[From Bill Powers (930305.1330)]
Joel Judd (930305.1130) --
Nice to know we have at least one volunteer for coordinator. I
think we're not quite ready for that yet -- unless we are. Why
not start by keeping a record of all the suggestions?
···
-----------------------------------------------------------
Martin Taylor (930305.1215) --
Well, I didn't expect everyone to like my proposal. And by the
way, I dissasociate myself from your statement
But it's your baby, so you get the casting vote.
If that's the case, I won't even vote. What I know about control
theory doesn't make me an authority on journals, and even if it
did, I would expect all good PCTers to insist on controlling for
what they want, not for what I want. I have noticed that this is
generally what they do anyway, whether I give them permission or
not.
If there were a Journal of Living Systems, based around PCT as
the foundational idea, then it should incorporate all aspects
of development of PCT. That includes not only experiments, but
theory and explanation.
I can't disagree with that. Experiments, theory, and explanation
are the toolkit of science. But I don't think you can leave
experiments out and still have science. My opinion.
How do you make a "reliable demonstration" of how a proposed
Decision Making Entity interacts with one or a set of ECSs?
So far, you don't. That's why we toss such ideas and many others
around on the net. There's no point in taking them seriously
until we can think of a way to show that no other explanation
appears possible. I wouldn't submit an article on the DME, pro or
con, in the present state of the idea.
Or of whether a program-type ECS can be segregated from the
sequency-type ECSs that presumably support it?
Ditto. Until someone thinks of an experiment that will
distinguish between these concepts, they're just concepts to me.
Worth thinking about, but not worth publishing for posterity.
How do you make a reliable demonstration that PCT-based methods
of teaching a second language work? As Joel Judd once
remarked, you can't even replicate such things, even if you can
experiment with them.
Then I guess we don't yet have anything publishable on that
subject. If we don't have anything to say, should we publish
anyway?
The theoretical discussion of what could be, must be, and
cannot be, is as important to PCT as demonstrations of one or
two level models that fit real data.
I totally, heartily, and vehemently disagree. Nature pays no
attention to what we think could be, must be, and cannot be. All
such musings are based on incomplete premises and inadequate
knowledge. Maybe somewhere in the realm of such ideas are things
that ARE. But without a demonstration, a test against what is
observed, there is no basis for either accepting or rejecting
such notions.
Those are just demonstrations. They even mislead many people
as to where the power of PCT lies.
Insofar as PCT has any power, its power lies in its ability to
explain phenomena. If you don't compare what PCT says against
phenomena, it has no power at all.
And I think Bill violates his own principles (committing Crime
2 or Crime 1) by his dogmatic rejections of group statistics.
If maxim 1 of PCT is that you can't do anything about
information you can't get, maxim 2 must be that you can use
information you can get. It IS valuable to the educator to
have 70% of the children rather than 30% understand say
Pythagoras' Theorem. A change in teaching method that achieves
this is a good thing.
It's a good thing for the teacher at evaluation time, and for 7
out of 10 of the students, although you can't say in advance
WHICH 7. Insurance companies, casinos, doctors, teachers, and
generals benefit greatly from knowing the odds. But knowledge of
the odds tells us nothing about why 30 per cent of the students
didn't learn the theorem. Population measures have their uses,
but helping to build a model of individual human behavior isn't
one of them.
Can you (have you) achieved 99% accuracy in modelling in
a situation where a 4-level hierarchy is necessary?
No.
Do you expect to?
I expect someone to. When they do, I hope to read about it in the
journal.
The proposal that articles might improve on the theory seems to
contradict the suggestion that all propositions should be
accompanied by reliable demonstrations.
I don't see the contradiction. If an article proposes a change in
the theory, and demonstrates that it is necessary in order to
account for phenomena in an experiment, surely the theory has
been improved by this.
So you don't want theoretical articles. I do.
I don't want articles that are JUST theoretical. That doesn't
mean they won't appear in the journal. It just means that I won't
be much interested in them. I'm of the school that says the way
to find out if men have more teeth than women is to open some
mouths and count.
Fundamental principles of mathematics or physics won't go away,
and need no demonstration. The demonstration they need is that
they apply to the situation at hand.
Fine, then demonstrate that they DO apply to the situation at
hand. Just claiming that they do, or ought to, isn't enough for
me. The way in which fundamental principles are applied always
depends on premises and assumptions; that is the great loophole
in all pure theory.
By the way, the only reason that we have reliable principles of
physics is that someone DID demonstrate them.
On reviews: in principle I like the electronic review idea,
but particularly in PCT diagrams are very important.
That's a touchy point. I don't see how we can avoid snail mail
until somebody finds a simple universally usable way to transmit
GOOD diagrams over the net.
As I just posted to Greg, I see no need for volunteer
reviewers. I receive a stream of articles for review,
sometimes from journals I never heard of.
I was trying to accomplish two things: first, to reduce the load
on the coordinator; second, to remove from the coordinator the
power to decide who gets reviewed and who does not. Every little
reduction in power helps. By setting up the review process as a
direct relationship between authors and reviewers, we also make
the connection more personal and the reviewers more answerable
for what they say.
--------------------------------------------------------------
Chuck Tucker (930305) --
The idea of publishing everything submitted, as we do on the net,
is appealing. But I don't think we can afford it. The CSG's
annual membership fee of $45, plus the student fee of $5, covers
the cost of printing and distributing Closed Loop four times per
year -- but it wouldn't support a much larger effort. And we
can't go on forever expecting Greg Williams to provide his superb
skills and valuable time for next to nothing.
Since we don't have the resources to publish everything
submitted, we have to choose somehow. My mechanism for choosing
is to raise the standards and get rid of the sort of fluff that
fills the conventional journals. CSGnet already publishes
everything that anyone submits, so we don't have to give that up.
After we've batted these ideas around for a while longer, I'd
like to see a group effort to establish criteria for
publishability that don't violate our sense of scientific honesty
and fairness. I think we can think of guidelines that emphasize
good science and eliminate politics. But we'll see.
Mary has had an excellent idea, which is to set very high
standards even if this means only one or two articles per issue,
or none. The remainder will be the usual fare from the net. The
title of the journal, Mary suggests, would be Closed Loop: a
journal of living control systems (a nice ambiguity). Just add
the subtitle and we're there.
------------------------------------------------------------
Best to all,
Bill P.