S-R is for the Boids

[From Rick Marken (971109.1110)]

Craig Reynolds:

I think my [boid] behaviors have more of a flavor of fuzzy
logic, stimulus-response systems, based on 3d vector values.
The connection between perceptual variables and controlled
variable (if those are the right terms) is rather indirect.

As Bill Powers (971107.1527 MST) noted, what you think of as
a fuzzy logic or stimulus-response system operates in a closed
loop; the stimulus (distance to other boids) is having an effect
on responses (boid movement) _while_ the responses have an effect
on the stimulus. If this system protects the stimulus variable
from the effects of disturbance (and it apparently does because
your boids clearly protect the distance between themselves and
other boids from disturbances such as movement of the other
boids) then it is a closed loop, _negative feedback_ control
system. Once you know this, you can design your system using the
tools of a well developed engineering discipline (control theory)
instead of trial and error.

My point is that if you build a good bridge, it doesn't matter
that you built it for the "wrong reason".

Bill Powers (971107.1527 MST) --

In PCT, we try to model not just behavior, but the behaving
organism

I think this is an important point to understand. I agree that
it probably doesn't matter in the behavioral animation world that
the boids are successful for the "wrong reason" (because they
are stable negative feedback control systems, not because they
are S-R or fuzzy logic systems). I imagine that behavior animators
want to see animation that works and could care less about _why_
it works.

But, as is apparent from your Web site, there are people
outside the behavior animation community -- for example,
people in the artificial life (ALife) community -- who see
the boids as a model of real _living systems_. In fact, the
boids _are_ a model of real living systems -- a control of
perception model. But, since the boids are described as S-R
or fuzzy logic systems, the ALife people seem to have concluded
that S-R and fuzzy logic are reasonable models of the behavior
of living organisms, which, of course, is the misconception
under which life sciences have labored for at least the last
200 years.

The next question, then, is "so what"? Who cares whether the
behavior of living systems is S-R or control of perception?
Aren't these just different ways of looking at the same process
anyway? After all, the S-R way of looking at boid behavior led
to the development of a model that is equivalent to a control of
perception model. So what difference does it make whether we see
the behavior of living systems as S-R or control of perception?

The answer to this question is also the answer to your question
about my control theory demos. You said:

However I may be missing the Big Picture, I'm not sure what
lesson to take away from these demos.

The Big Picture is a picture of why it makes a difference
whether behavior is seen as S-R, fuzzy logic, model based
output, or control of perception. It makes a difference, I think,
because behavior actually _is_ the control of perception; it
makes a difference because control of perception (like the
earth going around the sun) is the way things actually work.
All the demos at my Web page illustrate this point. They also
show that, while behavior often _looks like_ a response to
stimulation or model based output (just as the sun looks like
it's going around the earth), that is not what is really going
on.

Of course, whether it makes a difference _to you_ (or anyone)
that living organisms are actually controlling their own
perceptual experience depends on what your own goals and
interests are relative to living organisms. If all you want to
do is build convincing animations of various behaviors then
understanding behavior as the control of perception is probably
irrelevant (though I think that developing animations on the
basis of PCT could lead to far simpler and, ultimately, more
lifelike animations than what we have now).

If, however, what you want to do is understand human (or
animal) nature, then understanding behavior as the control
of perception is absolutely essential. Such an understanding
influences the very way you go about studying human nature.
Instead of studying behavior by looking for the variables
that cause people to act the way they do (the conventional
approach to behavioral research) you test to determine the
perceptual variables that people actually control. This is what
is illustrated in "The Test for the Controlled Variable" demo.
The computer is set up to determine which square's movement
(which perception) you are trying to control.

I'm glad you found the demo somewhat "spooky"; what the
computer is doing is something that's a lot like "mind
reading"; it's really "intention reading". The computer
(using the method of "the test for the controlled variable")
can determine which of the three squares you _intend_ to move.
In a sense, the computer looks past your behavior (the movement
of the three squares) to the perceptual variable that you are
controlling. Spooky, but cool, I think; and something that
could not have been done without an understanding of the Big
Picture (the fact that behavior is the control of perception).

Bill Powers says:

Wish I could program in Java, like Rick.

I think you'd be better off if your wish didn't come true
this time. I still don't feel like I can really program in
Java. I think you would be much better off wishing you could
program in Java, like Craig.

Best

Rick

ยทยทยท

--

Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/