Jim's musings on "summarizing" touch on a theme that Murray Gell-Mann
raised at the Complexity and Strategy conference. As I'm intently
working out many of the ideas from that conference, I'll make a first
attempt at this one here.
Jim says,
> But - and this seems a paradox - the one kind of transformation that
> resists all my efforts at formalization is the operation of "simplifying"
> or "summarizing" such a set. Which facts will I drop from my "summary"?
> That depends on who I'm summarizing for, why I'm summarizing at all, what
> other information is available or already known to me and to my intended
> audience.
>
> In fact this operation of "summarizing" a body of information cannot be
> defined solely in terms of the information being summarized.
My idea is that we can either redefine "summarizing" (probably not a
good idea) or come up with a different term (preferably rooted in a
meaning consistent with what follows).
I interpret Murray to be saying that the basic idea is that
"complexity", in the domain of information , refers to the ability to
compress information so that the shortest expression contains the
greatest amount of information AND that information, when played
"against" new and varied situations produces similar results. That
is, when it "meets the world" it decompresses - almost automatically.
And in the decompression produces useful, more or less repeatable,
results.
And, as Jim says, "No man is an island -- and neither is any single
fact." Murray says, both the compression and the "decompression" are
context dependent.
The challenge of a good summary is that, when played back against
both the various memories of the original (and far richer event) and
against similar events or intended application circumstances, it
produces results similar or usefully related to the original or its
intentions. Maybe we should call this "a compression".
And then I discovered what followed. Wonderful! It enriches the
above.
Which picks up an idea from Francisco Varela, also detailed at the
same conference. (It was a wonderful, stimulating experience - if
you haven't gathered I think so by now.) Francisco makes the point
as strongly as he can that we are trapped inside a mechanistic view
of information that is far from the case for living beings.
Information doesn't "travel" and it isn't "sitting in front of us" in
any way. Our senses receive a perturbation (a "poke") and the
ingoing stimuli meet some outgoing (generated from purely internal
sources) and, if there is a match or "resonance", then information is
created. ("in" "form" "ation = formed from within).
The relevance of feedback loops may be in mortal danger in most
operations of human sensing and intelligence.
> the information flow is altered; seems like there's a lot less need for all
> those elaborate presentations that we used to have...
According to this view, information flowing will be more like a
"standing wave" than in single thing flowing along a tube. If we
continue the idea of "resonance", then we might say that the
information at each stage is connected to the information at each
other stage - and the connections has as least as much to do with the
context of the information-creator at that stage than any initial
conditions.
This leads to a lot of work and a lot of new questions. But also to
the possility of both useful "compressing" if not summarizing and of
principles which can be developed for the process that we are
denoting by the term "information flows".
-- Michael McMaster Michael@kbddean.demon.co.uk