Universal Appetite Program (UAP)

[From Bill Curry (2002.05.07.1030 EDT)]

Rick Marken (2002.05.06.1400)

Yesterday, Linda thought of another familiar example of what looks
suspiciously like a UEC-based phenomenon: the horse accelerating back to
the stable after a ride.

Here's my data:

In my 15th summer I was employed by a local horse trader to board and break
his "barn-sour" horses. One of these nags, appropriately named Fruitball,
would vigorously resist leaving the barn from the get go. He'd defiantly
circle like a Sufi as I frantically hopped around on one foot trying to get
the other in the stirrup. Upon attaining my seat, he would rear, buck,
then set off diagonally at a stiff trot. As the distance from the barn
increased he became increasingly tractable. When finally out of sight of
the barn he became a model of comportment, neck reining flawlessly could
even guide him with my knees--but only as long as vector of forward
movement was away from the barn.

Upon turning barn-ward, i.e., when the angle between the forward vector and
the bearing to the barn was less than 90�, the barn homing reference took
over. This held true when miles from home. On the homeward leg, his gain
for getting to the stable went off the chart. Only extremely stern efforts
on a curb bit could keep him from bolting forward.

To my terror, I once experienced the full output of this control system
when a rotten chin strap broke and he took the bit in his teeth. No effort
on the reins could restrain him and I was left holding onto the pommel for
dear life. As the stable loomed into sight, I saw, horror of horrors, that
someone had left the barn door open. I hunkered down and prepared for the
worst. We entered the barn at a full gallop with me hugging his neck to
clear the beams. Then he planted his feet and skidded forward to a stop
beside his stall door. I continued on--and narrowly missed making a Wiley
Coyote silouhette in the opposite barn door. This, I learned, was hard way
to make a living.

Horse trainers attribute this behavior to the horse's controlling for
eating oats at the end of a ride (or whenever for that matter). My strategy
was to tie a bag of oats behind the saddle and ride out of sight of the
barn. I'd dismount holding onto the reins and feed him a handful of oats to
get his attention, and sweet talk him a little. Then I'd turn him
barn-ward while holding some more oats in my free hand. As he would start
to move forward I would disturb his progress with a whiff of oats then turn
him away from the barn for a few steps and feed him. Upon returning home,
no oats were given .

Over the course of several days, we got to the point where I could mount in
the barn without any dancing and ride away sans the theatrics. More
importantly I could walk him home right into the barn to dismount. He did
retain a vestige of the old system in that he'd go home more willingly than
when heading out, but it was controlled at low gain.

Hypothesis: The CV here was a program to "eat oats" and the sequence was go
to barn, stop, enter stall, and eat. This was a well oiled control program
operating in automatic mode.

TEST: I propose that a barn-sour horse's appetitive control systems
accounts for the _apparent_ increasing error and increasing output in the
run-hellbound-to-the-barn control system. When I disturbed the
go-to-barn-stall segment of the sequence by providing the oats in the
field, it was not defended and was dropped in short order, signifying (to
me :slight_smile: that that the horse-to-barn direction/distance system was under
conditional hierarchical control of the horse's programmatic goal to eat
oats.

Discussion: Bill critiqued my similar hierarchical analysis of the rat
restraint experiment by saying that eating error can't exist when eating
isn't occurring. My proposal is that the appetitive control systems in
essence amounts to eating in imagination mode--the gustatory systems are
activated, salivation primes the mouth for ingestion, attention focuses on
these sensations of eating (and necessarily the remembered food location).
We know these "hunger" systems experience increasing error with elapsed
time from the previous meal, and that they are activated by the presence of
food or even thoughts of food. There's a lot of human research that shows
that cellular energy deficits get the body biologically primed in advance
of eating. For example, increased levels of digestive hormones are found in
mouth and gut before ingestion occurs. It 's a reasonable hypothesis (to me
:wink: that this anticipatory eating error and related output will also
increase until (and only until) eating commences.

FYI: I actually hesitated to respond to your post, Rick, because of the
thin-skinned "how-dare-they object!" reactions that you and Bill have
expressed when others have posed what they obviously thought were
reasonable alternative explanations for the phenomenon under discussion. I
find this hypersensitivity to alternative views and understandings
indefensible (even if they are expressed with high gain). It certainly
chills the atmosphere of open discourse for me, and I assume, others.

For the record, I have no ax to grind except the furtherance of my own
understanding, no fixed idea of PCT correctness, and _zero_ claims to
expertise. I freely admit the problem here may be due to my obtuseness in
failing to fully grasp the issues involved. If so, I want it to be exposed
SO I CAN REORGANIZE!

My best,

Bill C.

···

--
William J. Curry Capticom, Inc.
bill@powerseed.com 603.756.9933

[From Rick Marken (2002.05.07.1400)]

Bill Curry (2002.05.07.1030 EDT)

In my 15th summer I was employed by a local horse trader to
board and break his "barn-sour" horses...

Great horse tale. Thanks.

FYI: I actually hesitated to respond to your post, Rick, because of the
thin-skinned "how-dare-they object!" reactions that you and Bill have
expressed when others have posed what they obviously thought were
reasonable alternative explanations for the phenomenon under discussion. I
find this hypersensitivity to alternative views and understandings
indefensible (even if they are expressed with high gain). It certainly
chills the atmosphere of open discourse for me, and I assume, others.

Aren't differences in references amazing! Bill's reactions chilled the
atmosphere for you and cleared it for me.

Best

Rick

···

---
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bill Powers (2002.05.08.0728 MDT)]

Bill Curry (2002.05.07.1030 EDT)--

>Here's my data:

In my 15th summer I was employed by a local horse trader to board and break
his "barn-sour" horses. One of these nags, appropriately named Fruitball,
would vigorously resist leaving the barn from the get go. ... As the
distance from the barn increased he became increasingly tractable. When
finally out of sight of
the barn he became a model of comportment, neck reining flawlessly could
even guide him with my knees--but only as long as vector of forward
movement was away from the barn.

So far so good -- the efforts to return to the barn _decrease_ as distance
from the barn _increases_.

Upon turning barn-ward, i.e., when the angle between the forward vector and
the bearing to the barn was less than 90�, the barn homing reference took
over. This held true when miles from home. On the homeward leg, his gain
for getting to the stable went off the chart. Only extremely stern efforts
on a curb bit could keep him from bolting forward.

Perhaps you could amplify on this a little. When facing toward the barn,
did Fruitball's efforts go immediately to maximum, or did they increase as
the distance to the barn decreased? If the latter, this would fit with the
approach-gradient experiments I mentioned.

> We entered the barn at a full gallop with me hugging his neck to

clear the beams. Then he planted his feet and skidded forward to a stop
beside his stall door.

Ah, so he finally reached the region where decreasing distance to the goal
resulted in decreasing effort to move in that direction (in fact, reversed
effort, decelerating to a stop).

Horse trainers attribute this behavior to the horse's controlling for
eating oats at the end of a ride (or whenever for that matter). My strategy
was to tie a bag of oats behind the saddle and ride out of sight of the
barn. I'd dismount holding onto the reins and feed him a handful of oats to
get his attention, and sweet talk him a little. Then I'd turn him
barn-ward while holding some more oats in my free hand. As he would start
to move forward I would disturb his progress with a whiff of oats then turn
him away from the barn for a few steps and feed him. Upon returning home,
no oats were given .

Over the course of several days, we got to the point where I could mount in
the barn without any dancing and ride away sans the theatrics. More
importantly I could walk him home right into the barn to dismount. He did
retain a vestige of the old system in that he'd go home more willingly than
when heading out, but it was controlled at low gain.

Hypothesis: The CV here was a program to "eat oats" and the sequence was go
to barn, stop, enter stall, and eat. This was a well oiled control program
operating in automatic mode.

Fine, I'll buy that as a possibility. The question is not about what the
controlled variable is, but how the amount of output relates to the amount
of error. In a normal control system, output decreases as error decreases.
Without deviating from the normal arrangement, we can postulate that there
is a limit on the output,. beyond which no more output can be produced no
matter how large the error is. But the UEC would require that the output
actually _decrease_ as the error goes beyond some particular size.

TEST: I propose that a barn-sour horse's appetitive control systems
accounts for the _apparent_ increasing error and increasing output in the
run-hellbound-to-the-barn control system.

But that's the normal control case. If the UEC were to be observed, we
would see an apparent increase in output as the error _decreased_ (moving
toward the barn), and a decrease in output as the error _increased_ (moving
away from the barn). You reported that you did see the latter.

When I disturbed the
go-to-barn-stall segment of the sequence by providing the oats in the
field, it was not defended and was dropped in short order, signifying (to
me :slight_smile: that that the horse-to-barn direction/distance system was under
conditional hierarchical control of the horse's programmatic goal to eat
oats.

Weak defense against disturbance that increases error (in fact, a collapse
of resistance) is what we expect from the UEC. But which is the error you
mean? There is clearly a return-to-the-barn control system, whatever it is
that is setting that reference signal. If the reference signal of going to
the barn is set by a program-level system, it is the subordinate system
than handles the details of controlling for position, not the program-level
system.

Discussion: Bill critiqued my similar hierarchical analysis of the rat
restraint experiment by saying that eating error can't exist when eating
isn't occurring. My proposal is that the appetitive control systems in
essence amounts to eating in imagination mode--the gustatory systems are
activated, salivation primes the mouth for ingestion, attention focuses on
these sensations of eating (and necessarily the remembered food location).
We know these "hunger" systems experience increasing error with elapsed
time from the previous meal, and that they are activated by the presence of
food or even thoughts of food. There's a lot of human research that shows
that cellular energy deficits get the body biologically primed in advance
of eating. For example, increased levels of digestive hormones are found in
mouth and gut before ingestion occurs. It 's a reasonable hypothesis (to me
:wink: that this anticipatory eating error and related output will also
increase until (and only until) eating commences.

Well, that's a lot of assumptions to check out. Can't we just focus on
whether or not the gradient of error and effort reverses outside a certain
range from the goal-position? If it doesn't, we have an ordinary control
system. If it does, we have a "UEC" type system -- however we explain the
observed relationship, and I'm not pushing for any particular explanation
(higher-order systems, built-in, etc.). I can't follow these complicated
arguments any more. Too old.

Best,

Bill P.

···

FYI: I actually hesitated to respond to your post, Rick, because of the
thin-skinned "how-dare-they object!" reactions that you and Bill have
expressed when others have posed what they obviously thought were
reasonable alternative explanations for the phenomenon under discussion. I
find this hypersensitivity to alternative views and understandings
indefensible (even if they are expressed with high gain). It certainly
chills the atmosphere of open discourse for me, and I assume, others.

For the record, I have no ax to grind except the furtherance of my own
understanding, no fixed idea of PCT correctness, and _zero_ claims to
expertise. I freely admit the problem here may be due to my obtuseness in
failing to fully grasp the issues involved. If so, I want it to be exposed
SO I CAN REORGANIZE!

My best,

Bill C.

--
William J. Curry Capticom, Inc.
bill@powerseed.com 603.756.9933

[From Bill Curry (2002.o5.23.1000 EDT)]

Rick Marken (2002.05.22.1800)

>Bruce Nevin (2002.05.22 15:11 EDT)

> The ill-mannered horse bolting for the barn to be fed is I believe not
> an instance for reasons that I described:

I think this one is still pending. Your explanation (hierarchical
control) might be right; but the NEC could be right, too. We don't know,
from observation alone, whether the horse gave up the goal of getting to
the barn (as per your explanation) or not (as per the NEC explanation).

I was out of pocket last week and am still catching up. Here's some more
horsing around in the form of a response to:

Bill Powers (2002.05.08.0728 MDT)

     >Me:

>out of sight of
>the barn he became a model of comportment, neck reining flawlessly. I could
>even guide him with my knees--but only as long as vector of forward
>movement was away from the barn.

So far so good -- the efforts to return to the barn _decrease_ as distance
from the barn _increases_.

Intensity of effort was stepped rather than linear as I recall. Strong
resistance accompanied with a considerable bit of coercion until we got out of
sight of the barn (like gettting a reluctant kid off to school). Seemed like
the whole out and back scenario was a learned sequence.

>Upon turning barn-ward, i.e., when the angle between the forward vector and
>the bearing to the barn was less than 90�, the barn homing reference took
>over. This held true when miles from home. On the homeward leg, his gain
>for getting to the stable went off the chart. Only extremely stern efforts
>on a curb bit could keep him from bolting forward.

Perhaps you could amplify on this a little. When facing toward the barn,
did Fruitball's efforts go immediately to maximum, or did they increase as
the distance to the barn decreased? If the latter, this would fit with the
approach-gradient experiments I mentioned.

My recollection is that I applied a consistent level of restraint on the
homeward leg until we got in sight of the barn. At that point Fruitball's
efforts to defeat restraint escalated significantly. My prediction for a
free-running, non restrained case from any distance would be that effort would
immediately approach the high limit and be sustained until the barn was in
sight, when it would max out. Come to think of it, that's not a prediction
either. I remember being tossed in a ditch by Rip, another of my charges, when
about three miles from home. I'd only had him a couple of days, and in this
case he ran full tilt via a different itinerary of back roads to reach the
horse trader's barn some five miles away where he had been stabled for quite a
while. I was able to reach some high ground where I caught momentary glimpses
of his posterior receding in the distance. He was flying. I was limping.

>Horse trainers attribute this behavior to the horse's controlling for
>eating oats at the end of a ride (or whenever for that matter). My strategy
>was to tie a bag of oats behind the saddle and ride out of sight of the
>barn. I'd dismount holding onto the reins and feed him a handful of oats to
>get his attention, and sweet talk him a little. Then I'd turn him
>barn-ward while holding some more oats in my free hand. As he would start
>to move forward I would disturb his progress with a whiff of oats then turn
>him away from the barn for a few steps and feed him. Upon returning home,
>no oats were given .

>Hypothesis: The CV here was a program to "eat oats" and the sequence was go
>to barn, stop, enter stall, and eat. This was a well oiled control program
>operating in automatic mode.

Fine, I'll buy that as a possibility. The question is not about what the
controlled variable is, but how the amount of output relates to the amount
of error.

But what if the goal is a temporal, i.e., "get oats ASAP" rather than a
locational CV? See my following analysis.

>TEST: I propose that a barn-sour horse's appetitive control systems
>accounts for the _apparent_ increasing error and increasing output in the
>run-hellbound-to-the-barn control system.

But that's the normal control case. If the UEC were to be observed, we
would see an apparent increase in output as the error _decreased_ (moving
toward the barn), and a decrease in output as the error _increased_ (moving
away from the barn). You reported that you did see the latter.

Actually, normal control _was_ my point. Rick proffered the accelerating
horse-to-stable example as a potential NEC situation. Based on my field
experience I saw it as normal sequence control. When we hypothesize the
horse-to-barn distance as the horse's CV, the return journey does appear to be
an NEC case (increasing output, decreasing error). If we entertain the idea
that "get oats ASAP" is the CV then increasing error and increasing output
would be consistent with what was observed.

>When I disturbed the
>go-to-barn-stall segment of the sequence by providing the oats in the
>field, it was not defended and was dropped in short order, signifying (to
>me :slight_smile: that that the horse-to-barn direction/distance system was under
>conditional hierarchical control of the horse's programmatic goal to eat
>oats.

Weak defense against disturbance that increases error (in fact, a collapse
of resistance) is what we expect from the UEC. But which is the error you
mean?

Alternatively, isn't a weak or absent defense against disturbance evidence that
the horse-to-barn homing system failed the TEST? An observer would say the
horse was striving to get to the barn when he was just striving to "get oats
ASAP" When I changed the oats location from barn to hand-in-front-of-nose, the
barn-homing system was summarily dropped. I could drop the reins and he'd
follow me anywhere.

There is clearly a return-to-the-barn control system, whatever it is
that is setting that reference signal. If the reference signal of going to
the barn is set by a program-level system, it is the subordinate system
than handles the details of controlling for position, not the program-level
system.

Undisturbed, the horse does get from its current location to the barn but that
seems to me like a necessary sequence element of getting oats ASAP. Reminds me
of the story about Einstein deeply engrossed in problem solving to the point he
was totally unaware of his walk across the Princeton campus from office to
home.

The leader of the Indy 500 controls at high gain to get the checkered flag and
a big trophy. To do so he has to cross the finish line first. If we posit the
distance to finish line (a locational goal) as his CV, it would appear that
there's a NEC function (increasing effort, decreasing error). Alternatively, if
he must get there first (a temporal goal), then normal control explains it
(max effort, max error).

My best,

Bill C.

···

--
William J. Curry Capticom, Inc.
bill@powerseed.com 603.756.9933