🚀 go-pugleaf

[From Bill Powers (990502.0326 MDT)]

Bruce Abbott (990501.1330 EST)--

>In asserting that this implication is surprising, you have begged the
>question of whether this is in fact an implication of Simon's statement.  In
>fact it is not, and your assertion that it is, merely raises yet another
>straw man for you to vanquish.

I was trying to say something a little complicated and it may not have come
out right. I'll try again.

If Simon's statement was as Fred Nickols quoted it:

"Complex actions can reflect complexity in the acting entity's environment
and do not necessarily reflect a corresponding complexity in the acting
entity's internal organization."

... with the "can" and "not necessarily", then the only way in which
Simon's observation would introduce some new element would be for us to
interpret it as meaning the apparent complexities of behavior DO represent
ONLY complexities of the environment. Any other interpretation makes his
insight banal.

If we interpreted it to mean that some unknown fraction of the complexities
of behavior reflect complexities in the environment, and the rest (how much
also being unknown, of course) represent complexities inside the organism,
the "insight" would fall rather flat, since it would then be essentially
self-evident. "Some part of the complexities of behavior, from 0 to 100
percent, reflects complexities in the environment." That's an insight?

>Simon stated that complexity in observed
>behavior may reflect the complexity of the environment rather than
>complexity in the organism.  There is nothing in that statement to imply
>that _all_ complexities of behavior are due to environmental complexities,
>and certainly the statement says nothing one way or another about whether or
>not part or all of the observed behavior is or is not produced spontaneously
>by the organism.

The same thing applies to your way of putting it. When you say "may", you
allow for the possibility "may not." I assume that you would not be
claiming an insight for Simon if he were referring to the "may not"
alternative. The only claim worth special attention would be a claim that
the environment IS responsible for all or nearly all of the complexity. If
he meant only that some unknown part of the complexity is due to the
environment, between 0 and 100%, he wouldn't have been saying much.

>Before we get "surprised" by an implication, be should first be careful to
>establish that it actually _is_ an implication.

That's what I am doing.
>
>Second point:  In his little ant example, Simon assumed that the ant had and
>maintained the goal of reaching its nest.  So at least with respect to that
>variable, the reference was assumed to be fixed.

>Third point:  References change for some reason.  If the ant temporarily
>changes its heading reference as required to navigate around a pebble, that
>change, which adds complexity to the ant's path, was necessitated by the
>presence of the obstruction in the ant's path, and therefore can be
>attributed to complexity (presence of obstructions) in the environment,
>interacting with what may be a relatively simple organization within the
>ant.

Nothing external "necessitated" a change in the reference heading. A lump
of rock lying on the ground can't "necessitate" anything; it can't do
anything at all but lie there. The change in path can't be attributed to
"complexity." Complexity is not an agency, a variable, or a function: it's
a subjective opinion in an observer. "Gee, that path looks complex."

The real relationships involved are exact and quantitative, like all
macroscopic physical relationships. As in the Crowd program, the paths
taken by an ant are consequences of the behavior of control systems in the
ant and any active disturbances in the environment. At any moment, the
direction is determined by the momentary goals inside the ant and the
momentary states of the ant's perceptions, the perceptions being affected
equally by the ant's actions and the state of the environment. This seems
to be a concise and accurate way of stating the PCT intepretation of
observations.

In this interpretation, we do not have "complexities" affecting
"complexities." We don't have objects "necessitating" actions. All we have
are relationships among variables. You're defending a superficial way of
representing and understanding behavior as if it has some sort of
scientific meaning.

Why not attribute the change in reference heading equally to the ant? The
ant could have changed its reference heading to the left, to the right, or
not at all (it could have climbed over the obstacle without deviating to
the left or right). One can argue that it is the ant's internal complexity
that allows it to circumvent the simple passive obstacle. The "complexity"
represented by A rock resting on the ground, it seems to me, is no greater
than the "complexity" represented by an adjustment of six legs each with
three or four degrees of freedom so as to turn the ant's body to propel it
to the left, right, or over the rock while continuing to track the trail
home (or however it gets there).

>Given the assumption that the ant is attempting to reach its nest, one
>would not expect it to "spontaneously" reset this reference for no reason
>and head off in another direction.

That's begging the question, too, isn't it? You'rew simply assuming that
there are no spontaneous changes in the reference direction, and then
deducing that any changes in direction of movement must be due to external
forces, which is what you just assumed. If the ant's reference is to head
home, it can't also "spontaneously" not have the goal of heading home. But
of course if it's not bound by your assumption, it _can_ "spontaneously"
head in some other direction. You're assuming that the reference does not
change, and then using this assumption to "deduce" that environmental
complexities must have caused any changes in path. The assumption that the
reference direction doesn't change seems to imply that something can change
the ant's direction of movement without the ant's changing the reference
direction.

In the Crowd program, apropos of this, the reference direction of movement
is affected by error signals in the destination-seeking, obstacle-avoiding,
and person-following control systems, simultaneously.

How do you know an ant can have a reference level of "heading for the
nest?" Its actual reference level might be "keep the ant-scent increasing."
This trail could lead in a complex winding path back toward the nest, thus
accounting for the ant's ending up at the nest without the ant's having any
perception of "home" or a spatial direction toward home. The same winding
path might produce the impression in the observer that the ant is avoiding
obstacles, when in fact it is the trail that avoids obstacles. Or perhaps
an "obstacle" is identified by the fact that the ant's path deviates when
the ant nears some object, whereas when the path doesn't change, and the
ant goes right over a similar object, the object is not recognized as an
obstacle.

>A real ant may of course change its
>heading because, for example, it chances upon an odor trail that it prefers
>to follow, but Simon was presenting a simple example in order to make a
>point about the fact that complexity of behavior need not result from a
>complex organization of system, and in that context, introducing these other
>possibilies that may distract the ant would serve only to muddy the point
>the example was intended to illustrate.

But his basic assumption is probably wrong. The "odor trail that it prefers
to follow" (on the ground and airborne) probably IS the way home. The ant
has no mysterious sense that can tell it where the nest, well over its
horizon of visibility after a foot or so of travel, is, unless it can use
stellar navigation, solar navigation, polarization of skylight, large
landmarks like plants and trees, and so forth. It doesn't just magically
know where the nest is.

I have always found analyses like Simon's irritating, because they use
terms with no particular definition and seem to imply relationships that
exist without any particular mechanism. Words like "complexity" sound
impressive, but don't actually mean anything useful. I'm reminded of
Ashby's "Law of Requisite Variety" which, by the way, says that the
"variety" of behavior must always _match_ the variety of the environment
being controlled, and so contradicts Simon. Excuse me -- MAY contradict
Simon, depending on whether we take the "may" or "may not" route. I just
love predictions that can mean whatever the facts later reveal.

Best,

Bill P.