Geese game also involves learning a new
perception, but at a rather more abstract level.

I agree. If Fox a'n Geese is like Missionaries and Cannibals, then the
perception the kids have to learn to control is a rather complex
algorithm.

I'm not sure what kind of reorganization is involved in the new
perception in Tanner's study.

useless skill I learned while doing my doctoral research -- I would say
that it's not so much a new perception that is learned. It's learning
_which_ existing perceptual variable(s) to control in order to
accomplish the task.

But the results do seem to point in the same direction--not
perceiving and then suddenly being able to perceive the new structure.

Yes. That's the interesting aspect of the data to me. It suggests that
kids are basically behaving randomly -- reorganizing in an unsystematic
way -- until they suddenly hit on the perceptual algorithm which, if
controlled, lets them solve the puzzle _regularly_.

I think there may be ways to design these kinds of studies (using
puzzles like Fox 'n Geese) so that the transition from random
reorganization to systematic control can be seen more clearly.

Best regards

Rick

···

-----Original Message-----
From: Control Systems Group Network (CSGnet) on behalf of Rick Marken
Sent: Sat 12/13/2003 5:50 PM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Reorganization

From my experience with learning to detect signals in noise -- a
---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Dick_Robertson · December 16, 2003, 3:01am

[From Dick Robertson,2003.12.14.2200CST]

Rick Marken wrote:

[From Rick Marken (2003.12.13.1550)]

> Dick Robertson (2003.12.12.2330 CST) --

> Well, that might look a little like our results, but the graph isn't
> long enough to see where it would go.

I agree. I wish he had let the kids keep solving the problem until they
were able to consistently make it through with no error. The data, by
the way, are from a dissertation done in 1980.

> Their approach seems to make the research unduly complex.

I still don't really know what their method was. But I think it was
pretty simple. They just had kids (1st and 2nd graders) solve the Fox
'n Geese problem, which, if it's like Missionaries and Cannibals,

I believe they are identical.

is a pretty hard problem. What they seem to be plotting is the number of
moves until the kid gets the problem into a failure state, which
probably occurs when the foxes outnumber the geese on one side of a
divide. It looks like each game is a trial that can last from 1 to 20
moves. I think the problem is solved if the kid can make 20 moves in a
row without getting into a failure state. What seems to be plotted is
individual data showing the number of moves on each trial until a
failure state occurs.

Oh, OK that seems to make sense. I never got that straight. I got caught
up trying to figure out what purpose their complicated math was serving. I
still can't figure that out.

The blurb says this experiment was conducted over a three week period
so I don't know how much time there was between each trial. It looks
like each kid did 27 trials. If these trials were done over a three
week period there could have been days between some trials. I'd like to
know whether this is the case or not. If there is a lot of time between
some trails, that may account for the fact that the kids who seem to
solve the problem (indicated by a run of 20 move trials) always have
some trials during this run where they fail.

It looks like the data is relevant to reorganization because the number
of trials until failure seems to vary randomly until suddenly you start
getting 20 move (solution) trials. One kid (the last one, ID =15) even
seems to show a distinct plateau before final solution. From a PCT
perspective (and consistent the results of your study with Glines) this
would suggest that the kid is able to control the problem at one level
(relationship, perhaps, where the kid can control the relative number
of foxes and geese) but not the level that allows solution (program,
perhaps, where the kid can control the move contingencies that produce
the goal result, which is probably to have all foxes and geese on the
other side of a divide).

Martin Taylor (2003.12.13.09.24) --

> I think the Fox and Geese game also involves learning a new
> perception, but at a rather more abstract level.

I agree. If Fox a'n Geese is like Missionaries and Cannibals, then the
perception the kids have to learn to control is a rather complex
algorithm.

That would make sense. If you consider the hunt for the correct algorithm a
program, it would be controlling for the successful sequence of moves. As
the kid is learning by trial and error he might stumble on a winning
sequence, but without reorganizing to "perceive" the algorithm he could hit
now and then on a winning sequence randomly.
That would be similar to what some subjects did in R&G. They stumbled on a
"win"
(i.e. reversing the machine score to below 0) but were not able to state the
winning concecpt (anticipate) and hence would stay on the penultimate
plateau when rerunning the game.We could ordinarily not get them to rerun
until they could state the concept. They would say, "Well, I won, didn't
I?" and consider their volunteering over. One difference as I see it is that
in R&G there were 3 distinct plateaux when they game was proporly won and
the S had the concept. (1) control button=light relationship. (2) control
light sequence (thus minimizing machine score) (3) anticipate sequence
(reversing machine score). In the Willet & xxx study I think there would be
only two to a win. (1)(some measure of time/movement, either time to win or
lose, or number of trials to win or lose. (2) minimizing of time/trials
upon reorganzing gain the winning algorithm.

What do you think?

> From my experience with learning to detect signals in noise -- a
useless skill I learned while doing my doctoral research -- I would say
that it's not so much a new perception that is learned. It's learning
_which_ existing perceptual variable(s) to control in order to
accomplish the task.

> But the results do seem to point in the same direction--not
> perceiving and then suddenly being able to perceive the new structure.

Yes. That's the interesting aspect of the data to me. It suggests that
kids are basically behaving randomly -- reorganizing in an unsystematic
way -- until they suddenly hit on the perceptual algorithm which, if
controlled, lets them solve the puzzle _regularly_.

I think there may be ways to design these kinds of studies (using
puzzles like Fox 'n Geese) so that the transition from random
reorganization to systematic control can be seen more clearly.

I wonder if it might not be as simple as telling the kid, "There is a way to
win this game in 20 (minimum number) moves. Please keep playing until you
can do that 3 (x) times in a row."

···

Best regards

Dick R.

Richard_Marken2 · December 16, 2003, 7:28pm

[From Rick Marken (2003.12.16.1130)]

Dick Robertson (2003.12.14.2200CST)

One difference as I see it is that
in R&G there were 3 distinct plateaux when they game was proporly won and
the S had the concept. (1) control button=light relationship. (2) control
light sequence (thus minimizing machine score) (3) anticipate sequence
(reversing machine score). In the Willet & xxx study I think there would be
only two to a win. (1)(some measure of time/movement, either time to win or
lose, or number of trials to win or lose. (2) minimizing of time/trials
upon reorganzing gain the winning algorithm.

What do you think?

I agree. I think that's what we see, when there is a plateau. The kids first
learn how to make the moves (unfortunately they don't report time to make a move
in the data I sent) and then they get the algorithm, if they get it.

Best

Rick

marken.vcf (92 Bytes)

Kenneth_Kitzke_Value · December 18, 2003, 12:43am

[From Kenny Kitzke (2003.12.17)]

Martin, you have made some terrific posts in the last few weeks with great ellucidation of PCT/HPCT concerning conflict, reorganization, etc., that even the Master applauded. I do so too and apologize for not saying this earlier. My intent is to encourage more posts from you for they have helped me and others grasp some details about the theory and the model.

I assume you are not defining a sudden change in grasping a new perception as a necessary condition for “reorganization?” Or, are you?

Martin_Taylor2 · December 18, 2003, 8:38pm

[Martin Taylor 2003.12.18.1537 EST]

[From Kenny Kitzke (2003.12.17)]

<Martin Taylor 2003.12.13.09.24>

<I think the Fox and Geese game also involves learning a new
perception, but at a rather more abstract level. I'm not sure what
kind of reorganization is involved in the new perception in Tanner's
study. But the results do seem to point in the same direction--not
perceiving and then suddenly being able to perceive the new
structure.>

I assume you are not defining a sudden change in grasping a new
perception as a necessary condition for "reorganization?" Or, are
you?

Not at all. Discovering a new perception is only one of many possible
_results_ of reorganization.

Martin