Selection, Adaptation & Control

[from Gary Cziko 941005.1944 GMT]

Preface:

Several CSGnetters know that I have been working on a book over the last
few years entitled _Without MIracles: Universal Selection Theory and the
Second Darwinian Revolution_. I think Popper would have liked it (too bad
he won't see it now) and it owes a lot to Donald T. Campbell's
"evolutionary epistemolgy" and "general selection theory."

The book is now being reviewed for publication as a Bradford Book by MIT
Press (but would be heavily marketed as a trade book if accepted--my big
chance for fame and fortune). It looked like it was going to be accepted
for publication in early summer, but a late negative review caused the
editorial board to defer its decision (ouch) and order a whole new round of
reviews (ouch again). The score is now 6 to 1 in favor of publication with
the next editorial board meeting set for October 26 (wish me luck!).

So here's your big chance to make an impact on the book and even be
acknowledged for doing so. I've got some questions which need some
answering. I know that at least Martin Taylor, Bill Powers, Greg Williams
and Rick Marken are interested in evolution so I hope to get some reaction
from them. Perhaps there are others who can also help out--I particularly
need someone who knows something about physiological adaptation, the kind
that leads to bigger muscles in weighlifters and higher-capacity hearts and
lungs in endurance sport athletes and high-altitude dwellers like Bill and
Mary Powers.

Questions:

1. Does it make any sense to think of the normal functioning of a properly
working control system as a type of blind-variation-and-selection process?
In private communication, Greg Williams said:

In a sense, control systems
actually are dynamical variation-and-selection machines. Now, that's an
interesting wrinkle I hadn't thought about before.

I suppose the thinking here is that the output function of a control system
makes no pre-calculation of how much output to put out. It just gives
whatever its amplication and energy source allows. It only "hits the
brakes" when it gets feedback. Without the feedback, the output function
is totally stupid, totally blind. The stability and success of the control
system is thus due to after-the-fact selection, not a priori calculation.
In spite of my "radical selectionism," I hesistate characterizing control
systems in this way, although reorganization is certainly seen as a
variation-and-selection process in PCT. Comments? Reactions?

2. How does physiological adaptation work? Is all or some of it also
based on a type of variation and selection process, or do muscles just
immediately "know" to get stronger after they have been exposed to stress?
I read somewhere that the liver is made up of many different types of
cells, each one best a detoxifying certain chemicals (like alcohol, organic
solvents, etc.). Those cells that are used most are thereby selected and
increase in numbers, while those that don't are eliminated. So this looks
like a selection process (the immune system is now a classic example). But
what about muscles, hearts, lungs, number of red blood cells, bone density,
etc. I realize that all these types of physiological adaptation can
perhaps best understood as the consequences of control (e.g., increasing
the number of red blood cells after moving to a high altitude controls for
the body's ability to deliver fixed amount of oxygen to the cells), but
does the adaptation process itself involve a type of variation and
selection of the type we see in PCT re-organization and liver adaptation?

Adaptively yours, Gary

P.S. If anyone knows of Internet-based discussion groups or news groups
where I could ask my physiologically adaptation question, please send me
the address or newsgroup name.

Gary,

Contact Tom Smith--address in Continuing the Conversation
circa 1990. He and K. U. wrote book-length ms. on
cybernetic approach to evolution. Tom is Ph.D.
physiologist.

Dennis Delprato

[From Rick Marken (941006.0740)]

Gary Cziko (941005.1944 GMT) --

Questions:

1. Does it make any sense to think of the normal functioning of a properly
working control system as a type of blind-variation-and-selection process?

Well, it depends on what one thinks of as sense. Does it make any sense
to think that Republicans care about anyone but themselves and other rich
people? Does it make sense to think that Clinton is not doing well as
president when he has done (and has tried to do, with little success thanks
to uncompromising, mendacious republicans and reactionary democrats)) more
good than any president since FDR? Does it make sense to think of the
NORMAL functioning of a control system as a type of blind-variation- and
selective retention? The answer to all these questions (to a sensible guy
like me) is NO. But, then, when it comes to verbalization, one man's sense
can be another man's nonsense. For example, Greg Williams answers the
question this way:

In a sense, control systems actually are dynamical variation-and-selection
machines.

Actually, I agree with this becuase Greg left out the "blind" part. I
interpret "blind variation" to mean "random" variation. In most control
systems, the variation in the output is neither random (it is a determin-
istic function of the error signal) nor is it blind (since it is in a
loop, it is a continuous function of the results it is producing).

I suppose the thinking here is that the output function of a control system
makes no pre-calculation of how much output to put out.

But it does. o = f(e) so the amount of output (o) given a certain error (e)
is "precalculated" by f(). I suppose that there is less "pre-calculation"
if f() is changing randomly over time, as it does (slowly) in the lowest
level control loops in living control systems. But even then, the very
existance of a particular f() for even a brief instant of time means that
the output that results from whatever error value is present at the
input to that function at a particular instant is determined by the f()
that exists at that instant.

There is one kind of control that does qualify as "blind variation and
selective retention". This is the unusual (but powerful) form of control we
call E. coli. In E. coli, there is no f() relating e to o; error is not re-
lated to the particular o that occurs; it is related to WHEN the o occurs
(where o is the direction in which E. coli moves). e determines the time
between (random) changes in o (call that time delta). So in E. coli,
delta=f(e).

Once f() determines that it is time for a change in output (a "tumble")
the new output is selected randomly from all possible outputs (from
the entire 360 degree range of directions in 2 space, or the 360^2
degree range in three space). The output is random so the effect of this
output on the variable E. coli is controlling (change in sensed concentration
of an attractant -- this variable is the "basis of selection) is also random.

If the result of the random change in output is an increase in the
difference between controlled variable and reference signal (the selection
criterion), then error increases, delta is shortened and there will be
another tumble soon;If the result of the random change is a decrease in
the difference between controlled variable and reference signal, the
error decreases, delta increases and a change in output is delayed (because
the process is delayed). In E. coli there is blind (random) variation in
output and selective retention (via increased delta) of those outputs
that move the controlled variable toward the reference signal.

I would describe ordinary control as systematic (functional) variation
to produce selective "retention" of a controlled variable at a reference
level (ie. control). I would say that E. coli control is blind (random)
variation that produces selective "retention" (control). The fact that E. coli
selection (control) works so efficiently was one surprising result of Bill
and my "selection of consequences" studies.

Best

Rick

[Martin Taylor 941007 19:30]

Gary Cziko 941005.1944

1. Does it make any sense to think of the normal functioning of a properly
working control system as a type of blind-variation-and-selection process?
In private communication, Greg Williams said:

In a sense, control systems
actually are dynamical variation-and-selection machines. Now, that's an
interesting wrinkle I hadn't thought about before.

I suppose the thinking here is that the output function of a control system
makes no pre-calculation of how much output to put out. It just gives
whatever its amplication and energy source allows. It only "hits the
brakes" when it gets feedback. Without the feedback, the output function
is totally stupid, totally blind. The stability and success of the control
system is thus due to after-the-fact selection, not a priori calculation.
In spite of my "radical selectionism," I hesistate characterizing control
systems in this way, although reorganization is certainly seen as a
variation-and-selection process in PCT. Comments? Reactions?

I would not characterize normal control as "variation and selection." A
"variation and selection" controller would, to me, act like the e-coli
system. It would move randomly around its space, faster and/or with bigger
moves if the error was large than when the error was small. Such a system
would have no sense of direction, as a normal control system does--always
moving the perception in such a way as to reduce the error. But it would
be more robust than a normal control system if the sign of the effect of
output on perception through the environment kept varying.

Variation and selection is effective if you don't know which way you need
to go. If you do, you go that way, and that's what a normal control system
does. That's more effective.

2. How does physiological adaptation work?

There may be people here who know. Unfortunately, I won't be able to ask
them until at least next Thursday, by which time I'll probably have forgotten
you asked the question. Send me private e-mail on Thursday if you want me
to find out if any of our physiologists can tell me.

Martin

Tom Bourbon [941006.1050]

[from Gary Cziko 941005.1944 GMT]

Questions:

1. Does it make any sense to think of the normal functioning of a properly
working control system as a type of blind-variation-and-selection process?

In one sense, I don't think so, for reasons summed up nicely in
     From Rick Marken (941006.0740)]:

"I would describe ordinary control as systematic (functional) variation
to produce selective "retention" of a controlled variable at a reference
level (ie. control). I would say that E. coli control is blind (random)
variation that produces selective "retention" (control)."

A control system makes a very precise calculation of how much output to put
out; whether that is the amount actually put out (muscle fibers fatigued,
O2 levels low, presynaptic motor neurons slow on the uptake of previously
released transmitters, waste products accumulating, etc), or whether the
amount put out has the appropriate effect (when it combines with disturbances
in the environment, for example), are other matters. o = f(e), always. The
issue is whether f remains constant or varies.

Rick talked about the "E. coli" process in control. In the example he used,
there is no f relating e to o, but there _are_ examples in PCT modeling (not
necessarily in print, yet) where we have used the E. coli procedure to
modify f, thereby adaptively changing the way e is related to o. Those
are instances where an adaptor loop senses and has a reference signal for
its own perceptions of the error signal in the primary controlling loop;
when the time integral of sensed error exceeds the reference level, the
adaptor loop puts out a random small change to the output function, f, of
the primary control loop.

In _that_ sense, my answer to your question # 1 would be, "I think so." A
control system equipped with a secondary adaptor loop demonstrates a
"blind-variation-and-selection process" in its setting of f. Whether I say
yes or no to your question all depends on the level of detail we use in our
model for ". . . the normal functioning of a properly working control
system. . .."

2. How does physiological adaptation work? Is all or some of it also
based on a type of variation and selection process, or do muscles just
immediately "know" to get stronger after they have been exposed to stress?

I remember Bill Williams, who now lives just down Black Lick Road from Greg
Williams (no relation), telling me about some interesting work on changes in
the density of bone (as in the skeleton of a live human) in which the
author(s) characterized the changes as adaptations to changes in stress
applied to the bone. (If so, this would be one of the few lines of work
with humans in which the word "stress" has something like a specific
meaning. :slight_smile: ) I don't recall the sources, but maybe Greg can run down the
road and ask Bill.

As to whether adaptation in general might be an E.coli-like process, I'm not
sure if there is research that has explicitly tested that possibility. (The
bone studies suggest that someone _might_ have done this kind of work.)

I can imagine some adaptation being the result of an E.coli-like process,
and other adaptation being garden variety control. In the latter case, for
example, a person as lifter of weights uses muscle to achieve the
newly-selected intended results of perceiving weights lifted. The muscle
cells that serve as output devices on the bottom level of the person-as-
perceptual-control-hierarchy have never before in their lives put out so much
output. They are subjected to new mechanical and physiological disturbances
that they might oppose, as cell-level control systems operating as
simple o = f(e) devices. There might or might not be E.coli-like adaption
at work here.

I'll try to locate some specific references if I can.

Later,

Tom