Wednesday, January 11, 2006
Additional reading:
Cory Casanave's paper on Data
Access
work on ontology for biological signal pathways
[126]
ß parallel discussion in generative
methodology bead thread
Discussion on informational invariance
[342], [343], [344], [345], [346], [347], [348], [349], [350]
Note from Judith Rosen
Hello Folks,
I'm pleased to help elucidate the concepts my father developed, in
this discussion. I hope it proves useful for you. Paul Werbos wrote:
the Rosen definition clearly does NOT
allow for emergent complexity. If one can start with simple underlying rules or
axioms or dynamics (either in forwards time or symmetric time or hypertime),
and generate all kinds of emergent complexity by applying these rules over a
huge field of entities... I wouldn't want to rule that out a priori.
It does actually depend on how you define "emergent
complexity". What the Rosennean definition declares impossible is the
creation of complex organization by accretion. In other words, one cannot take
a system with simple organization and add more components onto it and turn it
into a complex system. One can create very complicated, intricate systems that
way, but the organization will still be simple.
One seeming gray area involves the "addition" of a
complex system to a simple system, such as adding a human being to a computer.
That does, indeed, create a new complex system, but it has nothing to do with
the organization of the system we call "a computer" or with the process
of addition. Rather, it has to do with the human being. In fact, what that does
is create an extended human system whereby the computer then becomes a
component. Needless to say, the human being was already a complex system, via
his/her organization as a living system. So, it's not really a case of
complexity by accretion even though it can, perhaps, seem like it from a
certain perspective.
Emergent behaviors of the sort you are describing-- involving
time-- are already based on complexity because time is complex.
What Robert Rosen said cannot be possible (where computation is
concerned) is to completely digitize a complex system.
For instance, it is not possible to completely digitize the entire
system of mathematics-- which, as a human generated "language", is a
complex system in its own right. In order to completely digitize anything, one
has to render it in pure syntax, with all semantic referents removed, in its
entirety. When this was attempted with the system of mathematics, it simplified
the entailment and that generated all sorts of paradoxes.
To completely digitize something is to "take it out of
time" and make the infinite artificially finite. This changes the system,
if the system is based on relational aspects of organization. And that is the
essence of the Rosennean definition of complexity.
(On emergent behavior”) [1]
There are widely differing definitions of "emergent
behavior" or "emergent phenomena". If we have models of some system
and our models predict certain behaviors that seem to be borne out by
causality, then our system does something not predicted by the model, is that
"emergent behavior"?
I would argue that it probably isn't. It is a case of our model
being too limited in relation to the actual system being modeled. In other
words, there is more entailment involved with the actual system than we have
managed to encode into our model.
The Rosennean definition of emergent behavior is a relational one.
It has to do with the fact that an interactive relation between two or more
things can cause effects (behaviors) that could never be predicted by the
entailment of the things, themselves. We were recently discussing this on the
Robert Rosen discussion list, so I will copy in my description:
First principles: Empirics, as I learned the term, has to do with
verifying the existence of something by making it apparent to one's own senses
via some reproducible means. This originally had to do with "things",
as in material things. However, as science developed technologies which are
capable of allowing us to see wavelengths of light and hear wavelengths of
sound that are beyond the scope of our senses, or make visible to our eyes
living organisms which are too small for our focusing capacity, or detect such
things as radioactivity... it has become apparent that many of the forces
around us are not material things and that all of these forces had existed all
along but we had been oblivious to them prior to the development of technological
means for boosting our perception to a new level. The technology allows us to
observe and/or measure many things and forces that we were not able to
apprehend before except, perhaps via their effects, or perhaps not at all.
We have also, in the course of human history and scientific
exploration, discovered that there are forces which are caused by the
interaction of two or more "things"-- interactions which have effects
that could not have been predicted by the behavior of either of those things individually.
As chemistry has advanced, we have found that chemical compounds are different
from elements and that the compounds made from the elements rarely bear any
resemblance to the elements from which they are created.
My favorite example is Sodium Chloride (table salt), made from
Sodium (a toxic metal) and Chlorine (a deadly poisonous gas). The fact that
sodium chloride is absolutely necessary for human life yet it is
"made" up of elements that could end human life is a significant
fact, in my view. There are many other examples of this sort of relational,
interactive effect being well known and well documented. For example; in the
field of medicine, there is such a thing as synergistic activity; of nutrients
when ingested together versus taken separately (such as iron and vitamin C) and
such things as drug interactions, or drug-and-food interactions (like
acetaminophen and alcohol mixed in the human physiological system, causing
liver damage; or aspirin and vitamin E, causing dangerous blood-clotting
problems when ingested together).
A synergistic effect is quite real, perfectly measurable and
reproducible... and yet it isn't an actual thing, like an energy field or
radioactive particles. It's an effect--one which only exists in these examples
when certain things are interacting and then; only within a set of very
specific relations (one of which is time).
There is the further situation, well documented in many different
fields of science (including medicine) whereby two or more effects can interact
with one another, causing entirely new effects which, again, could not have
been predicted by any amount of study of the individual effects on their own.
Such an effect is every bit as measurable and reproducible as some material
thing is, yet what are the ingredients? What are the causes of this
phenomenon-- any given effect (proximate or otherwise)? Much of the time,
science doesn't know the answer to that question! We merely know that these
effects are real and cannot be ignored because, for example, many are deadly.
Consider a well-known deep-sea diver's malady: "The Bends".
Some of the relational aspects which have been "proven"
to be significant are straight-forward things like "dosage"; physical
or temporal proximity (did you ingest these two things at the same time? How
long between ingesting one to ingestion of the other?) ; length of time of
exposure (to radioactive material or to underwater pressure, etc). However,
other relational aspects are far less straight-forward. Consider things like gender,
age, body-fat percentage... It is a fact that ten short dives underwater does
not equal the effect of one long dive, even when the ten short dives add up to
far more cumulative time under pressure than the single longer dive. Why?
Science can tell us the causal basis for this, but what I'm after
is the fact that the relational situation where one has to worry about the
Bends is very specific and any other configuration of the same set of relations
may not have the same effect. In fact, there are documented, reproducible
examples of how to make what is usually a fatal interactive effect a very
beneficial interactive effect. My littlest daughter had to have bo-tox
injections into her calf muscles, last year, in an attempt to correct an
unbalanced set of interactive muscles which are responsible for the position of
the feet and ankles. Bo-tox is Botulinus Toxin: one of the most deadly toxins
known to science. It is routinely used now (aside from the wrinkle-cure use) in
cases of spastic muscles and can help in all sorts of physical rehabilitation
settings for disabled people, including people with severe cerebral palsy.
What is the difference between death and benefit? It's not dosage.
It's keeping the toxin localized, in the muscle! If my daughter had ingested
the same amount as was injected into her muscles, or if it had been mistakenly
injected into the bloodstream, she would have died rather quickly.
All of the above is by way of pointing out that the concepts that
my father developed are far from extreme, left-field concepts that would seem
outlandish to normal, intelligent people... AND we do already have plenty of
scientific ways for verifying relational effects and relational interactions.
So why is it such a stretch to begin seeing relations as having
the same or greater importance as material "ingredients"? Or
relational effects acting as if they were "things"? If relational
effects are measurably proven to be acting as if they were material things,
then we must consider them as such, it seems to me. And why is it so hard to
comprehend that in living system organization, all relations and relational
effects act like additional "parts" along with the material
components and structure? Given that temporal and proximate relations can have
such a huge impact, why is it so hard to accept that these relations are both
utilized and exploited by organism physiologies? Indeed, why is it so difficult
to see that life, as a set of behaviors of an organism, is-- itself-- a
relational EFFECT? Why is it so difficult
to see that this particular relational effect would not arise in any system
with a different type of organization that included different relations?
In Robert Rosen's development of complexity, emergence of new
properties and behaviors are due to the power of relational interactivity to
generate new effects, which could not exist without the relational aspect.
That is the essence of his view of what "complexity"
means. A complex system, then, is one which incorporates relations and relational
effects into its organization, which makes the system one which cannot be
fractionated without destroying the system (we can't take an atom apart and
have it still be an atom, nor can we take an organism apart by the same token).
Only simple systems can be taken apart and put back together, with the system
resuming all of its entailments once reconstituted. This is because simple
systems are organized with far less relational involvement (the only relational
aspect for a machine, for example, is the purely structural/configuration
type).
Does this help illuminate? Please let me know where I'm falling
short in explanation if need be.
Judith Rosen
[1] Paul Prueitt has separated text above to make more paragraphs so that the concepts are seen individually and can be discussed individually. The section title “On emergent behavior” was also editorially added – to separate the concept of complexity form the concept of emergent behavior. This separation is critical when evaluating the differences between John Doyle’s concepts and Robert Rosen’s concepts.