Control, si! Selection by consequences, No!

[From Rick Marken (941227.1115)]

Bruce Abbott (941227.10 EST) --

Here's the diagram you requested:

Thanks. Now I can see that your "selection" model is an
implementation of the PCT reorganizing system; the controlled
perception is a binary variable that could be called "survival": the two
possible values of this variable are "survive" and "dead". The implicit
reference for this variable is "survive". As long as the controlled
variable has the value "survive", there is no error and, hence, no
action. When the controlled variable has the value "dead" there is
action -- a change in the parameters of the system. The change remains
in place as long as the system has the value "survive". So, like the PCT
reorganization system, parameter changes that result in improved
control delay reorganization; parameter changes that result in poorer
control lead quickly to another change in parameters. The result is a
"selection" model that is a control system; this system controls the value
of the variable "survival", keeping it at the value "survive". Here is my
version of your diagram, with the reference signal and controlled
variable (survival) explicitly labelled:

                                                 > Environment of
                                                 > reorganizing system
                                random |
                            +---->parameter -----|----------+
                            ^ variation | |
                            > > >
                            e | V
               r(survive)-> C <--p (survival)<-- |-----Nutrient Control
                                                 > System
                                                 > ^ |
                                                 > > V
                                                 > Environment

Note that I have also shown the environmental boundary of your
reorganizing system; it is not the same as the environment in which
the "Nutrient control systems" operate.

Because you have built a control system, its behavior must be
viewed as selection OF consequences rather than selection BY
consequences. The consequences, as you note, are the state of the
"survival" variabley. You say:

The essential fact here is that the consequences (in this case
survival/death) determine whether or not there will be a new round
of (random) selection.

Now you can see that this is not the case; it's the difference between the
value of the perceptual variable, p,(survival) and the reference signal,
r, that determines whether or not there will be "a new round of
(random) selection".

In this sense they act as criteria by which the program selects whether
or not to generate a new set of control parameters.

And now you can see that this is incorrect as well; the criterion that
determines whether or not a new set of control parameters are to be
generated is the setting of the reference signal, r. It is with respect
to r that it is determined whether or not the consequences (the values
of the survival variable) should lead to the generation of a new set of
parameters.

Given that the consequences, as criteria, determine what decision
the program will make, one can fairly say that selection is determined
by consequences.

I assume that you can now see that this is a misconception.
Consequences are not criteria; reference signals (INSIDE THE
CONTROL SYSTEM) are. Selection is done by the control system,
NOT by the consequences of the control system's actions. I really
think you should read my "Blind men..." paper -- very carefully;-)

If you still think that consequences "determine what decision
the program will make", try changing the setting of the reference signal
in your model. You do this by changing the part of your program
where it says

if bug = "dead" then
        change parameters

to

if bug = "survive" then
        change parameters

This changes the reference for survival from "survive" to "dead";
"survive" now creates an error in the survival control system. Now the
_same_ consequences have very different effects than they did before
the change in reference signal; consequences (like "dead") that used to
cause control parameters to change now cause no change; consequences
(like "survive") that used to cause no change now cause the parameters to
change. The result is that the controlled variable (survival) is
maintained at the new reference level, "dead". This change in
behavior was "selected" by a change in the control system; it was not
selected by consequences; the consequences are the same as before
(survive or dead).

Purposeful behavior (such as that done by your reorgnizing system)
always involves selection OF consequences; selection BY consequences
has nothing to do with purposeful behavior (control).

Best

Rick