Derivatives and Clarifications

<Martin Taylor 940401 18:00>

I've not been posting to CSG-L for some time, for two reasons. First is
the time taken by the ongoing sleep study, which leaves little time for
anything else. But now we will have two weeks rather than one to deal with
the run about to complete and to prepare for the next one, so things are
slightly relaxed. Enough, anyway for a few comments on items from the
last three or four weeks. The second reason is that I have been spending
a lot of time communicating and consulting with Bill P., both on the
tracking studies in the sleep deprivation experiment and on the matter of
information theory in the analysis and prediction of the behaviour of control
systems. We have not come to a common position, but I think we both
understand things a little differently from the way we did a month ago.
So I will not comment further on that yet.

Rick Marken (940310.2130)

But I am worried by the
request that Martin clarify his prediction if we misunderstood it. I can't
imagine a clearer claim (or prediction) than the one that Martin made
above ("But the derivatives [of perceptual signal and disturbance] are
not independent"). When we showed a plot indicating that these
derivatives ARE independent, Martin said that this result was
EXPECTED. So I can only presume that Martin's "clarification" of
the prediction of non-independece of the derivatives is that they ARE
independent. Is that clear?

Funny perception! I quote the data that you say shows the derivatives
to be independent:

        >Bill Powers (940302.1300 MST)
        >
        >The correlation of the derivative of the disturbance with the
        >derivative of the delayed perceptual signal is 0.268.

That does not look to me to agree with your statement. The result that
was expected was that the correlation is positive and that it is not large.
The data look to me to be as expected. Still. The derivatives are
certainly NOT independent. Is that clear?

I am afraid that this "clarification" process could go on forever; Martin
makes a prediction, we ask if that's what he really means, he says "yes"
(usually in a private post, as he did to me when I asked if he really
meant to say that the derivatives are not independent), we collect the
data, it doesn't fit the agreed on prediction, we think we have falsified
the prediction but then Martin wants to "clarify" what he really meant
to say.

There are several reasons for clarification, but as you can see, the above
is not one of them. I may be wrong, but I work with the presumption that
we are all trying to discover scientific truth rather than to achieve
political dominance. This being so, it is essential that the people
concerned have the same idea about what is being said.

There are many reasons why they may not, foremost among these being
differences in the technical background of the participants. But there are
other reasons, including that the writer misstated the concepts intended
(as the writer determines on re-reading what was written) or that the
readers misread and responded to what looks to the writer as something that
was not said or intended.

One thing that leads to clarification NOT of what was said, but of the
conceptions behind them, is a failure of prediction when what is predicted
is agreed among the participants. That kind of failure is productive. It
leads to improvement in understanding the way the world works. The other
clarifications affect only the participants' understanding of each other.
But they are necessary. And judging from experience, they very often need
to be tried in many forms, because some particpants find it very difficult
to modify their perceptions of what the others are trying to say.

Given time, I hope that clarifications of all kinds will continue to be
useful. It is the truth of the way the world works that eventually
determines whether they have succeeded in their intent--that we all
advance our understanding of the world.

I'll make one prediction (that won't need clarification) -- I predict that
every time you get a result that is inconsistent with one of Martin's
predictions you will find Martin saying that you had not been "clear"
about what he really meant.

That, I think, reflects simple meanness of spirit. If a prediction I make
is inconsistent with results, I hope that it will involve a clarification
of my concepts, not of my statements about pre-existing concepts.

ยทยทยท

=========================

On self-organized structures and "e-coli" model for reorganization
(this comment triggered by the dialogue between Hans Blom and Bill Powers
in mid March)

E-coli has long served as a model for the effectiveness of random reorganizing
processes. In view of an earlier discussion on the inutility of the
concept of "self-organization" in a control hierarchy, I thought it might
be worthwhile to draw your attention to a pair of articles about self-
organization brought about by e-coli. I mean the actual bacterium that
tumbles about, not some computer model of it.

In "Physics Today" Feb 1994 (p 21 of a special issue on physics and biology)
there is a beautiful picture of a self-organized structure of aggregates
of e-coli that moved "randomly" in chemical gradients based on substances
they themselves excreted. The structure is very reminiscent of the spirals
of spots you might see if you looked end-on at a pine cone, or at the
yellow centre of a sunflower. Dots composed of clumps of e-coli are
very regularly spaced, despite the movements that led to the spots having
been random in the sense in which this movement is usually discussed on
CSG-L when we talk of reorganization.

The symmetry of the structures built by e-coli presumably depend on an
equal symmetry or uniformity of the substrate on which they move. But
the stability of the organization is remarkable. It seems to me that
in a large hierarchy of control systems acting on an outer world,
reorganization is likely to build stable, mutually influenced, structures
of equivalent complexity and (where the world permits) symmetry. The
mechanism, though, is likely to be different. If I am right, reorganization
in a control hierarchy is driven by the Bomb more than anything else,
so that the structure as a whole becomes critically stable rather than
showing the strong stability of the e-coli structure in a symmetrical stable
chemical world.

In the article from which the Physics Today picture is drawn, several
different kinds of structure are shown. You may want to look at them:
E.O.Budrene and H.C.Berg, "Complex patterns formed by motile celss of
Escerichia coli" Nature, 1991, 349, 630-633.

=====================
Have to go back to the sleep study now. Hope to get to more of the
backlog later in the next two weeks. I have 117 messages saved, from
different source, all of which either demand a response or some action.
Not all are going to get it.

Martin