[Martin Taylor 2004.05.13.2310]
[From Bill Powers (2004.05.13.1007 MST)]
Martin Taylor 2004.05.12.16.37 --
My argument for the necessity of PCT is based on the
very wide-ranging CLAIM of classical physics that entropy increases
over time in closed systems. It doesn't make accurate prediction of
behaviour, but it makes some prediction, and it does so at zero cost
in added hypothesized degrees of freedom for anyone who has accepted
the basic laws of classical physics. It is therefore very strngly
supported. To deny it, one has either to rework classical physics, or
invalidate the argument that classical physics applies when the
entity being examined is a living organism.I reserve a little corner of skepticism about that claim, too. It is
possible to design a control system on paper which works without any
mention of entropy, so the basic principles of negative feedback control
apply independently of the role of entropy.
I guess I haven't made myself as clear as I might have done. I
thought it was only Marc who didn't understand. So I'm reposting here
the relevant passage from my Editorial in the PCT issue of IJHCI (in
which issue, as some contributors may not know, PCT was applied at
many perceptual levels and different application areas, from simple
tracking through general Human-ComputerInteraction to self-image and
crowd behaviour). I've added a couple of short paragraphs at the
bottom to what was originally published, because the extract as
published leads on to other material. (The following was previously
placed on the ECACS Forum at
<http://www.ecacs.net/cgi-bin/discus/show.cgi?tpc=192&post=527#POST527>\).
···
--------------------
The Necessity of Perceptual Control Theory
There is one overwhelming fact about life: to survive, any organism
must in some way stabilize its essential internal chemistry in the
face of disturbances from a turbulent outer world. At least one major
theory proposes that the species that survived the mass extinction
that ended the Cretaceous were preceisely those best able to achieve
this stabilization in the face of large disturbances. Our bodies are
thermodynamically unstable, and decay away as soon as we cease to act
to counter the influences that would destroy us. Every living thing
has ancestors all of whom behaved so that they stabilized their
internal chemistry at least long enough to propagate their genes.
Protection against the buffeting of the outer world can be done in
two ways, and life employs them both. The first way is to develop
passive armour, such as a membrane, skin, or shell. But the armour
cannot be so perfect as to isolate the organism completely from the
outer world; if it did, the organism would die an entropic death.
Just to sustain a minimal internal organization, any living thing
must at least take in high-quality energy from the outer world, and
excrete less organized waste energy. Most do much more.
The second, and more important, way an organism can protect itself
against the disturbances of the world is to counter them as they
occur, actively and powerfully. To do this, the organism must be able
to sense important states of the outer world; it must be able to
compare the sensed states with desirable conditions for those states;
and it must be able to act to influence them so that it can bring
about and maintain the desirable conditions.
"To sense" means to alter some internal state, such as a chemical
concentration or a neural firing rate, in correspondence with changes
of something in the outer world. In Perceptual Control Theory, such
an internal state is called a "perceptual signal," and the value of a
perceptual signal is a "perception." To stabilize a state near some
reference condition is the technical definition of "control." Hence,
when an organism is countering the disturbances form the outer world,
it is controlling its perceptions.
"Perception," in PCT, carries no connotation of consciousness. It is
just the value of a signal. A perception may be related directly to
the current state of some property of the outer environment, but very
rarely is the connection so direct. The value of a perceptual signal
in most cases depends not only on the current state of the physical
variables that affect sense organs, but also on the current values of
other internal variables, which often depend on the history of the
organism.
The external variables that affect the value of a perceptual signal
might be very complex and context-dependent-such as those that relate
to the perception of the commitment of one's government to democratic
values-or they might be as simple as the rate at which visible
photons impinge on the retina, which relates to the perception of
brightness. Both perceptions-democratic values and brightness-
according to Perceptual Control Theory, are just values of a signal,
and neither need be conscious to be effective.
Bringing a perception of some state to a desired (reference or goal)
value with which it is compared, and maintaining it there, is control
in the strict engineering sense of the word. The perception of the
external state is what is stabilized, not the external state itself,
and still less the action that the organism uses to influence the
external state-hence "Perceptual" Control Theory. For this reason,
PCT has a core tenet: "All behaviour is the control of perception."
The actions that stabilize the perception may vary dramatically as
the environmental influences change, but a well controlled perception
varies only when its reference value changes.
Not all perceptions can be controlled. Many, such as the perception
of the height of the sun in the sky, exist as signals that one's
actions cannot influence. Other perceptual signals, such as a
perception of the democratic commitment of the government, can be
influenced only slightly by one's action. Yet other perceptions can
be controlled with ease, such as, for most people, the perceived
location of one's hand.
Perceptual control systems: engineered and human
It is not only living things that contain signal values. Inanimate
objects such as computers do, too. If those values are consequent on
states of the world outside the computer, they may legitimately be
called "perceptions." If a computer acts so as to maintain at a
reference level a signal value that corresponds to a state of the
outer world, it is controlling a perception, just as a living
organism might do. This does not mean that the computer is considered
to be alive, but it does mean that the analytic techniques of control
can be applied to the relevant operations of the computer, just as
they can to the relevant processes of a living thing. Human and
computer both control perceptions by sensing some aspect of the world
and acting to influence the resulting signal toward a reference
value, in the face of external influences that would alter its value.
From a thermodynamic point of view, a control system acts as a
refrigerator for its immediate environment. This means that it has to
use a flow of energy that allows it to export entropy to a part of
the environment unlikely to affect its controlled perceptions. In
colloquial language, no perceptual control system should "foul its
own nest."
Another way of looking at the situation from a physics point of view
is that this necessary energy flow almost certainly will give rise to
some kind of self-organized system with internal feedback loops
(implicit and continuous, as in the structure of a vortex, or
explicit and possibly even discrete). Furthermore, self-organized
systems of this kind tend to evolve toward the edge of chaos, and we
should expect the Perceptual Control networks that constitute an
individual organism to do likewise.
Martin