\... Echo!

From Greg Williams (921015)

Isaac Kurtzer (921014)



Hi Isaac, from the Family Man with the longest hair in the room. It would be
helpful to others on the net if you put your name and date at the beginning of
each of your posts (something similar to the format I used at the beginning of
this post would be nice; most netters have adopted Gary Cziko's suggestion of
date code as YRMODY in numbers). It makes it easier to keep track of who
replied to what. Thank you.


PCT can predict at a distance (i.e. into the future significantly further
than the immediate). it is highly predictable whether you will sucessfully
drive to the store without wrecking the car, or killing yourself or
other people --given that is your intention. however, immediate
predictions (actions carried out to oppose disturbances) will be known
only if you know "other things" : certain system characteristics
(ex. gain), and the vector sum of all disturbances at a given time
(which could be inferred by the actions but that is circular), and
the so-called "initial conditions" (this point can be and probably will\
be thrown away). that is a lot of things to know, especially if you
are not the manipulator of them all which probably isn't the case. i
realize that some of the disturbances may be remote but 1) there effects
aren't 2) by definition if it is a disturbance it disturbs. this, in my
opinion, is a rewording of stable end by various means.

Knowing an organism's controlled variable allows the knower to predict the
desired outcome which is sought by the organism. I.e., knowing that a rat is
hungry, because you have kept it away from food for several hours, you can
predict the OUTCOME when the rat is provided with food: it will eat. But
predicting the PARTICULAR actions (outputs) used by the hungry rat to satisfy
its desire for food if it is provided -- that is, the PARTICULAR actions the
rat uses to eat -- is, as you say, difficult. What is often possible in
practice (ranging all the way from cases involving Skinner boxes to everyday
"wild" human behavior) is, given good guesses about an organism's desired
outcome and knowledge about some of the organism's other characteristics and
about some of the characteristics of the organism's environment, to be able to
predict that the organism will use actions in a certain class of possible
actions to attempt to achieve the desired outcome. A hungry rat in a Skinner
box with a lever which releases food will perform an action in the class of
"moving the lever." The Porsche driver approaching a left-hand turn which he
wants to negotiate successfully will perform an action in the class of
"turning the steering wheel counterclockwise." The size of the predicted class
of actions will vary from situation to situation. I claim that often in social
interactions, party A, who is controlling for some perceptions which depend on
actions of party B, can successfully control IF B's actions, upon which A's
controlled perceptions depend, are in a certain class of actions (the class
"desired" by A). Furthermore, I claim that often in social interactions, party
A can successfully arrange (disturb or carefully not disturb) B's environment
so that party B indeed performs actions in that certain class of actions. So,
to control his/her perceptions depending on B's actions, A needn't predict B's
actions EXACTLY, and (on the basis of observations) I claim that such control
is often achieved.

From a fellow student grunt (I'm always learning!),