Experimental Framework
Abstract
The following text outlines a possible experimental structure
for the Closing the Loop series. Drawing upon the ideas of Maturana
and Varela's autopoietic units, elementary systems theory and
the ideas of autocatalysis in general, the framework presented
here may contain the right ideas to frame various series of experiments.
In connecting these ideas, we come to a framework that guides
exploration, leading us to some of the more interesting questions
and hopefully allowing us to determine which parts of the answers
are in fact interesting.
The following text surfs over various ideas from systems theory,
whether of the biological, electromechanical or social, and collects
the ideas of a calculus of variations. We start with the biomechanical
unit, which we take to be our elementary unit of interest. Looking
at the way in which this unit relates and refers to other units,
biomechanical and other, we develop an idea of a calculus of variations
on the flows into and out of this unit. From here we begin to
frame some questions relating to the analysis of these flows and
theit interrelations.
In the unadulterated state, the biomechanical unit is an object
experiencing perceptual influences that act upon certain interfaces,
causing variations in the system. This system reacts at many levels,
as many as there are levels to the biomechanical system. In particular
we are interested in the autosomatic aspect, the automatisms,
the responses of the biomechanical unit outside of some kind of
conscious "control". Closing the loop of autosomatic physiological
data to perceptual permutations using various models will allow
the analysis of various biomechanical constraints and capabilities.
The Biomechanical Unit
Biomechanical systems must have a border, a limit, a skin that
defines their boundaries with the outside world. Our biomechanical
unit is a system with an internal effect occurring influenced
(though not necessarily caused) by the exterior world, permutations
effected by the world; this will be called the input. Certain
properties of the unit will be visible, measurable, observable,
these will be regarded as the system's outputs. Note that we do
not necessarily make some kind of quantum assumption about observations
necessarily effecting the system. We situate ourselves somewhat
semiclassically, a pretense of objectivity. This will be of course
seen to be a false assumption, but rather than fall into raptures
of doubt and uncertainty, we propose to take a seperate approach
that validates the subjective perception of the experimental subject
as an objective value. But we shall return to this in more detail,
for now let us assume a classical objective stance.
Previously we have dealt with a school of biomechanical thought
that relies upon a conscious effort, the higher brain functions,
attempting to re-unify the so-called "higher" and "lower" brain
functions, traditionally divided along the frontal lobe / cerebellum
axis, or even trying to reunify the forebrain and the physical
completely. This is not the way to progress for now, we need to
balance our approaches. Whole body intelligence, the autosomic
systems, immunity, waste disposal, healing, growth, unconcious
reactions, these are the touchstones. In particular, the areas
of the body that are traditionally outside the control of the
concious mind, at least in the Western tradition (wherein we find
ourselves).
Biomechanical units abound. Perhaps the canonical example of a
biomechanical unit is the human individual in its public or private
sphere. But this is only one particular example. Moving sideways
we can naturally look at other mammals and animals, plants and
other meso-scale living entities. Falling down the ladder of scales,
we can regard various subentities as biomechanical units; the
microbes that surround us, either in their single states or as
a mass, the digestive tract in which they collaborate, the immune
system which may attempt to fight them off. Taking as a cue the
idea of an assembly of microbes being a biomechanical unit, we
can also look at groups of animals, humans included, with this
lens, treating various collections of bodies as a single biomechanical
unit. Between these scales we can begin to enclose certain pieces
of hardware into biomechanical units, students with collections
of books in an exam, people with artificial limbs or pacemakers,
the bicyclist or hang-glider. In all these cases we have some
defined body that makes up the unit. Though in all cases there
is some flow across this boundary, there is also some idea of
a skin, a surface at which this flow takes place, that defines
the edge of the biomechanical unit as it is defined with this
particular point of view. We also see that we will often subsume
one unit inside another, the immune system inside the person inside
the computer user inside the hacker network.
A calculus of variations.
"Information is a difference that makes a difference". We investigate
a calculus of differences, of variations, a systematisation of
the process of altering perceptual constraints, correlations of
input variations to output oscillations. Genes are not specific
pieces of DNA, they are genotypic changes in DNA structure that
cause changes in the phenotype, the other DNA being environmental
as is the cellular environment or even the effect of DNA from
other bodies, one might even say ("The Extended Phenotype") that
the limits of the phenotype expression of genetic material is
not necessarily limited by the unit body that it is carried within.
The partial derivatives in the physicist's arsenal presume that
the other variable are held constant and we can look at infinitessimal
changes in one variable causing infinitessimal differences in
the function. Game theoretic analysis is often based around the
idea of locating strategies that are optimal given that the other
participants hold their strategies fixed. We are surrounded by
systems and sciences where objects are analysed by varying one
small part of them in isolation and monitoring the changes in
the overall structure, then attempting to recombine these changes
in a way that allows us to predict global behaviour under many
simultaneous changes.
In linear systems, we can add the effects of different variations
of input, the responses add in the same way the inputs add. In
experimental systems we can restart the machine, push the reset
button, restart the simulation. We are dealing with biomechanical
systems where this will often be difficult or downright impossible.
Linearity is probably our first, and definitely a main analytical
problem. The response of a system to a sum of changes is not the
same as the sum of the individual responses to these changes.
This is the ever-present failure of reductionist approaches, reiterating
it is almost banal. But the expression science has its etymological
roots in the same place as schism and sword, that is, to cut,
to seperate. The way that science manages to be so successful
is to reduce systems to elementary pieces and to analyse them
in isolation, cut off from richer interactions with a complex
environment. It is emphatically not the case that this approach
is futile, it is apparent that this approach has many rewards.
Although many people cry for the end of reductionist approaches
to understanding the world, the methods will continue to be used
for the simple reason that they work. The reason that they work
has a lot to do with where one defines the border of the system
in question. If a system is divided in such a way that there is
little interrelation between the parts, and that interrelation
can be simply defined and analysed, then an analysis of the system
as the sum of the two subsystems will be successful. If, however,
there is no such division, then the system in question must be
regarded as a whole, it is "irreducible" in a strict scientific
sense. Subsystems that interact linearly can be seen as examples
of decomposable systems, the behaviour of the system as a whole
is simply the sum of the behaviours of the parts. Systems that
have other such summing machanisms are useful, but perhaps the
greatest problem is to locate the natural lines of separation.
This determination of natural lines of seperation may be helped
by the ideas of a theory of biomechanical units. If one can divide
a system into a collection of biomechanical units in some kind
of systematic way, units with well defined interrelations, perhaps
even very simple interrelations, than one can begin to analyse
these units individually, and perhaps even find ways to sum the
behaviours of these systems in such a way as to obtain the behaviour
of the whole system as a sum of its parts.
The second major problem indicated above is that of the "start
state". We cannot reset most biomechanical systems, once they
have been started, they are off and running and there is no red
button to reset, reinitialise and restart them to observe their
behaviour once again in the same of a different context. There
is no simple solution to this, no clever rewording of the problem
where we attempt a workaround by redefining our terms, or even
by developing new terms. It would seem that one of the ubiquitous
features of biomechanical systems is the existence of long term
correlations, memory effects, stored information, a history of
sorts that comes with every nontrivial system. It may be the case
that such problems can be overcome with careful work, but somehow
we doubt it; perhaps even the attempt to find ways of returning
to zero is morally suspect when dealing with human or other living
subjects. The definition of black boxes of various kinds, whether
they be models of memory or computationally intractable systems,
oracles and such, may be a road out of this mess. Attempts to
take the unanalysable aspects of a system and to reframe them
as a generic but unknown dynamic system may pan out. A general
systems theory begins to deal with this by attempting to define
the complexity of a system independently of the internal structure
of the system; ideas of dimensionality, free variables, universality,
degrees of freedom occur repeatedly in economic, sociological,
mathematical, physical, psychological and computation models of
complexity.
Permutations
Regarding a biomechanical system as a unit is of great help in
analysing its structure. The definition of borders of the unit
can be at times difficult, but in view of this difficulty there
is also the benefit of probable correctness. Many different definitions
of the border of a unit, or rather, there are many different overlapping
biomechanical units in our world, many contained within one another.
In this knowledge, it is often good, and will be of value to our
researches, to define biomechanical units that include several
other biomechanical sections, or even some non-biomechanical parts.
In particular, taking a biomechanical observer and the observed
object, we can regard this collection as a biomechanical unit.
We can then modify the connection between the observed and the
observer, using various mappings or permutations. The inputs to
this biomechanical system are not light levels or fluctuations
of pressure, but rather the parameters of modifications to these
quantities. This new, larger system is once again a biomechanical
system in the same way that the previous system was.
Now we can apply a science of variations, we are acting in a system
with an intrinsic structure, we do not take as our parameters
the entire information flow, but rather the permutations to this
flow. We do not attempt to refer to an external observed object
as a seperate and thus objectively defined object, we are more
interested in the perception of this object by the subsumed biomechanical
entity. The input to this unit then becomes the modifications
to the perceptual flow between the object and the observer, this
can be more carefully defined and investigated than attempting
to deal with modifications to an arbitrary input.
Closing the Loop
It is to be expected that the observables of the system, the output,
relate something of the internal structure of the system. Even
mathematically in systems theory it is provable that certain properties
of the internal sytructure of a dynamical system can be determined
using only the time series of the output, for instance the dimension
of the internal state. These observables can also be used to measure
certain meta-phenomena. The action of forming a loop, of using
the variations in this output to vary the input parameters in
a causal and preprogrammed way, is the overarching framework of
the program. The resonances and oscillations of this loop, of
the parameters and variations thereof that can be discerned upon
it, are the area of focus.
The multitude of methodologies and metaphors that one can employ
here are a bonus, the testing of falsifiable pseudoscientific
theories is one of many aims. It is important that we scavenge
as many possible theories of control or feedback as we possibly
can from the various fields in which these theories are developed.
Given various theories of control, feedback in linear and nonlinear
systems, interactions of agent complexes as models of the mind,
market forces and psychological modelling in the stock market,
toy universes plundered from technical institutes, we begin to
develop ways of interacting with biomechanical units, to develop
experiments that test theories in all possible situations, or
only in one, hopefully interesting situation.
Conclusion
The above discussion of loop closing and scientific modelling
leads us to a conclusion where we expect that experiments formulated
in such a context, feeding output data via appropriate modification
methods back to vary the input, may lead to some interesting phenomena
in the biomechanical pseudosciences. Abstraction and modelling
play an important function here, moving into new realms of free
association, but we also note that the experiment must remain
in the forefront if we are not to fall into traps of impotence
that are often associated with ivory tower or corporate career
scientism or garage crackpot paranoia.
Version 2 tb april 98