control by consequences; controlled evolution

[From Bill Powers (941228.1030 MST)]

Bruce Abbott (941228.1150 EST) --

Selection, as I said in my post (941227.1000 EST), is done by the
program (in the IF DEAD statement) and not by the consequences.

Good, we're in agreement. Thanks for catering to my literal-mindedness.

Given that reference, the consequences determine what will happen on
the next loop of the program.

Depends on what you're referring to by "what will happen." The immediate
result is a new set of random parameters, a refueling, and a resetting
of position if necessary. What will happen as a consequence of those
changes is not exactly "determined" by the previous consequence, given
all those random changes.

This is all that I mean by "selection by consequences" in this example,
and I still hold that there is nothing semantically or logically
incorrect about it.

I agree, if you mean to be speaking loosely and metaphorically rather
than technically. Technically, "selection by consequences" does not say
the same thing as "Selection by some system on the basis of previous
consequences and reference signals defining desired consequences." If we
agree that the former is simpy shorthand for the latter, there's no
problem. Are there many EABers who would understand "selection by
consequences" in that way?

However, in the interest of clarity in communication, I'll try to
refrain from using this phrase in the future (except, of course, to
annoy Rick). (;->

Oh, I agree. Rick reacts so beautifully to being annoyed. Typical S-R

Do I perceive correctly that you and Rick now agree with me that this
simulation DOES provide an analog of the reorganization process?

It's a start. As Rick pointed out, it's a bit difficult to think of
reorganization across organisms. Your control system for detecting death
and launching a new organism isn't really inside any organism, is it? If
we think of complete control systems carried across generations in
working order by DNA and other inherited cell structures, we can make
some sense of it (although one would expect inheritance to occur
_before_ death, not after).

The weakest part of this model is the resetting of the fuel supply and
the relocation of the new organism -- this kind of outside help wouldn't
be available in a real reorganization process. A real reorganizing
process would always have to start where the previous one left off.


     A few years back I was thinking about the problem that evolution
     seems to "run" at different speeds at different times. In the
     geological record, some organisms appear to move virtually
     unchanged through the eons and then, bam!, rapid evolution occurs.
     I came up with the hypothesis that unfavorable changes in the
     environment might place an existing group of organisms under
     considerable stress (poorly corrected disturbances), and that under
     the resulting changes in the organism's internal environment, the
     physiological mechanisms of reproduction might work less
     "faithfully," producing more uncorrected transcription errors,
     crossovers, and so on.

That's the same sort of idea that got me started thinking, back in the
1980s, about mutation as an action by the organism. It's a short step
from seeing stress as an error signal to proposing a reference signal
and a controlled variable -- and an output function.

There are some problems, however, because way down there at the level of
genetic biochemistry, the controlled variables aren't going to be at the
level we recognize when we look at the behavior of whole organisms. A
period of excessive heat, for example, wouldn't have its effect through
disturbances of controlled body temperature; that's much too gross a
level. The critical effect would have to be on the concentration of some
biochemical substance that an intracellular biochemical control system
could sense. The actual path by which this variable is disturbed might
be very indirect -- resulting, for instance, from a disturbance of body
electrolyte balance and changes in diffusion rates across cell walls.

Also, if we're talking evolution, we can't be talking about control
processes confined to a single organism's lifetime. Somehow, the control
systems in question have to be passed along from one generation to
another, complete not only with their organization, but with the values
of the variables in the system. There's enough room in "junk" DNA
sequences to hold the required information, but we musn't forget that a
lot more is passed along than just DNA. A great part of the actual cell
is passed along, both in sexual and in asexual reproduction. It's not
hard to see how an entire operating biochemical control system could go
right on working through the reproduction process, being duplicated in
each offspring and with the current values of the system variables also
being reproduced.

      This would of course increase the genetic variability of the
     offspring, thus providing a wider range of options to test in the
     unfavorable environment. As seems to be the case with most of my
     "original" ideas, someone else was already thinking along the same
     lines and soon published it (I don't remember who, but the person
     is now somewhat famous for having thought of it).

Some time after I had started musing along these lines, I came across a
book called "The neck of the giraffe" which seemed to propose something
similar (although not in control-system terms). Just a minute, I'll look
it up on the Colorado library net ...

OK, it's

Hitching, Francis (1982). _The Neck of the Giraffe: where Darwin went
wrong_. New Haven: Ticknor & Fields.

I'd have to re-read the book to say just what I found in it of interest.

     It did occur to me that, if this "loosening" of control over
     reproductive fidelity were in fact adaptive, then by definition it
     would have been "selected" for and thus mechanisms would have
     evolved with the correct relationship between stress and fidelity.
     However, in my conception the relationship was a simple byproduct
     of a FAILURE of control rather than the product of a control

In a paper that's due to come out some year in "World Futures" I develop
a whole fantasy about the origins of life based very nearly on your
idea. If you ask what the first control system would have been, the
answer must be that it controlled fidelity of reproduction (I said
"accuracy"). In a primordial soup, there are three types of feedback
possible: neutral, positive, and negative. Only negative feedback from
the effects of a molecule on the substrate that gives rise to it would
tend to resist changes in the substrate that would affect fidelity of
reproduction. So very rapidly, those molecules in a negative feedback
relationship to the substrate would predominate. It's inevitable, as far
as I can see, that living molecules would quickly become control systems
controlling those aspects of the substrate that affect accuracy of
reproduction. So I ended up wondering not about mutations so much as how
reproduction could possibly be as stable as it is. The obvious answer
(to a control theorist) is that the environmental variables capable of
affecting fidelity of reproduction come under feedback control.

Somewhere down the line, we would have non-zero reference signals being
evolved, so that even if the environment changed, the local variables
that affect fidelity of reproduction would be held in non-equilibrium
states by the outputs of the developing biochemical control systems.

Failure of control would then have exactly the effect you posit. The
variables that preserve fidelity of reproduction could no longer be
maintained in the required states by the outputs of the control systems,
and they would begin to change. As soon as they begin to change,
reproduction begins to vary again -- mutation! This offers a very neat
explanation for "punctuated equilibrium."

And of course, mutation stops only when a new control organization comes
into being (if it does -- it may not) that can once again stabilize the
environmental variables on which fidelity of reproduction depends.
That's the basic principle of reorganization, present from the start!

I think this makes a very pretty tale, especially as it extends PCT to
the very origins of life.

     If such a mechanism exists (yours or mine), then the stability of
     creatures like the clam and the cockroach would imply either that
     these creatures' control systems are superbly adapted to cope with
     a wide variety of environmental disturbances, or that they function
     in an environment that is especially stable. Either way, the
     "variation" mechanism would not be brought into play.

Right on. It's probably a combination of effects: good control systems,
but good enough only for a particular environment (called a "niche").
Change the niche too much, and control will start to fail, leading to a
burst of mutations and either extinction or better control systems.

I haven't thought about this for a while, and find that it still makes
sense. Maybe there's something to it.

Bill P.