[From Bruce Gregory (980729.1615 EDT)]
I am reading a chapter with this title in an earlier draft of Bill's new
book (which Amazon.com seems as yet unaware of). It raises many questions
for me, but I'll start with the most elementary.
"PCT offers simple definitions of reward and punishment. A reward is
anything that reduces error; punishment is anything that increases it."
Intrinsic error? Total system error? Some other error? Imagine you want me
to practice my game of horse shoes. To encourage me, you offer to give me a
dollar for every ringer I throw. You are clearly trying to control my
behavior, but it is not clear to me that this is an example of reward as PCT
uses the term. Whatever error is being reduced seems to have little
connection with my ability to accurately throw horse shoes. What
reorganization is going on that is slowed by the reward? Am I going astray
by trying to apply a technical PCT term to an inappropriate situation?
Bruce Gregory
[From Bill Powers (980729.1500 MDT)]
Bruce Gregory (980729.1615 EDT)--
Intrinsic error? Total system error? Some other error? Imagine you want me
to practice my game of horse shoes. To encourage me, you offer to give me a
dollar for every ringer I throw. You are clearly trying to control my
behavior, but it is not clear to me that this is an example of reward as PCT
uses the term. Whatever error is being reduced seems to have little
connection with my ability to accurately throw horse shoes.
That is right. It has to do with your desire for a dollar.
Best,
Bill P.
[From Tim Carey (980730.0700)]
dollar for every ringer I throw. You are clearly trying to control my
behavior, but it is not clear to me that this is an example of reward as
PCT
uses the term.
Just a technical point Bruce ... in PCT the term "behaviour" refers to the
working of the entire loop and "actions" refers to the muscle activity that
an outside agent can observe. Rewards, then, would be used to control a
person's actions and not their behaviour.
Regards,
Tim
[From Bruce Gregory (980730.0955 EDT)]
Bill Powers (980729.1500 MDT)
That is right. It has to do with your desire for a dollar.
I assume I can safely say the following. Any voluntary exchange is rewarding
for those involved. Salaries are rewards. We reward ourselves whenever we
exercise control. The government rewards its citizens in ways they are often
unaware of. I often unintentionally reward someone.
Bruce Gregory
[From Bill Powers (980730.0934 MDT)]
Bruce Gregory (980730.0955 EDT)--
I assume I can safely say the following. Any voluntary exchange is rewarding
for those involved. Salaries are rewards. We reward ourselves whenever we
exercise control. The government rewards its citizens in ways they are often
unaware of. I often unintentionally reward someone.
This seems to involve the translation, "reward" --> "giving somebody
something he wants." That doesn't quite capture its meaning for me, because
it leaves out the strategy of the giver: using the reward to induce the
rewardee to do what the rewarder wants. I discuss this at some length in
the new book.
Best,
Bill P.
[From Bruce Gregory (980730.1200 EDT)]
Bill Powers (980730.0934 MDT)]
Bruce Gregory (980730.0955 EDT)--
>I assume I can safely say the following. Any voluntary exchange
is rewarding
>for those involved. Salaries are rewards. We reward ourselves whenever we
>exercise control. The government rewards its citizens in ways
they are often
>unaware of. I often unintentionally reward someone.
This seems to involve the translation, "reward" --> "giving somebody
something he wants."
No, I was using your definition: A reward is anything that reduces error.
Bruce Gregory
[From Bruce Gregory (980730.1238 EDT)]
The chapter "Control by Reward and Punishment" frequently talks about
reorganization in the context of changing behavior or goals. Since
controlling at the program level leads to changing behavior and changing
goals, is controlling at the program level necessarily a result of
reorganization?
Bruce Gregory
[From Bill Powers (980730.1041 MDT)]
Bruce Gregory (980730.1200 EDT)--
This seems to involve the translation, "reward" --> "giving somebody
something he wants."
No, I was using your definition: A reward is anything that reduces error.
I believe that giving somebody what he wants reduces error. But there's a
different problem here: reward as the thing or situation used to to reduce
the error, vs reward_ing_ as a strategy for controlling the behavior of
others. English lets us use the word "reward" either as a verb or as a
noun, so my statement was ambiguous.
Best,
Bill P.
[From Bill Powers (980730.1046 MDT)]
Bruce Gregory (980730.1238 EDT)--
is controlling at the program level necessarily a result of
reorganization?
Yes, in the sense that the very existence of a program-level control system
is the result of reorganization, according to my theory. However, once such
a system has become organized, it can systematically vary lower-level goals
in such a way as to prevent further reorganization.
Best,
Bill P.