reorganization, evolution

[Avery.Andrews@anu.edu.au 950808.1535]
(Bill Powers (950806.2300 MDT))

>The big challenge is to devise a genetic strategy that doesn't give the
>organism credit (by letting it reproduce) for going partway to a
>solution of a problem. The evolution of internal reference signals gives
>the organism an internal criterion against which to judge reduction of
>error. In standard natural selection, there is only one error: death. I
>think there are limits to what this crude method of selection can do in
>a finite time.

But there is a gradient in evolution; you can have more or fewer offspring
before the final oops. Or a fixed behavioral strategy (eating things
that taste like so) can lead to the untimely death of more or fewer
of your offspring, due to chance factors (if poisonous organisms that
taste like so are rare, few of your offspring will die from eating them;
if they start to get more common, more will, and strains with different
dietary preferences will begin to have more of an advantage).

Avery.Andrews@anu.edu.au

[Avery.Andres 95080918]
(Bill Leach 950806.22:44 U.S. Eastern Time Zone)

I don't think there's much left of a substantive nature to debate about
my little speculations, since being a speculation implies a high
likelihood of being wrong! But two little examples.

Here's my idea of
a `society of control systems'. Larvas have a reference level of 1
for being `surrounded by food'. As they are `surrounded by food' to
a lesser and lesser extent, they secrete more and more of pheromone
A. Nurses have a reference level of 0 for for pheromone A. As they
perceive a greater and greater concentration of pheromone A, they get
motivated to collect food from the larder and slather it onto the
perceived source of pheromone A. Monitors have a reference level of
1 for `full larder', and as the larder gets less full, they secrete more
and of pheromone B, and Gatherers have a reference level of 0
for pheromone B; the more of it they smell, the more inclined they are
to go out and get food, and stick it in the larder. Four different
control systems, in four different individuals, interacting to produce
an ecologically significant result of `well-nourished larvae'.

And re the idea that humams are ill-adapted to their environment.
Suppose we found a population of creatures that looked like cheetahs,
but were very bad at running: 10% never figured it out at all, while
the rest seemed to run at some surprisingly low percentage of their
theoretical maximum efficiency, but almost all of them could be brought up
to speed, with no apparent ill-effects, by a bit of intervention from
a skilled animal-trainer from a circus. This would be a really wierd
situation. One possible explanation for it is that the creatures were
recently developed from ones with a different way of live, in which running
did not figure prominently, so that natural selection had not had time
to stabilize some behavioral inclinations that would lead to it being
learned properly.

If something like this is true of people, we still don't know when the
brain started cooking up ways of life that are militarily effective, but
hard for individuals to learn. There is certainly a tendency to connect it
with rise of large-scale agriculture, irrigation, etc., but maybe
it's been going on a lot longer than that, as I think I interpret
you as suggesting.

Avery.Andrews@anu.edu.au

[From Bill Powers (950809.0400 MDT)]

Avery.Andrews (950809) --

     Here's my idea of a `society of control systems'. Larvas have a
     reference level of 1 for being `surrounded by food'. As they are
     `surrounded by food' to a lesser and lesser extent, they secrete
     more and more of pheromone A. Nurses have a reference level of 0
     for pheromone A. As they perceive a greater and greater
     concentration of pheromone A, they get motivated to collect food
     from the larder and slather it onto the perceived source of
     pheromone A. Monitors have a reference level of 1 for `full
     larder', and as the larder gets less full, they secrete more and of
     pheromone B, and Gatherers have a reference level of 0 for
     pheromone B; the more of it they smell, the more inclined they are
     to go out and get food, and stick it in the larder. Four different
     control systems, in four different individuals, interacting to
     produce an ecologically significant result of `well-nourished
     larvae'.

These are truly "social control systems" in that each control system can
operate _only_ by having a physical effect on something under control by
another organism. Each organism controls the variable it senses by
disturbing a variable sensed and controlled by another organism.

To make this picture more uniform, we should try to describe all the
processes from the standpoint of the controlling system. The larva, for
example, is not equipped to perceive that it is "surrounded by food".
That is an observation from an external observer's point of view. This
could be verified by coating the particles of food with plastic that
prevents diffusion of anything from the food into the larva. Even though
the larva is "in fact" (i.e., from the wrong observer's point of view)
"surrounded by food," (the observer's characterization of a geometrical
relationship), its controlled variable is at zero, and it excretes
pheremone A as if maximum error existed -- which is true.

What the larva is controlling is some internal result of being
surrounded by food. If we could identify the chemical substances that,
impinging on the outer surface of the larva, affect the perception it is
controlling, we could shut off the production of pheremone A by
supplying this substance to the larva even though it is not "surrounded
by food."

So by this method we would arrive at a definition of the controlled
variable for each organism that is closer to what that organism is
actually controlling. The others are left for the reader as an exercise.
In such situations, the external observer's way of describing the
relationships reflects the observer's perceptions and interests, which
may be somewhat related to the variables actually under control, but are
likely to be mostly irrelevant.

When you finish describing each control system in terms of its own
perceptions, you find that the objective state of the environment drops
out of the description, leaving each system controlling only what it can
perceive by producing whatever action will affect that perception. The
external observer can see the mechanisms by which the actions affect the
perceptions, but the individual participants can't. What the observer
sees as a "social" system is really nothing of the sort: it is merely an
interaction. No one participant has any way of knowing that there are
other participants -- at least under the stated terms.

Since the individual organisms other than the larva seem to be
controlling variables that are of no benefit to themselves, we can
wonder if this picture could possibly be correct. Let's suppose that we
find that the gatherers do not eat during their lifetimes. They perform
their function and die, without ever having benefited from their own
actions. How could such gatherers ever have evolved? Their own fitness
is exactly zero. They never reproduce themselves; they exist only to
support the reproduction of a different organism. So if all the
Gatherers died out, there would be no mechanism to replace them and the
whole community would die.

It seems that this can't be correct. Something they do must have effects
that cause more Gatherers to be reproduced. One possibility is that
Gatherers have a reference level for a certain number of others like
themselves, which translates into a certain level of a perception
affected by other Gatherers in the vicinity. When the level of this
perception falls, each Gatherer begins secreting a hormone into the food
that causes a certain number of the larvae to mature into more
Gatherers. If too many Gatherers are present, they stop secreting the
hormone. So the Gatherers could control their own numbers even though
they themselves have no reproductive organs and transmit no DNA to
succeeding generations.

This begins to sound a lot like the role of cells in an organism. The
original cell reproduces and eventually the cells begin to differentiate
into different kinds of organisms. These organisms happen to be immobile
and attached to each other, but each one controls what matters to it by
acting chemically on its environment. While only a very few of these
cells actually participate in reproduction, all of them have effects on
the whole organism that influence its reproductive success, and in the
long run thereby influence their own reproductive success even though
they pass no DNA on to the next generation. The organism called "the
brain," for example, can select mates according to criteria existing in
the brain and thus influence the next generation of brains.

Some of Darwin's female finch brains, for example, select mates for
large (visually perceived) body size. These mates have large beaks and
can crack certain tough seeds that smaller finches can't manage, and can
thus survive on parts of the island where smaller birds can't survive in
a drought. The female finch brains thus assure themselves a niche where
more of the same kinds of female finch brains, which also select large
mates, will continue to exist. These brains can control the secretion of
hormones that influence the circumstances under which the body organism
has an urge to reproduce, and so can have many influences on the next
generation.

What we observe, of course, is the steady state of all the control
systems that influence the circumstances of reproduction, and of course
we interpret the result -- mistakenly -- as selection by the
environment.

···

-----------------------------------------------------------------------
Phil Runkel (8 August 95) --

     The author advised readers not to think about the way of swinging
     the arm, placing one's feet, hitting the ball with the center of
     the racket, or any such thing, but instead to pay persistent
     attention to the spot where you want the ball to go--the target on
     the other side of the net where you want to see the ball appear.

This assumes that you have already learned how to control swinging the
racket, placing the feet, etc. and can rely on those control systems to
achieve whatever reference conditions they are given. If you've never
held a racket or bounced a ball or bounced yourself around on a tennis
court, this method would probably not work very well.
------------------
To answer an outstanding private query: There are circumstances under
which reinforcements are provided to animals in such a way that too few
can be obtained to sustain life. These are the circumstances, I believe,
that generally pertain when a sustained positive relationship between
behavior rate and reinforcement rate is observed. In these conditions,
life is maintained by providing the reinforcer between experimental
sessions; typically there is an hour or two of experimental session
followed by return to the home cage where the deficit is made up over
the rest of the 24-hour period (usually in a limited way so as to
maintain a low body weight or low fluid intake). Keeping the rat in
these experimental conditions indefinitely would kill it.
-----------------------------------------------------------------------
Best,

Bill P.