[Martin Taylor 980628 10:01]
[From Bruce Nevin (980627.1947 EDT)]
The far more interesting question is why coercion is not the rule of the
day? Why control systems do what we call forbearance, tact, politeness,
forgiveness, compassion, apology, and so on. There are some ready
suppositions, many of them stemming from the observation that coercion
cannot be successful and "coercion" requires continuous vigilance and
control systems, even weaker ones, resist disturbances to their control,
even those produced by stronger ones, may control for not being in the same
environment with a bully, may agree about that and cooperate in it, and so
on. It would not be all that difficult at least to sketch a model of this,
so that we could begin to model components of it with the larger picture in
mind.
I've read only as far as May 13 in the "coercion" debate, but I did see this
one. Forgive me if I cover ground already discussed in the intervening month.
This question has been thoroughly discussed more than once in the time I've
been on CSGnet, so I don't think I will be saying anything new by asserting
that the answer is a natural, one might almost say inevitable, consequence
of reorganization in a context of other reorganizing hierarchies. Here's a
rewrite of the old argument.
Let's start off with a few assumptions about reorganization:
1. Reorganization can affect perceptual input functions, ECU output
functions, linkages among existing ECUs (Elementary Control Units), and
the generation of new ECUs as well as the elimination of existing ECUs.
In other words, it can affect what one perceives and what one does to
influence one's perceptions.
2. Reorganization is more likely to happen when the behaviour of the
perceptual control hierarchy is not keeping the intrinsic variables near
their genetically set reference levels.
3. Reorganization is more likely in a region of the perceptual control
hierarchy subject to sustained or increasing error than in a region of
the hierarchy that is controlling successfully.
A consequence of these assumptions is that so long as there are parts of
the hierarchy that are not controlling well, and at the same time the
intrinsic variables are not being well maintained near their reference
levels, the poorly controlling parts of the hierarchy will change. Changes
may affect what perceptions are controlled or what actions are used in
controlling perceptions.
···
------------------------------
Consider first a hypothetical organism that lives in an environment empty
of other living (reorganizing) organisms. The environmental feedback
functions available to it are fixed by the laws of physics and chemistry,
and by the material resources it can perceive and manipulate.
When this organism acts to control a perception, its actions may have side
effects that influence other controlled perceptions, reducing the precision
of control for these other perceptions. Over time, reorganization will
tend to reduce the cross-influences, making the action outputs of each
ECU tend to become orthogonal to the perceptual inputs of other ECUs.
(Parenthetically, in my readings of messages up to May 12, I saw several
assertions that conflict can happen only when two ECUs try to control
the same perception at different reference values. This is false--at least
the word "only" makes it false. We discussed this all a few weeks earlier.)
It is likely that the reorganization within this organism not only will
tend to make the output distribution functions of the various ECUs
orthogonal to the perceptual input functions of all the other ECUs, but
also will tend to align the output functions with their own perceptual
input functions. If they didn't, energy would be unnecessarily expended,
increasing the likelihood that some intrinsic variables would not be kept
near their reference levels. The consequence is that the controlled
perceptions within a layer of the organism's hierarchy will tend to become
orthogonal to each other. Conflict, in the weaker sense of "mutual
disturbance of control" will be effectively eliminated, and the organism
will eventually come to control optimally in the environment of physical
laws and resources in which it finds itself.
So far, so good. Reorganization tends to eliminate the mutually disturbing
influences of one control system on another. But what of those influences
by which the actions of one control system reduce the strength of
external disturbances on another, or by which the actions of one enhance
the effective output of another? It is quite possible for the side-effects
of the actions of one ECU to shield the perceptual signal of another from
some source of disturbance, or to enhance the power of the other's
output.
Not only is it possible, but if there are enough ECUs having inputs from
enough different components of the environment, it is almost certain that
there is at least one ECU whose actions shield some other. (I use "shield"
to include both a reduction in the disturbance influence and an enhacement
of the power or gain of the ECU being shielded). And as Stuart Kauffman
showed in "At Home in the Universe," such a possibility leads almost
inevitably, if there are enough ECUs, to the existence of a circular
chain of side-effect benefits, whereby the actions of A improve the
ability of B to control, the actions of B help C, of C help D,....,
and of Z help A.
A set of control systems that includes a cycle of side-effect support will
be _more_ stable and more able to control against _all_ the external
disturbances than will a set in which the members merely control orthogonal
perceptual functions with orthogonal output distribution functions. Such
a set will tend to be more stable against reorganization than a set of
purely orthogoanl control units, and the more such mutual support cycles
the set contains, the more stable it will be.
But what of the ECUs in the set that are not shielded by the actions of
others? The actions of the other ECUs are still likely to disturb its
perceptions. The ones in the cycle will control nicely, but the ones left
out will have more difficulty--and will tend to reorganize. Their actions
may be countered by the actions of the stable cyclic set, in which case
they may reorganize to use different actions to influence the same
perceptions. Indeed, this countering of their actions may be an aspect of
the shielding of some ECUs by others in the set. They may also reorganize
by changing the perceptual input function that defines the CEV--the function
of environmentally observable entities corresponding to their perceptual
signal.
Reorganization is not a property of a single ECU, at least not under the
assumptions at the head of this message. It is a property of the
hierarchy, though it may be focused around a poorly controlling ECU. If
the intrinsic variables stay near their reference levels despite the poor
performance of one ECU, there will be little reorganization. The ECU can
continue indefinitely to perform poorly. But this situation is unlikely,
given that the action of the poor ECU is likely to disturb the perceptions
of other ECUs, and a poor controller necessarily emits more action for
a given disturbance than does a good controller. So it is more probable
that an ECU outside of the stable mutual-support cycles will change
through reorganization than that it will not. In changing, it may come
to a structure orthogonal to the rest of its level of the hierarchy, but
it will be more stable if it finds a structure in which it is at least
partly shielded by the actions of the stable set. More subtly, it will
also be more stable if its actions come to shield some other ECUs,
because in this case the intrinsic variables are more likely to
be stabilized by increased effectiveness of the control actions of
the shielded ECUs.
The upshot of all this is that in our hypothetical isolated organism, the
hierarchy (if it is large enough) will most probably come to consist of
ECUs, most of which fulfil specialized roles (as seen by an outside
analyst). Each, in performing its own little control function, does things
that help to stabilize the perceptual signals of some few other ECUs.
But there may also be a few "bad citizen" ECUs that do not participate
in this mutual support structure. These will in many cases be reorganized
so that they at least do not disturb the rest of the structure, if they
don't come to help it. The "structure" consists--from an external
viewpoint--of ECUs specialized for many different functional roles. These
"functional roles" are the side-effects of the control actions of the ECUs.
The ECUs themselves just control their own perceptions, and cannot sense
their effects in their "functional roles." Nevertheless, for the stability
of the hierarchy as a whole, the functional roles are important.
One should note that reorganization results in improvement only in the long
run. Many reorganization "moves" lead to a condition worse than the original,
and so it is possible--even likely--that new "bad citizens" will be created
from time to time. Over a long period, however, more and more interlocking
cycles of mutual support will be created, and "bad citizen" ECUs will
have a harder and harder time controlling, leading to their quicker
alteration by reorganization, and their assumption of beneficial roles
in the community of ECUs (beneficial, that is, to successful stabilization
of the intrinsic variables).
---------------------
Now consider our organism in the same physical environment, but with the
addition to its environment of many other organisms, of the same or
different structures.
The foregoing argument can be repeated, almost word for word, substituting
"organism" or "control hierarchy" for "ECU," with one _major_ exception.
The intrinsic variables that influence the rate of reorganization are
_within_ each organism, rather than being external to the individual
ECUs we considered above. This means that reorganization is _local_ to
the individual. Does this matter to the argument? I think not in any
important way.
If an organism is controlling well within its environment--i.e. bringing
its perceptions to their reference levels reliably, despite disturbances
from other organisms--it is unlikely to reorganize. If, in the environment
of other control hierarchies, it cannot control well, it is likely to
reorganize until it does. And if the actions of one hierarchy could
shield another from external disturbances, and there are enough hierarchies
in the environment, mutual support loops will almost inevitably form.
Such suppoort loops are the substructures of societies, whether the
societies be of slime-mold fruiting bodies, of chickens or of philosophers.
Hierarchies whose actions tend to disturb the perceptions of other
hierarchies are likely to find those actions opposed. That's how "shielding"
often works. The "society" coerces (in the sense the word was being
used prior to May 12) obnoxious (or noxious) organisms, whether they be
of the same species or another. Members of a stable support cycle are
fulfilling "roles," and the more cycles they belog to, the more simultaneous
roles they play--all in the course of controlling their own perceptions
as best, and a selfishly, as they can.
------------------
If you look at it in this way, as a matter of the stability of systems
against reorganization, you find that "forbearance, tact, politeness,
forgiveness, compassion, apology, and so on" are all aspects of shielding
against disturbance among organisms capable of language--and I think among
many other organisms as well. Think of dog packs or poultry, for example.
When one member improperly plays a role as defined by the "rules" of
the society, bad things happen to it/him/her. The disturbing influences
are mitigated either by reorganization within the "bad citizen," or by
active shielding (banishment, jail, shunning) or withdrawal of services
(refusing to help the bad citizen, firing from a job...).
Life is based as much on altruism (providing benefits to others with no
immediate prospect of return benefit) as on competition (achieving control
by denying to others the possibility of control using the same environmental
variables). This is true at the basic level of cellular chemistry, and it
is true for individuals within a society--human or otherwise.
And that is why it is a good thing to pay taxes. You may not know precisely
who will benefit from your tax payment, but it is almost a sure thing that
you will benefit from someone's taxes in the long run. Perhaps the best
economist was Jesus, who said you should cast your bread upon the waters,
and it will return to you manyfold. It is the greedy who starve, in the end
(if they are not guillotined first). In something I read (I can't remember
whether it was before May 12 or in one of the random messages since
then that I have dipped into) Mike Acree said that he did not see how
anyone who knows PCT could avoid being an anarchist. In light of the
foregoing, I don't see how anyone who knows PCT would dare to be an
anarchist!
Martin