CONCVIVI, Sunday evening 7/18/99

[Abbott_Bruce]

[From Bruce Abbott (990603.1100 EST)]

Bill Powers 990603.0000 MDT)

Bruce Abbott(990602)

I'm not going to comment on your post in detail. We have some deep
disagreements about how to analyze behavior, Bruce. I wish they were as
clear to you as they are to me. I wish that Third Party were around,
someone enough smarter than both of us to explain just where this
disagreeement lies in terms we could both agree to. But I guess we'll just
have to muddle through on our own.

Consider the following:

Bruce:

No claim is made that there is anything magical about the food. It is the
pigeon that earns the food, not the food that earns the pigeon. There is
nothing magical about an iron bar, either. Yet a magnet will attract an
iron bar, and not one made out of copper. The hungry pigeon will respond
when responses produce food, but not when they produce a hopper full of
sawdust. There is something about the food, under these conditions, that
the pigeon finds "attractive" (as we commonly say), but it is the pigeon
that plays the role of the magnet, not the food.

But the pigeon is more like an electromagnet than a permanent magnet. It
has to be hungry before making food contingent on its behavior will produce
a change in that behavior, just as the electromagnet must have power before
it will attract the iron bar. And even that analogy is not complete,
because it does not take account of myriad other factors. The boy who is
offered $10 to mow the lawn may not mow the lawn, because $10 is not enough
to compensate for the punishing effect of the work or because he can get
more immediate enjoyment from hanging out with his friends than from having
$10 in his pocket.

These two paragraphs reflect the view that an organism is like any piece of
matter, having properties which determine how it will be affected by
external forces or influences. The implication is that if you can specify
the surrounding circumstances well enough, and know the properties of the
organism, you can predict how these external influences will make the
organism behave.

Bill, I appreciate this post very much, especially its positive tone, but
your reply is based on a misunderstanding of my position. The analogies I
presented were intended to make certain points clear. The systems involved
are not control systems, but that does not mean that I reject control theory
or believe that control systems are not central to the organization of
living organisms.

Let's take another look at my examples. First, the soap bubble. Soap
bubbles are not control systems, but they _are_ equilibrium systems: the
surface of the soap film contracts owing to surface tension until the
pressure inside the bubble is high enough to resist further shrinkage of its
surface. The result is that the bubble (unless disturbed by other forces)
assumes the shape of a sphere, a shape that encloses a given volume with the
minimum surface area. A physicist might say that the bubble minimizes the
surface area -- in a sense, it "computes" the shape that will do this (a
nice analog computer, wouldn't you say?).

I raised the example of the soap bubble in response to your statement that
you didn't believe that pigeons could carry out the complex computations
required for maximizing reinforcement rate. What I am trying to get across
is the idea that we observers may see that the _result_ of the process is
that reinforcement rate is maximized (if that were true), and might describe
this by saying that "the pigeon maximizes reinforcement rate." But that no
more implies that the pigeon is doing mental arithmetic than that the soap
bubble is solving differential equations. In either case, simple process
may produce the result.

The magnet example was used to try to disabuse you of the notion that in
EAB, bits of food have magical properties that "make" pigeons do things. I
did not mean that analogy to be carried forward as a literal interpretation
of how the "magnet" (organism) is organized. You can only carry an analogy
so far before you run into trouble with it. The main points I tried to
bring out in stark relief were these:

  (1) If we analogize the pigeon's attraction to food and a magnet's
       attraction to an iron bar, then the food is the iron bar,
       not the magnet. Thus, if delivering food contingent on some
       behavioral act changes the liklihood of that act being repeated
       (under those same conditions), it is because of the pigeon's
       internal organization, not because the food has magical
       "control" over the pigeon.

  (2) This does not mean that the food has no special properties.
       There is something about the food that the pigeon is attracted
       to, owing to the pigeon's internal organization and current
       state. Again this is like the magnet and the iron bar, in that
       to be attracted by the magnet, the bar must have certain
       physical properties.

  (3) Like the attraction between an electromagnet and the iron bar,
       the pigeon's attraction for food depends on the state of its
       internal system. When the animal is not in the right internal
       state, food will not serve as a reinforcer. Different individuals
       may be organized somewhat differently; I might be willing to pay
       for steak but not for oysters, you the reverse. What may serve
       to reinforce behavior can only be determined empirically, on a
       case-by-case basis.

All of this I see as independent of the question of how the internal system
of the pigeon or the person is organized. Whatever the structure, it
behaves in ways that conform to observations. One of those observations is
that under well-controlled conditions, behavior changes in predictable ways
when certain acts are consistentently followed by certain consequences. The
scientific problem is to explain this observation.

Skinner's approach was to ignore the internal organization and instead to
develop a set of empirical principles relating manipulations of the
organism's inputs to the behavioral outcomes of these manipulations. This
strategy is in line with Skinner's interest in developing a technology of
behavioral control that could be applied in a practical way to solve
real-world behavioral problems. Unfortunately, as you say, this strategy
leads to a lot of empirical manipulations that are not guided by any
particular theory, and, whether the Skinnerians realize it or not, to the
tacit adoption of a loosely-defined theory that lumps together different
phenomena under the same name and for the most part ignores the importance
of feedback. For progress to continue, at some point one must go beyond
empirical generalizations (the so-called "functional" approach) and begin
developing a theory of mechanism. PCT, of course, is such a theory. It is
clear to me that a central component of the complex mechanism we call the
living organism is the control system, and that any approach that fails to
recognize this is doomed to failure. I am less than convinced that the
specific organization of those control systems and other components which
you have proposed will hold up to empirical scrutiny, but I could say the
same of any theory as bold and detailed as HPCT is.

So no, I don't see the pigeon as a passive physical system being nudged this
way and that by environmental forces, despite my use of passive physical
systems (soap bubbles and magnets) to try to communicate certain ideas.
Pigeons, fundamentally, are control systems and must be understood as such.

Bruce

[Bruce_Gregory]

[From Bruce Gregory (990603.1445 EDT)]

Bruce Abbott (990603.1100 EST)

Pigeons, fundamentally, are control systems and must be
understood as such.

Seems clear enough to me...

Bruce Gregory

[Bill_Powers1]

[From Bill Powers (990603.1027 MDT)]

Bruce Abbott (990603.1100 EST)--
Bruce:

Let's take another look at my examples. First, the soap bubble. Soap
bubbles are not control systems, but they _are_ equilibrium systems: the
surface of the soap film contracts owing to surface tension until the
pressure inside the bubble is high enough to resist further shrinkage of its
surface. The result is that the bubble (unless disturbed by other forces)
assumes the shape of a sphere, a shape that encloses a given volume with the
minimum surface area. A physicist might say that the bubble minimizes the
surface area -- in a sense, it "computes" the shape that will do this (a
nice analog computer, wouldn't you say?).

Everything in this paragraph is fine until you get to "in a sense." Any
time you have to say "in a sense..." or "you could say that..." you should
stop and not say whatever it was that you were about to say. You're about
to commit a metaphor or a qualitative analogy, and neither of those things
belongs in a serious scientific discussion. If you're just messing around
you can say anything you want, but then what scientist would be interested
in what you're saying?

I raised the example of the soap bubble in response to your statement that
you didn't believe that pigeons could carry out the complex computations
required for maximizing reinforcement rate. What I am trying to get across
is the idea that we observers may see that the _result_ of the process is
that reinforcement rate is maximized (if that were true), and might describe
this by saying that "the pigeon maximizes reinforcement rate." But that no
more implies that the pigeon is doing mental arithmetic than that the soap
bubble is solving differential equations. In either case, simple process
may produce the result.

If that's so, then why not just describe the simple process? If you _can't_
describe the simple process, then you are in very great danger of assuming
that something you propose is simple to carry out when in fact it would be
impossibly complex, as complex as learning algebra would be for a pigeon.
That would be a sign that you should look for a different explanation.

Maximizing, if you try to build a model that can carry it out, is generally
a complex process that requires considerable knowledge, memory, and
computation. It's quite a lot more complex than controlling. And it's
ambiguous, because without added information from somewhere, the definition
of "maximum" is most likely to refer to a local maximum, with no indication
of whether a higher maximum is nearby or far away, or whether the maximum
is in fact the maximum possible. I don't think a pigeon has what it takes
to deliberately maximize or minimize anything. Nor do I think that
"maximizing the reinforcement rate" would be good for the pigeon -- it
would most likely be fatal.

The magnet example was used to try to disabuse you of the notion that in
EAB, bits of food have magical properties that "make" pigeons do things.

I know you don't like to hear it said that way, but as long as you claim
that a reinforcer can alter the probability of a response in a properly
deprived etc. animal, that is exactly how I will say it. No such effect has
ever been observed, nor will it ever be observed. The occurrance of an
event has precisely no effect on anything else. All physical effects are
conveyed by the transmission of forces, by chemical interactions, or (we
can allow) by fielde that can be measured. "Reinforcement" is none of these
things. It's the wave of a sorcerer's wand, which make the castle appear
not through any physical process but just because of having happened. That
is how magic works: things affect other things just by existing or
happening, not through any intervening processes. If you aren't prepared to
propose a physical process and lay it out in some detail, then like it or
not you can properly be accused of magical thinking.

To "disabuse" me of this "notion," you will have to show me how
reinforcement works -- the actual physical mechanism by which it can
increase the "probability" of a behavior in the way that will result in the
specific changes that are claimed to occur. If you can't do that, all you
can claim is that reinforcers _seem_ to have such effects under proper
conditions -- which leaves the door open to conmpletely different
intepretations of the same observations. And that, in turn removes the
status of reinforcement as a "fact" that needs only to be observed to be
accepted. It becomes an interpretation to which there are alternatives.

I
did not mean that analogy to be carried forward as a literal interpretation
of how the "magnet" (organism) is organized. You can only carry an analogy
so far before you run into trouble with it.

You have already run into trouble with it. I am not interested in analogies
unless you intend them to be taken literally.

The main points I tried to
bring out in stark relief were these:

(1) If we analogize the pigeon's attraction to food and a magnet's
      attraction to an iron bar, then the food is the iron bar,
      not the magnet. Thus, if delivering food contingent on some
      behavioral act changes the liklihood of that act being repeated
      (under those same conditions), it is because of the pigeon's
      internal organization, not because the food has magical
      "control" over the pigeon.

You cannot "deliver the food contingent on some behavioral act." You can
arrange for the food to be automatically delivered if some behavioral act
occurs, but you can't make the behavioral act (and thus the delivery)
occur, nor can you (or the food) change the likelihood of that act being
repeated. It is the internal organization of the pigeon that changes what
the pigeon controls and how it does so, thus affecting what we abstractly
call a likelihood (which is not a physical variable).

(2) This does not mean that the food has no special properties.
      There is something about the food that the pigeon is attracted
      to, owing to the pigeon's internal organization and current
      state. Again this is like the magnet and the iron bar, in that
      to be attracted by the magnet, the bar must have certain
      physical properties.

The food's properties consist of its physical and chemical properties. It
has no other properties, particularly not properties that depend on
something else, such as an organism. The pigeon is not "attracted to" the
food, as if the food exerted some force on it. The food attracts other
objects only according to the product of the masses and the inverse square
of the distance, and any electrostatic charge it may carry. If the distance
between the pigeon and the food decreases, that is because the pigeon makes
it decrease, with the intention of ingesting the food. The food has no way
to cause the pigeon to approach it, or itself to approach the pigeon.

(3) Like the attraction between an electromagnet and the iron bar,
      the pigeon's attraction for food depends on the state of its
      internal system. When the animal is not in the right internal
      state, food will not serve as a reinforcer. Different individuals
      may be organized somewhat differently; I might be willing to pay
      for steak but not for oysters, you the reverse. What may serve
      to reinforce behavior can only be determined empirically, on a
      case-by-case basis.

There is no attraction of the pigeon by the food. Food does not reinforce
(strengthen, increase the likelihood of) behavior. It has no properties
that would enable it to do this, so any appearance that this is happening
must be deceptive. All these descriptions you propose are nothing more than
metaphors, not to be taken literally. Like any metaphors, they can
represent a complete misunderstanding of what is actually happening, just
as if you said that the piece of iron attracts the magnet but the piece of
copper does not (which fits the observations just as well).

The whole problem we are having here, Bruce, is this incessant use of
metaphors as if they were statements of fact. You can persuade yourself
that a given metaphor is valid, but that is strictly a subjective matter,
depending on the basis you use for judging similarity. If someone else
fails to find the metaphor convincing, there is nothing you can do to prove
that it IS convincing -- all you can do is allow yourself to be convinced,
and hope that the other person will do the same. There is nothing
compelling or unvoidable about a metaphor.

All of this I see as independent of the question of how the internal system
of the pigeon or the person is organized. Whatever the structure, it
behaves in ways that conform to observations. One of those observations is
that under well-controlled conditions, behavior changes in predictable ways
when certain acts are consistentently followed by certain consequences. The
scientific problem is to explain this observation.

I disagree: the first scientific problem is whether this is the correct way
to describe what you observe. In the first place, by setting up
well-controlled conditions, you prevent a very important phenomenon from
being seen: you do not see that when disturbances of the consequence occur,
the behavior changes in just the way needed to bring the consequence to its
former state. The acts are consistently followed by certain consequences
because of a simple causal connection in the environment; it is clearly the
behavior that is causing the consequences. And it is the changes in
behavior that _control_ the consequence. Even in your vaunted VI schedule,
it is only the behavior that maintains whatever level of food delivery exists.

The behaviorist bias is always to place the inititiation and cause of
behavior in the environment, so it is the contingency that seems to causes
the behavior that produces the reinforcer. But a view that is at least as
valid is to put the initiation and cause of the reinforcer inside the
organism, which takes advantage of any means that may exist to make the
environment provide what it needs.

So your statement of the scientific problem is not the only one possible,
as you seem to believe it is. The "conditions" of which you speak are in
part the reason for the interpretation you propose; if you change those
conditions, so disturbances of the consequence are allowed, you will see
something happening that completely changes the picture: control.

Skinner's approach was to ignore the internal organization and instead to
develop a set of empirical principles relating manipulations of the
organism's inputs to the behavioral outcomes of these manipulations.

That Skinner thought he could "manipulate the inputs" to the organism shows
only what a distant view he took. He could rearrange the environment of an
organism, but the organism determined what the actual inputs were. Skinner
never realized this, because he was so obsessed with controlling the organism.

This
strategy is in line with Skinner's interest in developing a technology of
behavioral control that could be applied in a practical way to solve
real-world behavioral problems. Unfortunately, as you say, this strategy
leads to a lot of empirical manipulations that are not guided by any
particular theory, and, whether the Skinnerians realize it or not, to the
tacit adoption of a loosely-defined theory that lumps together different
phenomena under the same name and for the most part ignores the importance
of feedback. For progress to continue, at some point one must go beyond
empirical generalizations (the so-called "functional" approach) and begin
developing a theory of mechanism. PCT, of course, is such a theory. It is
clear to me that a central component of the complex mechanism we call the
living organism is the control system, and that any approach that fails to
recognize this is doomed to failure. I am less than convinced that the
specific organization of those control systems and other components which
you have proposed will hold up to empirical scrutiny, but I could say the
same of any theory as bold and detailed as HPCT is.

I don't know what will be left of HPCT 100 years from now; I suspect it
will be the concept of hierarchical control, but not the details we
presently use to illustrate the principles.

So no, I don't see the pigeon as a passive physical system being nudged this
way and that by environmental forces, despite my use of passive physical
systems (soap bubbles and magnets) to try to communicate certain ideas.
Pigeons, fundamentally, are control systems and must be understood as such.

I agree, of course. But how are we to rid ourselves of this meddlesome
collection of metaphors that confuse the issues and stand causation on its
head?

Best,

Bill P.