NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/bit/listserv/csgl/998

Text File | 1992-09-12 | 9.4 KB | 175 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!paladin.american.edu!auvm!VAXF.COLORADO.EDU!POWERS_W
X-Envelope-to: [e-mail]
X-VMS-To: @CSG
MIME-version: 1.0
Content-transfer-encoding: 7BIT
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Sat, 12 Sep 1992 20:52:06 -0600
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         "William T. Powers" <[e-mail]>
Subject:      Autonomy
Lines: 160

[From Bill Powers (920912.1400)]

Dennis Delprato (920912) --

RE: Autonomy.

You're coming very close to the position I think I have. As you say,
the development is delicate. On the one side there is the pit of
solipsism, with the world denied and made imaginary or subject to
whim. Down the other side lurk the positivists, who must eventually
advocate controlling people for their own good according to objective
criteria. At the apex where these two slopes -- one slippery and the
other studded with sharp rocks -- converge, there is, I think, a
narrow ridge which, followed to its end, will show us a self-
consistent view of the environment, of other people, and of ourselves.

Greg Williams is perfectly correct in wanting us to develop a theory
of interaction. To cast the problem in terms of manipulation, however,
is to beg a question, because we don't know, without analyzing the
possible mechanisms, what can actually be manipulated by another and
what can't.

The underlying question is what we mean by autonomy: in what respects
are organisms autonomous and in what respects not? Organisms clearly
depend on their environments (including other organisms) for
sustenance of various kinds. Furthermore they have inherited needs
about which they have no choice. Understanding autonomy is especially
difficult in a hierarchical system, where combating a disturbance at
one level entails altering goals at lower levels. Goal-directed
behavior is not, per se, autonomous -- in a hierarchy.

There are certain processes in the human organism that are carried out
simply because of the way the organism is put together, inside.
Control itself is an example, as is reorganization. The environment
contains no means of carrying out these processes for an organism,
either to help it or hinder it. A human being must acquire perceptual
functions that produce consistent perceptions, perceptions that vary
in sensible ways and relate to each other without contradiction. The
environment can't do that kind of making-sense; only a brain can. Once
a perceptual signal exists, it is the brain that must carry out
comparisons with reference signals, to generate an error signal. The
environment does not inform the human organism of how it, the
environment, should be, nor does it tell the organism what constitutes
a discrepancy with the organism's goals. And it is the inner
mechanisms of the organism that must find out how to convert error
signals into those choices, amounts, and directions of lower-order
actions that will make the errors smaller. The actual process of
finding those means, as opposed to what means will be found, is
strictly a product of internal mechanisms for change.

On the other hand, the organism can produce outputs, but it can't
determine what the consequences of them will be. All the brain can do
is drive the effectors. It is the environment that then serves up the
consequences. The environment can generate consequences on its own; it
can vary enough so that a given action seldom creates the same
consequences twice, and it contains independent forces that affect
perceptions. This is my best evidence for existence of a Boss Reality
that exists and has properties independent of my experiences.

Because the environment has properties and contains independent
sources of disturbance, the organism must learn by itself which output
variations will control its perceptions and which will not. Once the
organism has begun representing the external world in a particular
perceptual way, and once it commits to a goal, to reproducing a
particular state of that perception, it has no choice but to produce
an output that is sufficient to create that perception. The forms of
its perceptual functions determine how the environment will be
apprehended, but given those forms, it is the environment, not the
organism, that determines what actions will in fact be sufficient to
control the perceptual result. The organism must contain means of
discovering the effective actions, the means of control, but it has no
say as to what those actions will turn out to be.

When an organic control system resists a disturbance, it does so
relative to a goal-state of the affected perception. But that goal-
state itself is adjusted as it is because of other disturances that
have effects on higher-level perceptions, making them deviate from
higher-level goal-states. There is no end until we reach the top
level. Only there do we find the possibility of purely endogenous
goal-generation. Below the top level, the way goals are set is
determined by the perceptual structure that intervenes at lower
levels, and by the properties of the environment, which dictate how
resetting a lower-level goal will affect a higher-level perception.

Each level of control, therefore, comes into being through the action
of internal mechanisms for change and development, but the final
result, the control organizations that come into existence, must be
designed to work through the properties of the world that actually
exists, however we may perceive it.

Finally there are the intrinsic variables, their inherited reference
levels, and the process of reorganization driven by intrinsic error.
These are defined for the organism by its heritage. Thou shalt stay
warm, but not too warm. Thou shalt breathe, and eat, and drink or find
ways of acting that have equivalent effects. Thou shalt reproduce, and
find the experience pleasant. Thou shalt avoid injury, and find injury
painful. And so on through a list that is undoubtedly longer than any
we can now write down.

The environment can affect intrinsic variables, but it can't say what
will constitute an intrinsic variable, or a reference level for one.
The process of reorganization, being capable of altering anything in
the hierarchy of control, both perceptions and actions, overrides all
other considerations. The environment may determine what must be done
to affect a given perception in a given way, but it is the
reorganizing system that decides what will constitute a perception,
and whether any particular state of it is to be sought or avoided.

With respect, at least, to the current environment, these built-in
systems define the true autonomy of an individual. The individual will
learn to perceive the environment, and to control its perceptions
relative to particular goals, in ways that use the properties of the
environment and satisfy the requirements of the inherited control
system that I call the reorganizing system. In an advanced organism,
if the properties of the environment aren't sufficient to find a way
of eliminating intrinsic error, the properties of the environment will
be changed. The rain will not be allowed to fall, the wind will not be
allowed to blow and freeze, the sun will not be allowed to beat down
-- on this organism.

So autonomy, in any one lifetime, is awarded to the organism, in
particular to its reorganizing system. But that is only in one
lifetime.

One organism is a member of a species. There are processes of blind
variation and selective retention that span the links between
generations. The selective retention depends at least in part on
criteria endogenous to the species; even the blind variation may be in
part an act of the species rather than just a random effect of
something else. The species contains mechanisms for change. And it
must also contain reference signals that are passed down from
generation to generation, which define the reason both for starting
and for ending change.

I won't follow that trail any further; the point is only to show that
the same principles of autonomy can be extended into the past, with
the environment providing the stage and acting mindlessly to disturb,
but being incapable of carrying out the processes of change, which
continue to reside in the species. This trail leads all the way back
to the first molecule that stabilized the conditions on which the
accuracy of its replication depended, and thus gave birth to life.

Here and now, we live in a world whose properties are largely unknown
to us and which determine for us the effects of our actions. We,
however, choose our own goal structures, as our means of preserving
ourselves in the state that our natures tell us is right. Between this
ultimate personal autonomy and the impersonal events in the nonliving
universe, there comes to be a hierarchy of control systems that
reflects both our overriding inner needs and the conditions the
environment places on meeting those needs.

This is the best I can do for now by way of laying out the meaning of
autonomy and the relation of an organism to the world outside it. This
still leaves open the question of interactions among organisms;
organisms with similar organizations, and organisms assymetrically
related to each other. Whatever we decide to say about manipulation, I
think it will turn out to describe the situation presented here: one
organism learning how to affect its perceptions in the way it wants,
by choosing actions from among those that the environment says are the
only feasible ones.