Posted by
Phil Henshaw-2 on
Sep 03, 2007; 4:03am
URL: http://friam.383.s1.nabble.com/politics-and-cliques-tp524626p524637.html
Glen,
> Phil Henshaw wrote:
> > Gee, I don't know if it helps with your philosophy, but I
> think you're
> > making a common mistake with the inverse square relation. It's an
> > indicator of complex system organization, not a design
> principle. 'A'
> > implies 'B' but 'B' in no way implies 'A'. It's like a
> thermometer, if
> > a thermometer reads 98.6 it's likely you've found a human
> but heating
> > something up to 98.6 and trying to talk to it is nutty.
> The inverse
> > square metric is a time saving empirical tool for helping
> to locate and
> > investigate complex systems. You have to look into the
> system to find
> > what makes it organized, though.
>
> Hmmm. I don't think I'm making the mistake you're citing.
> But, it's plenty likely. I make all sorts of mistakes all
> the time. [grin] To be clear, let me paraphrase what you're saying:
>
> By saying the inverse power law is a result/indicator rather
> than a design principle, you're saying that the generators of
> these results could be multifarious and not determined. I.e.
> just because a system turns out to require things extreme
> behavior and circumstances doesn't mean the organization of
> the system is the only organization that could possibly
> achieve its objectives.
I may not be speaking directly to your actual phrase, describing what
you've gathered from complexity theory: "the extent versus the
objectives of control structures should show something like an inverse
power law to maintain a balance between diversity and efficacy." I read
that as meaning that you'd design an inverse square relation into your
control systems. I don't know what actual kind of controls you may be
thinking of, or how you'd measure their diversity or efficacy, of
course.
The 'generators' of the inverse square measure are the
self-organizations of the particular complex system you then try to
understand. If you design a procedure by which self-organization
develops it's quite likely it would behave like natural self-organized
systems and be structurally different every time, and still have metrics
like the inverse square distributions of their parts which are similar.
That there might also be various different kinds of solution to a given
objective is a separate issue to me.
> You're saying that another system may be organized
> differently, achieve the same objectives, and not exhibit the
> same extrema.
>
> Is that right or did I misunderstand you?
I'm not quite sure it addresses your question, but I'm was saying the
process by which complex systems evolve does not follow an inverse
square pattern or series of steps. The measure is generally only found
in systems after they have been built by other means.
> > The network science people seem to have a better way of
> using it than
> > the other mainstream science disciplines interested in the subject I
> > think. They're looking at complex systems from the inside
> out (though
> > maybe not having quite realized that networks are artifacts of the
> > complex systems they are embedded in). Their identification of the
> > elaboration and refinement of network connections during network
> > development as the origin of the inverse square metric and
> > 'scale-free' distribution of internal connectedness of natural
> > networks is very helpful. There should logically be some kind of
> > connection with the thinking of people taking an outside in
> approach
> > to complexity, but I have not been able to figure out what it is.
>
> I don't quite buy this. But, my criticism of it takes us on
> a tangent. I'll state my criticism anyway and if you choose
> to pursue the tangent, then so be it. [grin]
>
> I don't believe there is a fundamental difference between
> constructivism and formalism. I.e. one cannot study a system
> from the inside out without also studying it from the outside
> in, and vice versa. When one uses a phrase like "studying
> complex systems from the inside out", the phrase merely
> _emphasizes_ one part of the studying. Objectively, all
> studies involve an iterative approach that cycles between
> inside and outside studies. This seems to be true of
> everything from riding a bicycle to cosmology.
Well, it's not half well enough studied, but inside and outside
perspectives of organization in systems are so very different it takes
special care to keep them straight it seems to me. I'm not even sure if
one can discuss a system as having an inside (network cell of relations)
since I haven't heard the 'news' in the journals yet and it seems to
require a radical exception to the traditional view of determinism.
Isn't the traditional view that all causation comes from the outside
still the most widespread?
One of the differences between the two perspectives is the huge
difference inside and outside views is in the information content of
your observations. If your view of the world is based on an insider's
perspective of some self-organized 'hive' of activity, say a religious
or social movement, it may be extremely hard to make sense of an
outsider's view of exactly the same thing. The insider's view is of all
the internalized connections, and the outsider's view of essentially all
the loose ends. Getting them to connect can be very difficult.
> > As far as the limits of control, don't all complex systems have
> > significantly independent design and behavior? It seems
> to me that the
> > first thing anything with independent design and behavior
> requires is
> > basic respect, otherwise you make large mistakes with it,
> right? We so
> > often forget that finding the easy ways for independent
> things to get
> > along is a great design strategy. Nature seems to like it
> quite a lot
> > for evolutionary survival too!
>
> It's not clear to me what you're saying, here. But, I don't
> really believe in "design". Design is a cognitive fiction we
> use to rationalize/justify our behavior. The causes of the
> phenomena generated by a complex system are... occult,
> occluded, at least to some extent. I think that's why we
> call these systems "complex". And its for these reasons that
> simulation is such a necessary and powerful tool in the study
> of these systems. We can't readily find "laws" that compress
> the description of the system. For less complex systems, we
> can infer these laws.
In studying natural systems it's apparent that lots of intricate
'design' develops without any 'design'. I was using first sense above,
that complex systems may develop all kinds of organization and activity
that were neither preconceived nor predetermined. Whether you can find
useful 'laws' to describe complex systems I think is like other real
scientific questions, more dependence on whether you ask the right
questions.
I was looking for years for some clear evidence that the economic
systems all act as a single complex system, behaving as a whole. The
fact that the embodied energy of economic value (btu/$GDP) is
asymptotically approaching around 8000btu/$ in all the economies of the
world seems to say it's all one system in a highly useful way. The
self-organization of the economies gives us a conversion and equivalence
between a physical measure and what humans value. I expect these things
are lying all over the place, but we're just beginning to recognize
them.
>
> Ultimately, however, even when we can (seem to) achieve some
> descriptive compression, the causes of the behavior are still
> occult. But we gain some confidence through temporal and
> spatial extrapolation (we repeat experiments through time and
> check to be sure the results are the same and we have
> different people in different locations repeat the
> experiments to see if the results are the same). Through
> such indirect "validation", we come to trust that our
> compressed description is _correct_ or true. But,
> ultimately, the causes are still occult. There is always the
> chance of a black swan.
I think it's more productive, when you're well beaten, to accept that
systems with complex internal network designs we tend not to even see
are beyond our understanding. There's still good sense to making models
of things, and developing ways of determining if the models behave like
what it imitates.
One of the interesting subjects that came up at the SASO conference is
that no one in the information network control systems field seems to
know how to do that for self-organizing and self-adapting software
controls... except random experiment. That may lead to 'gaining some
confidence', as you say, but it's not the same as the narrowly defined
uncertainties of the deterministic controls of the past.
>
> So, "don't all complex systems have significantly independent
> design and behavior"? For the above reasons, my answer is
> _no_ because "design" is a figment of our imagination. A
> better answer would be that the question is ill-formed and
> unanswerable. Complex systems are not _designed_ at all.
> They grow and evolve through the propagation of happenstance.
Well, that's kind of abstract. It's a simpler issue when talking about
real things. Any ecology or social group, etc., will have different
networks emerge within them as they develop and so they will respond
differently too. That's all I mean by independent design and behavior.
When I speak of 'designing' complex systems, as I do for architectural
and planning projects, it's more about setting up a system learning
process. Discovering how to make links between previously disconnected
parts of communities takes an effort at exploratory learning about the
disconnected parts of your community. Once you then design ways to
link them the end product is their own creative interactions which the
planners would never think of.
Instead of 'propagation of happenstance' I'd use 'development of
opportunity'. The latter covers both truly random events and the
exploratory path-finding processes also prominent in self-organization.
Phil
>
> - --
> glen e. p. ropella, 971-219-3846,
http://tempusdictum.com> A random group of homeless people under a bridge would be far
> more intellectually sound and principled than anything I've
> encountered at the university so far. -- Ward Churchill
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: Using GnuPG with Mozilla -
http://enigmail.mozdev.org>
> iD8DBQFG2wGFZeB+vOTnLkoRAq3eAKCAI+vbRLTFlFKeT7WV7k3QEeh6KwCfRLQ9
> phZzjO6yjThsCbm74urmILA=
> =IWll
> -----END PGP SIGNATURE-----
>
>