Efferent retina, description and explanation

CSGnet Archive CSG1996
1

[From Rick Marken (961023.1320)]

Bill Powers (961023.0745 MDT)

I think those efferents in the optic nerve must be doing something to the
retinal perceptual systems. What, I can't imagine.

Bill Benzon --

But it probably has something to do with controlling perception.

Not necessarily. I think it is far more likely that the efferents in the
optic nerve are components of "image enhancement" operations (like lateral
inhibition). They might not even be components of control loops; for example,
the efferents may affect the firing rate of afferent neurons that have no
have no effect on the firing rate of these same efferents. I believe that
efferents worked this way in F. Ratliff's model of lateral inhibition in
the retina.

Bruce Gregory (961023.1245 EDT) --

I don't find it [Martin's approach] fundamentally different from what you
are saying. The attractor notion _describes_ how systems behave, but it
does not _explain_ why they behave that way. That explanation is provided
by the dynamical equations (and the constraints).

Have you seen Bill Powers' "Gatherings" program in action? I think it's
available on Dag's PCT Demos disk. Anyway, in one scenario (called "guru" I
think) a bunch of dots (representing individuals) converge toward and
eventually encircle a stationary dot -- the "guru". I am thinking that a plot
of the dynamical pattern of movements of these dots resembles an "attractor".
I presume that are a set of dynamical equations -- the attractor "theory" or
equations -- which (with the appropriate parameter settings) can generate an
approximation to the dynamical behavior of the set of dots. These equations
do not (to my mind) provide an explanation of the behavior of the dots; they
simply describe the behavior of the dots.

I believe that these dynamical equations _do_ provide an explanation for
behavior in certain circumstances: in particular, I believe that the
"attractor" equations are an explanation of the behavior of physical
phenomena like, say the behavior of marbles converging toward and ending up
in a circle at the bottom of a bowl. In this case, the parameters and
variables of the attractor equations would correspond to the functional
relationships assumed to exist between actual and theoretical entities in
this situation. Some parameters in the model would correspond to the assumed
downward acceleration of gravity; others would correspond to the actual
initial positions and masses of the "dots", etc. In this case, the dynamical
"attractor" equations are simply a physical model of a particular situation:
one where a bunch of marbles all fall into a bowl at the same time.

To say that these dynamical attractor equations _explain_ the behavior of
the dots in the gatherings program (which I beleive _is_ what Martin is
saying) is a huge mistake. It implies that the behavior of the dots in the
"Gatherings" program - - social behavior -- results from the same kind of
dynamical interaction of forces and masses that causes the dynamical behavior
of the marbles. It is a way of sneaking (by methaphor) the cause-effect model
of physics in as an explanation of what is likely to be a phenomenon based on
control processes.

Even if one is careful to remember that the description of the social
behavior using the attractor equations is just metaphorical, it is hard to
see what the value of such a metaphor might be. At best, it is just a
colorful and fun way of describing an observation; the dots in the
"Gatherings" program _do_ seem to be pulled toward an attractor -- the
guru. But at worst, it is misleading and directs attention away from the
basic explanation of all social behavior: control of perception.

I think Kent McClellend's approach to the study of social behavior shows
how much can be done by looking at this phenomenon in terms of PCT. Kent's
approach is based on observing how control systems _actually_ interact.
Rather than looking for metaphors for social behavior in the trendy science
literature, Kent actually sat down and started studying what happened when
control systems controlled the same or similar environmental variables. What
Kent has found (so far) is that different kinds of observed social
interactions, from cooperation to conflict, result when control systems
simply "do there thing" in a common environment.

Kent's is purely theoretical work (so far) -- just like Martin's. The
difference, to me, is that Kent is contributing worthwhile and new discoveries
about how control theory might actually relate to social behavior; and he's
doing it by studying how control systems actually interact. Martin is
contributing new discoveries about superficial similarities between control
theory and other theories that have nothing to do with control; I'm just not
interested.

Best

Rick

2

[Martin Taylor 961024 16:20]

Rick Marken (961023.1320)

Interesting to read the following, just after posting my response to Bill
Benzon, on Fourier transforms and phases. I think you are probably right,
(because, as so often happens, you wrote what I think--it puzzles me why
when I say so, you are so keen to say that I don't think what you write).

ยทยทยท

I think it is far more likely that the efferents in the
optic nerve are components of "image enhancement" operations (like lateral
inhibition). They might not even be components of control loops; for example,
the efferents may affect the firing rate of afferent neurons that have no
have no effect on the firing rate of these same efferents. I believe that
efferents worked this way in F. Ratliff's model of lateral inhibition in
the retina.

----------------------------

However, I don't "think what you write" in connection with the following
in the same post.

Have you seen Bill Powers' "Gatherings" program in action? I think it's
available on Dag's PCT Demos disk. Anyway, in one scenario (called "guru" I
think) a bunch of dots (representing individuals) converge toward and
eventually encircle a stationary dot -- the "guru". I am thinking that a plot
of the dynamical pattern of movements of these dots resembles an "attractor".

What would be the attractor? I'm afraid I don't see one. The orbits of the
different "followers" do wind up near the guru, for sure. But they don't
wind up at the same place, or following the same track, and they wouldn't
do so even if there were only one follower and the run was started over and
over with the guru tracing the same track and the follower started from
all possible positions in the field. So I'm afraid I can't immediately see
the attractor that you see. Perhaps in some higher dimensional space there
is one, but I can't see one in the 2-D space of the display, and even in
a 1-D space of "distance-to-guru" the plot of orbits versus time would
criss-cross (a sure sign that one is not looking at a properly dimensioned
dynamical system). Could you enlighten me, oh guru?

But, sorry, I forgot:

Martin is
contributing new discoveries about superficial similarities between control
theory and other theories that have nothing to do with control; I'm just not
interested.

Fine. If you aren't interested, just don't interfere in the discussions among
people who are or might be interested.

But at worst, it is misleading and directs attention away from the
basic explanation of all social behavior: control of perception.

You are starting with observations, and asking what is the explanation.
That's fine. I'm asking what one might expect to see if the explanation is
correct.

I'm interested in the _consequences_ of control, on the assumption that
"control" is the foundation of any explanation. I'm interested in the
_quality_ of control; I'm interested in the _learning_ of control, all
of which involve consideration of the environment in which control occurs,
and of the changing structures of control hierarchies. You, as you have
often stated, are interested only in the _fact_ of control, something I
rather tend to take for granted.

Even if one is careful to remember that the description of the social
behavior using the attractor equations is just metaphorical,

Oh, guru, it is the image of the marble in the bowl that is metaphorical,
not the description of the social system as a dynamical system. That's
abstractional, to coin a term.

The description of social behaviour, using the assumption that there are
interacting control systems that learn, is based on recognizing that
the system is a member of a class of systems whose behaviour has been
studied. So long as the assumptions cover the important aspects of that
behaviour (a limitation brought up by Bill P, with which I agree), then
any system that is a member of the class will have that behaviour. The
question is only whether the social system of interacting control systems
can be treated as a member of any specific class of systems that has been
studied. By analyzing the relevant aspects of the mechanism of interaction,
it seems that it can.

So, until proved otherwise, we assume that this field is a member of the
class of rectangles that measurement seems to show it to be, and we assume
that we know the length of its diagonals by applying to the lengths of
two sides the theorem of the mystic (really, he was) Pythagoras.

Incidentally, there's no conflict whatever between Kent's approach and mine.
Mine is based only on the idea that reorganization more often results in
reduction of conflict than in increase in conflict, overall. Kent's says
that conflict can happen or not, in an interaction.

Martin