Influence; e. coli model

[From Bill Powers (920907.0900)]

Greg Williams (920907) --

Let's stay with the basic ideas for a while and make sure we
understand each other.

Your "control by disturbance" is my "manipulation." Manipulation

works >as you say, requiring "cooperation" of the manipulee. I must
say, >however, that "control by disturbance" is important in another
way to >BOTH the controllee AND the controller, since they ARE both
controlling >in ways contributory to the outcome. And, in this sense,
I think what >we control for IS important to us.

I'm controlling variable v by means of action a. You apply a
disturbance d to v. This causes v to depart from v*, my reference
level for it, by approximatly d/myloopgain; my action causes a to vary
enough to produce this result, so that a is nearly -d. So by varying
d, you can cause a to vary. This is your control of my action. You can
have a reference level a* for a. My action a will come to a state of
about a*/yourloopgain.

By this way, this takes no cooperation from me -- I don't have to pay
any attention to my own action. All I have to do is continue doing
what I was doing before: keeping v near to v*. From my point of view,
your variations of d are just another disturbance. It doesn't matter
to me whether those variations are systematically aimed at a goal of
yours or are random.

By our postulates (which forbid conflict), I must not have any goal
that forbids my action a to be controlled by you. So my action can't
be "important" to me -- that is, I can't have any goal for it, but
must be willing to let it vary as you choose. I can't have any
preference for a particular state of my action if you are to be able
to control it without conflict with me. This procedure works best if I
(higher level) pay no conscious attention to my actions at all, but
just let the (lower-level) control system operate them.

It has just occurred to me that we are probably using the term
"important" in different ways. I suspect that you're using it to mean
_objectively_ important -- that is, important in ways unknown to me
but known to the manipulator (in the manipulator's opinion). The
manipulator may believe that if he can get me to bring my action into
the state a*, something good for me will result, or something bad for
me will be avoided. The manipulator may, for example, have the opinion
that if I could be maneuvered into learning the moves that result in
taking square roots correctly, this action by me will be of future
benefit to me even though I'm indifferent to it now and am not doing
it "on purpose.".

Is this what you're talking about when you speak of importance?

ยทยทยท

----------------------------------------------------------------------
Hans Blom (920907) --

I puzzled mightily over your e. coli Pascal model until I went back
and read your description: "In my model, coli has neither a counter
(memory) nor a sensor to sense gradients. I assume it is too
primitive for either."

What I was puzzled about was the lack of sensing of the gradient --
there can be no control system without sensing. I was saying to
myself, "Why, this is just a random-walk generator -- how could it
ever go up or down a gradient?" The answer is, of course, that it
can't, except by accident. This is why it always escapes off the
screen eventually.

To make this model behave like e. coli it's necessary to make the step
size, or the delay between tumbles, depend somehow on the sensed rate
of change of concentration. Your model is much too simple; the real e.
coli (the intestinal bacterium) does have such rate-of-change chemical
sensors, and does vary the interval between tumbles according to the
time rate of change of concentration.

What you have shown is useful, in that it shows that a bacterium that
is too simple -- that is not a control system -- can't do what e. coli
actually does.

Rick, why don't you post the code for our e. coli model? I'm sure that
Hans can translate the Apple-specific parts into his version of
Pascal.

By the way, Hans, could you set your right margin to about 70 or 72?
Your lines are exceeding the screen width here, and I end up with lots
of single words on a new line. I suspect this is true for most others
too.
--------------------------------------------------------------------
Best,

Bill P.