Autonomy and control

[From Bill Powers (920323.2000)]

Gary Cziko (920323.1410) --

If you are my disturbance pulling on your end of the rubber band, your
action WILL determine some of my perceptions, namely my perception of
where my hand is. But it will not determine where the knot of the >rubber

band is (unless of course you disturb beyond my capacity to >control).

Right. The level where disturbances can affect the controlled variable is
the level where your behavior (actions) depends on what the environment is
doing. At that level, your behavior (and your perceptions of it) are not
autonomous. But by giving up autonomous control over hand position, you
gain autonomous control over a consequence of hand position -- the position
of the knot.

The environment, however, can be more complicated than that: it may be that
control of a consequence of position of the knot might also be disturbed in
some way, so that you have to give up control of the knot's position in
order to control something else. For example, you might be trying to make
the knot stay over a randomly moving target. Now you can't control either
your hand position or the position of the knot. You have to let the target
determine where the knot is to be, in order to maintain the relationship
between the knot and the target. You have to let your hand move according
to the disturbance put in by the person at the other end of the rubber
bands. Now your autonomy applies only to the control of a relationship, and
neither to control of knot position or hand position.

Here's an even more complicated example:

In the 70s, when I was thinking up games (like Trippples), I devised a
rubber-band game that was played on a wooden board. The board was divided
into three parts by two lines parallel with the ends of the board, which
was about 24 inches long and 12 inches wide. In the middle part, which was
only about 4 inches deep, were four holes. Two of the holes were labeled
"A" and the other two "B", corresponding to players A and B, alternating
A,B,A,B across the width of the board.

The rubber bands were connected to a puck with a hole in it in which a
marble was placed. The puck could slide over the holes in the board, but if
it was centered, the marble could drop in. So the first aspect of the game
was for the players to move their ends of the rubber bands to try to get
the marble to drop in their own hole for a score (then they put the marble
back in the puck and started again from a neutral position).

The other aspect of the game came about from the fact that at the free ends
of the rubber bands were two more pucks; the players moved their ends of
the rubber bands by sliding those pucks around. Those end pucks also had
holes in them into which marbles were put. In each player's part of the
board, there were more holes. If the marble in the puck you were holding
fell into one of the holes on your side, the OTHER player got the score for
that hole.

So each player had to control for three things: the position of his own
end-puck (don't let the marble drop in a hole on your side); the position
of the other player's end-puck (try to influence the other player to drop
the marble at that end into a hole that scores for you); and the position
of the center puck (try to drop the center marble into a hole that scores
for you and keep it out of a hole that scores for the other person).

This was, of course, a game of conflict between autonomous systems. But it
was also a game of inner conflict, because as you maneuvered your puck
around trying to get the other person to give you a point, you could either
try to get the center puck over one of your holes, or maneuver the other
person into moving the other end puck over a hole at the other end of the
board -- and all the time not move your own puck directly over a hole on
your side. You had to decide what to give up control of and what to
control; you also had to guess whether the other person was concentrating
on maneuvering you to drop your marble in a hole at your end, or trying to
get the center marble in a scoring hole.

If you paid too much attention to avoiding holes on your side, the other
person could pretend to maneuver you toward a hole so that when you
counteracted that move, the center puck would be over the other person's
scoring hole. If you paid attention to getting the center puck over one of
your holes or keeping it away from the other's scoring hole, the other
person could put in a disturbance that would make you drop the marble at
your end into a hole. And of course you were trying to do the same thing to
the other person.

This turned out to be a very funny game; also very difficult and sometimes
so fast it was right at the limit of what was possible. Your attention was
darting all over the board, and you could feel things clashing inside as
you tried to avoid dangers and also make a score. I never tried to market
it. If anybody wants to try, pay me a chunk of whatever you get for it --
I'm saving up for a new computer.

As you control for variables of higher and higher level, the aspects of the
environment that can introduce disturbances become less "concrete" and more
"abstract." Think of playing the piano in front of an audience. The piano
isn't going to disturb you, and you're used to controlling against the key
resistances and so on. But if someone in the front row pulls out a
newspaper and starts reading it, this may disturb some very high-level
variables and make it difficult to put the appropriate expression into your
playing. Or suppose you're in the middle of a passionate exposition of
control theory, and you notice that the other person is stifling a yawn. If
you try to control for rapport or understanding, you're going to have to
change your style of delivery in some way. And it's the properties of the
other person that determine what you must do to maintain control of what
you're trying to do at the higher level.


Happy Memorial Day ( somehow that doesn't sound right)

Bill P.