[From Bill Powers (930616.0900 MDT)]
Still just hitting the high spots while writing the Paper.
Hans Blom (930616) --
Why do you say that they [cells] receive REFERENCE levels? In
my opinion, cells receive perceptions. Their reference levels
are 'built in'. The analogy with people is, in this respect,
complete. I'm very curious about your reasoning on this point.
I think it will be productive to consider that some inputs to
cells are sensory inputs, and some are reference inputs. In
chemical systems the situation is quite different from what it is
in neural systems. Neural signals can be isolated so they go only
where they are supposed to go and are insulated from each other.
In chemical systems, all the signals are dumped into the same
soup pot and the pathways have to be sorted out through their
chemical origins and selective binding processes at the
destinations. In the nervous system it's fairly easy to sort out
the reference signals from the perceptual signals because they
come from geometrically different places. In biochemical control
systems we have to try to identify the roles of chemical signals
in some way other than circuit-tracing.
There is more and more work on biochemical feedback systems. It's
pretty clear that negative feedback is present at all
organizational levels from organs to detailed processes inside
cells, even involving DNA and pritein synthesis. A couple of
years ago I came across a book (_Systems analysis of enzyme
systems_, I believe) by a couple of Japanese , one a biochemist
and the other a control engineer, in which many biochemical
feedback systems were modeled. Unfortunately, the authors were
most interested in how such systems reach a final operating
condition starting with wildly out-of-equilbrium conditions, with
the model behavior being shown only until the stable state had
been reached. But enough of the stable condition was shown in
some cases to reveal a perfectly good control system, reference
signal and all. The authors actually described one of the signals
as determining the level of the controlled concentration to which
the system converged, but didn't follow up the implications.
The comparator of the best control system was an allosteric
enzyme. This enzyme was "switched" (actually, driven with high
amplification) into the active state by one chemical substance
that acted to represent the controlled variable (the gross output
of the catalyzed chemical reaction), and was driven into the
inactive state by another substance that depended on other
concentrations outside the closed loop. The net activity of the
enzyme thus depended very sensitively on the difference of
concentrations between the independent chemical substance and the
other substance whose concentration varied with the controlled
concentration (i.e., a sensory signal). The enzyme catalyzed the
reaction that converted a pool of substrate molecules into the
output concentration that was sensed and controlled: that was the
output function.
By varying the concentration of the independent substance, which
is the reference signal, one could cause the closed-loop system
to produce the same variations in the controlled concentration,
quite independently of drains on the chemical product and
variations in the substrate. The authors didn't think of trying
that with their model, but it was obvious that it would work,
from the way their model snapped into a stable state once the
initial gyrations came close enough. If they had then varied the
reference concentration, they would have seen that the controlled
concentration tracked it very nicely, with no wild variations in
the other concentrations.
It's sort of funny that they missed this; the reason is the same
one for which Ashby missed the real significance of control in
behavior. If you think of a control process as one by which a
system gradually finds its way toward a state of zero error, then
you get interested in the trajectories. But if you assume that
the systems have all been in operation for some time and have
reached their stable states, from then on the only "trajectories"
that occur result from adjustments in reference signals, with the
controlled variables always closely tracking them. This is how I
think of the hierarchy: all errors are maintained at essentially
zero all of the time, with behavior always being under active
control on the time scale appropriate to the level of control.
The same might well apply to biochemical control systems. If
these biochemical control systems work as they seem to work, the
rules connecting various concentrations become very different
from those of the basic chemical reactions involved. The
concentration of one high-energy substance can be made to depend
on the concentrations of low-energy chemical signals in a way
that ignores the obvious chemical rules and imposes new ones. You
can begin to see an organization in the biochemical systems that
is simply invisible if you concentrate too closely on the
individual chemical reactions.
I put out some feelers to see if we could recruit some
biochemists familiar with this kind of modeling, so we could
collaborate on investigating these control systems further from
the PCT point of view. But there haven't been any biochemists who
have gone far enough with PCT to see the relevance of the ideas
and to be willing to look at their work from an outsider's
unorthodox point of view. Same story as in psychology, basically.
Actually, there are lots of control systems at higher levels of
organization of the organ systems. The pituitary seems to contain
a whole bunch of chemical comparators, the outputs being error
signals going to various organs and the outputs of those organs
being sensed (as negative feedback) by the pituitary. Reference
signals enter the pituitary through the neurohypophysis, the
signals originating in the brainstem. Every organ that produces a
chemical output is shut down by excesses of its own output
(negative feedback), and every organ receives both neural and
chemical reference signals from higher systems -- i.e., signals
that tend to increase the output of the organ. The control-system
organization is obvious.
I think there's a lot more to the organization of the body than
just a lot of cells living their independent lives in the midst
of a lot of other cells. That happens, too, of course, but it's
far from the whole story.
···
-----------------------------------
Who tells the liver cells that they are responsible for other
cells or for some grand collective? I think that you are
anthropomorphizing.
Then you don't understand what I'm saying.
--------------------------------------------------------------
Oded Maler (930616b) --
What is clear to me is that there are collective perceptions
for such entities, there are some collective reference, and
there is some mechanism that connects them thru actions.
If I'm anthropomorphizing, then you are mysticizing. Where do
these mysterious entities live? Are they floating about invisibly
in the atmosphere? By what mechanism do they have any influence
on an individual?
I think that these apparent collective properties simply emerge
from the interaction of individual control systems, as measures
like entropy and pressure emerge from the interaction of
individual molecules. I think they live in a conceptual space,
which exists inside your head. You're creating an allegory or a
metaphor, not a literal model of how things work. You are reading
something into the world that actually comes from your own
imagination. I say that there is absolutely no evidence for
collective perceptions, collective reference signals, collective
error signals, or collective mechanisms for action. I go along
with Tom Bourbon and Rick Marken: what's your evidence?
--------------------------------------------------------------
Best,
Bill P.