virus: Re: Philogeny, ontogeny and pers

Lumley_Ted/dal10_emlumley@dal.mobil.com
Tue, 28 Nov 95 13:02:22 -0600


Item Subject: Reply: Re: Philogeny, ontogeny
Tracy,

My personal interest in evolution is in the evolution of ideas (i.e.
learning) in complex topic areas. Your comment to Alexander struck a
nerve with me. That is when you said;

...............
"Your further talk of "environment-guided" leads me to think that either
you need to study evolution much more thoroughly, or you would do well to
take greater care in writing on the topic."
...............

While you moderated the "bite" of your comment in the final sentence, my
comments are less about intent and personal sensitivities and more about
system design.

My own exposure/engagement in email dialogues and other processes used to
evolve ideas suggests that there are two undeclared camps out there
(amongst email dialoguers), which corresond to the declared camps
embracing alternative evolutionary theories in biology and artificial
life. I need to lay out a bit of background here to make my point.

One camp is the "Means-oriented, component-scored" (MOCS)** camp which
believes that evolution proceeds via natural selection of components
(e.g. genes or "selfish genes" as Dawkins would say). This implies
optimization towards a "failsafe" system design in which each component
is as good as it can get to be.

The second is the "Purpose-oriented, behavior-scored" (POBS) camp which
believes that evolution proceeds via natural selection of the fittest
behaviors of organisms, or the fittest aggregate behaviors of groups of
organisms. How well behaviors support purpose, of course, depends on the
nature of the environment at the time of the behavior.

Naturally, the genes largely determine behaviors, but, as the POBS
proponents say; "Nature does not select on components (genes) or internal
organs". The POBS assumption leads toward system designs which are
"fault tolerant" rather than "failsafe". Fault tolerant designs allow
for small failures in components and avert "disaster" by having
components "cover for each other", as a good sports team might (e.g.
hockey, soccer, tennis doubles). This is probably why the "wiliness" of
organisms and groups of organisms seems to have played an important role
in survival and "fitness".

Where am I going with this? It is to suggest that the evolution of ideas
in discussions such as these email discussions can follow either the MOCS
or the POBS model, depending on which appears to be the most appropriate
to the purpose++. Perhaps it would be worthwhile for discussion groups
to make a conscious choice of which model they prefer to use.

The MOCS model (the dominant (Cartesian) tradition in social
organizations and discussions) doesn't worry too much about defining
purpose beyond the obligatory "Vision, Mission, Values" statement which,
once knocked off, opens the door to focusing on the perfection of the
means.

>From the data I have been exposed to in studying the behavior of
successful teams, I no longer doubt that in complex environments, the
POBS model is the superior choice. In complexity theory terms (e.g.
Gell-Mann's "Quark and the Jaguar"), one of the reasons for this
superiority associates with the exclusionary aspect of the MOCS approach,
where the focus is dominantly on the improvement of known components.
This can lead the evolving system into a suboptimum "fitness valley".
The inclusionary aspect of the POBS approach, on the other hand, does not
discriminate against new or modified components as long as they support
the right behaviors to achieve the designated purpose. Thus, in the film
"Apollo 13", the crew and the NASA team went into "POBS" simulation mode
and kluged components to achieve their purpose of bringing the crew back
home safely. Without a very compelling purpose, however, the more
mechanical MOCS model would likely have remained the dominant approach.

The second reason the POBS model is superior in complex environments is
because it implies a "fault tolerant" design. The MOCS model which
implies the optimization of each individual component runs into trouble
as the probability of component failure rises with the number of
components and their interdependencies. Thus in "Apollo 13", the failure
of a small coil in an oxygen tank (coupled with the MOCS, failsafe design
of the command module) had a cascading effect which brought the system
and crew to the brink of disaster. What kept it from disaster was the
adoption of a fault-tolerant POBS approach by the crew and the NASA
ground based team.

Having spent most of my life not being aware of my unconscious use of the
MOCS approach, I am now rather sensitive to whether there is a MOCS or a
POBS assumption underlying evolutionary "builds", whether these be in
business undertakings or in email discussions such as these.

So, the type of engagement in a POBS approach is to ask; Is the behavior
supportive of the shared purpose? If so, it deserves supportive feedback
and encouragement. If there are obvious errors incorporated in the
offering, these, by the way, should be jointly filtered out as a
secondary issue. Usually, however, "errors" are in the realm of opinion
rather than objective fact.

The scientific method itself constitutes a good rationale for not
evolving a complex system of ideas on the basis of optimizing each
component. This is because one can never prove the truth of anything,
only falsify it and move on to a better theory. In complex systems with
many components, deeper more unifying theories are more likely to emerge
from inclusionary approaches, even if there are a lot of impurities mixed
in (diversity is very valuable in the POBS approach). (e.g. you don't
need to question the accuracy of temperature and pressure measurements to
find out that there's a hurricane moving in on you, you only need to get
a broad distribution of perspectives on the current weather and look for
the deeper unities and consistencies).

Improving our knowledge of specialized topics such as extant evolutionary
theory per se is well worthwhile if that is the purpose, but if we are
looking for deeper, broader truths, then we do not know what we do not
know, and we need to take care that our addiction to analytical precision
does not cast the baby out with the bath water.

Perhaps email dialogues such as the "evolve" dialogue should split into
MOCS and POBS threads to test the relative effectiveness of these two
evolutionary approaches in the field of idea evolution? One could
observe the progress of both, but one's engagement in the dialogue would
have to respectively support, on the one hand; component optimization and
failsafe design (MOCS), and; behavior optimization and fault tolerant
design (POBS), on the other.

POBS emphasizes positive collaborative energies and trust as the fuel for
discovery, while MOCS emphasizes energies drawn from analytical
precision, competitiveness and individualism. My personal view is that
to deal effectively with increasingly complex problems, the latter "way"
must cede to the former.

Perhaps this is all old stuff which has been brought up in prior dialogue
which I have missed? If so, I extend my apologies for coming in "out of
synch".

Lastly, I apologize (to Tracy) for seizing on a perceived imperfection in
a component sentence (i.e. in reverting to MOCS mode) as an excuse for
expounding on my own area of interest.

Ted Lumley

* The terms and abbreviations "Means-oriented, component-scored" (MOCS)
and "Purpose-oriented, behavior-scored" (POBS) are for convenience. For
a discussion on the effectiveness of "bottom-up" (i.e. MOCS) and
"top-down" (i.e. POBS) evolutionary methods as are being studied in
artificial life, see "Evolutionary Programming in Perspective: The
Top-Down View", Plenary paper presented at the IEEE World Conference on
Evolutionary Computation, June, 1994. For similar ideas and evidence
supporting nature's choice of POBS, see "Beyond Natural Selection",
Robert Wesson, MIT Press, 1991

++ Ref. "The Valuated State-Space Approach and Evolutionary Computation
for Problem Solving" Keynote address to 1995 IEEE Conference on
Evolutionary Computation, Nov. 29 - Dec. 2, 1995, Perth Australia (i.e.
if you don't find this reference useful, it was at least fairly timely)