They be gods

CSGnet Archive CSG1995
1

From Tom Bourbon [950831.0041]

Earlier (Tom Bourbon [950830.1311]), I described a _science News_ report about
an article in _Physical Review Letters_. Reportedly, in the article, several
physicists wrote about the appearance of "coordination" and "cooperation" in the
movements of large numbers of "self-driven" particles. Reportedly, they said
their results would generalize to movements of living things, in groups.

From the Science News report, I could not discern which ideas originated with

the physicists, and which with the science writer. Now I know. Everyone who
thinks the physicists modeled their particles as producers of their own
movements, go sit in the corner. :slight_smile: Everything the simulated particles did was
determined by the god in the computer, but you wouldn't know that by reading
many parts of the article, especially those where the authors told how their
results could be applied to living things. That is unfortunate, because the
writer for Science News treated the original article as one that gave insights
into the collective behavior of living things.

···

================================================

T. Vicsek, A. Czirok, E. Ben-Jacob, I. Cohen & O. Shocher (1995). Novel type of
phase transition in a system of self-driven particles. Physical Review Letters,
vol. 75, no. 6, pp. 1226-1229.

Selections from the "ABSTRACT":

"A simple model with a novel type of dynamics is introduced in order to
investigate the emergence of self-directed motion in systems of particles with
biologically motivated interaction." (TB: I translate this sentence as saying
each of the particles direct its own motion and they all interact due to their
individual biological motivations. If all is as they say here, they might be
modeling living control systems)

"In our model particles are driven with a constant absolute velocity and at each
time step assume the average direction of motion of the particles in their
neighborhood with some random perturbation (eta) added." (TB: I translate this
second sentence as contradicting the one immediately before it. They seem to
say that all of the particles move, all of the time, at the same velocity, which
is assigned arbitrarily by the physicists, and that the directions of movement
for all of the particles are changed in accordance with calculations done,
moment by moment, by the physicists, or their surrogate -- the program running
in the computer. If my two translations are correct, the second sentence
suggests that the physicists played god with their particles, which they moved
about on the screen like inanimate objects. The article confirms the latter
interpretation.)

It is clear in the article that the physicists wanted to model some of the
_outward appearances_ of "coordination" and "cooperation," and that they
succeeded. It is equally clear that, in no way did their model represent any or
all of their particles as living systems. In fact, they modeled them as being
like metal filings, influenced by a magnetic field: "In this sense our model is
a(n) . . . analog of the ferromagnetic type of models . . .. The elementary
event is the motion of a particle between two time steps. Thus the analogy can
be formulated as follows: The rule corresponding to the ferromagnetic
interaction tending to align the spins in the same direction . . .is replaced by
the rule of aligning the direction of motion of particles in our model of
cooperative motion. The level of random perturbations we apply are in analogy
with the temperature." (p. 1226). William James would have loved it! In an
elegant early passage in his _Principles of Psychology_, James used the case of
filings drawn to a magnet as an example of an interaction that is at the
_opposite extreme_ from interactions between intelligent, purposeful living
things.

The physicists say, all in italics, "The only rule of the model is at each time
step a given particle driven with a constant absolute velocity assumes the
average direction of motion of the particles in its neighborhood or radius r
with some random perturbation added." (TB: In other words, each particle
changes its direction under direct control of the program, which plays the role
of a god-like perceiver of all particles within radius r. There is _no model_
of each particle as a system that perceives the average direction, or controls
its own direction relative to such a perceived average. This is a model of iron
filings, not of living things.)

The authors speak of how, under certain conditions, "all of the particles tend
to move in the same spontaneously selected direction" (p. 1227), but no single
particle ever selects any direction, nor do the particles in the aggregate
select anything. Every initial direction of each particle, and every change in
direction of any single particle, is imposed from outside.

When the authors talk of how their model and its results apply to living things,
all of their examples of animal behavior are descriptions of external
appearances. The only way their model applies to the behavior of living things
is as an arbitrary way of duplicating some of the outward appearances of
movements by large numbers of actors. The bottom line is that the particles in
their model are like iron filings, not like control systems, either living or
artificial. But the word is out, in Science News, that physicists can explain
cooperative, coordinated movements in groups of bacteria and social animals and
maybe even humans -- watch for it being cited in other places.

Later,

Tom

2

[From Bruce Abbott (950831.1000 EST)]

Tom Bourbon [950831.0041] --

Welcome back, Tom.

From the Science News report, I could not discern which ideas originated with
the physicists, and which with the science writer. Now I know. Everyone who
thinks the physicists modeled their particles as producers of their own
movements, go sit in the corner. :slight_smile: Everything the simulated particles did was
determined by the god in the computer, but you wouldn't know that by reading
many parts of the article, especially those where the authors told how their
results could be applied to living things. That is unfortunate, because the
writer for Science News treated the original article as one that gave insights
into the collective behavior of living things.

Not too long ago on CSG-L there was a brief mention that Richard Feynman,
that noted physicist and bongo-drum player, had declaired psychology to be a
"cargo cult science" that emulated the trappings of science without
duplicating the substance. This is, I'll wager, a common view among
physicists (not to mention Congressmen) of psychology as a field.
Physicists are used to dealing with phenomena that, although perhaps
complex, are fundamentally the products of simple processes which usually
can be isolated and studied under conditions which yield precise
relationships capable of clear mathematical description. The physicist can
be confident that every electron, every photon, every mole of hydrogen, will
behave the same, within experimental error, when subjected to the same
experimental conditions. I would argue that physics has been so successfull
precisely because its observational phenomena are so regular.

It's nice to see that physicists, when attempting to apply their findings to
living organisms, are just capable as psychologists of doing cargo cult
science. It was easy for Feynman to criticize: he never did any
psychological research himself. That's too bad: I would have loved to see
the result.

When the authors talk of how their model and its results apply to living

things,

all of their examples of animal behavior are descriptions of external
appearances. The only way their model applies to the behavior of living things
is as an arbitrary way of duplicating some of the outward appearances of
movements by large numbers of actors. The bottom line is that the particles in
their model are like iron filings, not like control systems, either living or
artificial. But the word is out, in Science News, that physicists can explain
cooperative, coordinated movements in groups of bacteria and social animals and
maybe even humans -- watch for it being cited in other places.

Right on the money, Tom. But for me this example does bring up a question
concerning models. If, as the authors say, their model is based on analogy
to the spin-alignment behavior of ferromagnetic atoms in a magnetic field,
then the analogy is to a balance of local forces acting on the "particles"
(e.g., fish). It is, as you say, a model of the observable behavior of the
particles and not of the particles themselves. It is as if a given particle
"looks around" and "sees" what the other particles around it are doing, and
then alters its direction of movement to be aligned with the average
direction of movement it perceives in those other particles.

The question I have is, is this model incompatible with a control model? It
seems to me that a collection of autonomous control-system particles,
controlling, say, their own perceived distances from their neighbors (and
perhaps the rate of change of those distances), might be described as
"looking around, seeing what their neighbors are doing, and altering their
movements" in the way envisioned by the authors of this article. The
control model is, of course, more fundamental (i.e., generative); the
apparent "rules" of the behavior of the particles emerge as a consequence of
the organization of each control system and the interaction of those
systems. What our physicist friends have done (perhaps) is to supply a kind
of Boyle's Law: a useful description, at a rather superficial level, of
certain relationships, which beg to be explained by a causal model.

This line of thinking leads me to question whether it is necessary to assume
that either the physicists or their computer are acting in some sense as the
god of the particles, any more than it is necessary to do so when describing
how the multiple independent control systems of the crowd demo get updated
by the program at each time-step. The "particle" description it not
necessarily invalid; it just operates at a different level of description.
The only problem with it I perceive is that its authors apparently do not
realize that the "law" of particle behavior they have descovered is a
product (at least in the case of living organisms) of a more fundamental
process (control). Had they modeled the particles rather than their
behavior, they might have been able to show under what conditions their
descriptive law of particle behavior holds _and_ under what conditions it
doesn't.

From another perspective, what these authors have done is interesting. _If_

living control systems control certain perceptions (distance from neighbors,
etc.) and _if_ they are packed densely enough, they may behave _as if_ each
control system is attempting to move in a direction that is the average of
its neighbors' directions; under these conditions living control systems,
matter aligning in a magnetic field, and perhaps many other physical
systems, may exhibit behavior consistent with the predictions of the same
model. Furthermore, this model suggests that one parameter, random "noise,"
has an important influence on the observed regularity of the behavior of the
particles as a group. It sounds to me that the role of "noise" in this
model is to perturb the directions of movement of the particles. In living
control systems, that role might be played by the actions of additional
control systems whose actions alter the direction of the "particle." For
example, fish on the periphery of a school may change direction in order to
investigate a patch for food or to avoid an approaching shark.

Regards,

Bruce

3

[Peter J. Burke 8:00am PST]
It is always amazing to me that physists will blithly step out of the
areas of their own expertise to write such gibberish. I would take one
possible exception with Tom's interpretation that the simulation had
a god doing the calculations for each of the particles. One could
interpret the calculation as being made within the particle, averaging
the movement in its "perceptual domain". Still, it is not a control
system, it is an SR system.
Peter

···

-------------------------------------------------------------------
Peter J. Burke Phone: 509/332-0824
Sociology Fax: 509/335-6419
Washington State University
Pullman, WA 99164-4020 E-mail: burkep@unicorn.it.wsu.edu
-------------------------------------------------------------------

4

From Tom Bourbon [950831.1858]

When I perceived an afternoon-long series of very intense thunderstorms,
the error signals that ensued, relative to my reference signal for
preserving my computer, resulted in an unplugged afternoon.

There were some interesting replies to "They be gods" (Tom Bourbon
[950831.0041]). I had described some work by physicists who said that the
movements of organisms, whose movements result in ordered patterns, can be
modeled as though the organisms' movements are governed by a rule that the
physicists use in a model for the movements of iron filings in a magnetic
field. In the replies to my post, I was especially interested in some
suggestions that the physicists' model could be, or might be, interpreted
as a model of perceptual control.

    [Peter J. Burke 8:00am PST]
    I would take one possible exception with Tom's interpretation that the
    simulation had a god doing the calculations for each of the particles.
    One could interpret the calculation as being made within the particle,
    averaging the movement in its "perceptual domain".

And

    [From Bruce Abbott (950831.1000 EST)]
    It is, as you say, a model of the observable behavior of the
    particles and not of the particles themselves. It is as if a given
    particle "looks around" and "sees" what the other particles around it
    are doing, and then alters its direction of movement to be aligned with
    the average direction of movement it perceives in those other
    particles.

Elsewhere in their replies, Peter and Bruce agree that the physicists'
model is S-R, or S-R like, but, in the quoted passages, they suggest one
might consider that the particles in the physicists' model act like
perceptual control systems. In my first post on the physicists' particle
model (PPM), I said the blurb in Science News gave insufficient information
for us to know whether (a) movements of the particles in the PPM were
controlled by external forces, or (b) the particles moved to control their
own perceptual signals, or something analogous to perceptual signals. The
original article, in Physical Review Letters, removed that uncertainty:
even though the physicists _talked about_ the modeled particles as though
they embodied properties of living systems, the particles were actually
modeled as passive objects whose movements were governed by external
forces, just as the movements of iron filings are governed by magnetic
fields. The only difference between particles in the PPM, and iron
filings, is that the particles were all given the same, arbitrary, constant
velocity. The authors tell us all of that, so it would be inappropriate to
assume they thought of their particles in any other way.

I believe Peter and Bruce are overly generous when they suggest that the
physicists might have thought of their particles as control systems. The
particles controlled nothing; the "rule" that operated in the physicists'
program played the role of an all-seeing observer, external to all
particles, and the actions of the observer -- a god-like observer --
completely determined the movements of each particle.

At one point, Bruce suggested,

    . . . This line of thinking leads me to question whether it is
   necessary to assume that either the physicists or their computer are
   acting in some sense as the god of the particles, any more than it is
   necessary to do so when describing how the multiple independent control
   systems of the crowd demo get updated by the program at each time-step.

But as Bruce said elsewhere, and as Bill Powers described in his reply, the
PCT model is a _generative model_, in which the workings of the
organization modeled _within_ each control system result in the system
generating behaviors by which the system controls its own perceptual
signals, with incidental side effects that we can notice, as observers
external to the controlling systems. The PPM is a purely _descriptive_
model that reproduces some of the things (movements, patterns, etc.) that
can be perceived by those who observer the particles, but that might not be
(and in the case of the PPM, are not) perceived by the particles
themselves. In the PPM, the particles do not perceive, hence, they cannot
control perceptions. In the PPM, nothing is modeled beyond external
appearances.

Bruce also said:

    The question I have is, is this model incompatible with a control
    model? It seems to me that a collection of autonomous control-system
    particles, controlling, say, their own perceived distances from their
    neighbors (and perhaps the rate of change of those distances), might be
    described as "looking around, seeing what their neighbors are doing,
    and altering their movements" in the way envisioned by the authors of
    this article.

I agree that you might see such things, were you to observe a group of PCT
systems, controlling for the perceptions you described, but, in that case,
the PCT systems you envision are not at all like the the particles
envisioned by the physicists. Similarities in the outward appearances of
behavior do not imply similarities in causes of behavior, a fact that you
acknowledge elsewhere in your reply.

Not surprisingly, I agreed wholeheartedly with the reply from Bill Powers
(Bill Powers (950831.0730 MDT)]). Bill expanded on the fact that the model
in PCT is generative; the model in PPM is descriptive. Bill drew exactly
the ties I had in mind, between the ideas in my posts, and the PCT modeling
of movements by E. coli. And his example of the movements of falling
raindrops is right on the mark. All of Bill's comments go directly to the
huge differences between the generative model in PCT, and the descriptive
one in PPM.

One last point. Bill said:

    What these physicists did was to impose a rule such that the particles
    moved in the average direction of the particles around them, plus a
    random variation. Behavior which fits this rule also fits the
    implications of the rule. However, it might also fit the implications
    of many other rules that could have been imposed.

Yes! That was one of the points Bill and I made in "Models and their
worlds" (available in _Closed Loop, or from the authors, or soon -- if I
get the file converted -- through the CSG home page on WWW). If one is
content to build a model that recreates only the outward appearances of
behavior, then there are many (perhaps an infinite number of) plausible
models. They can take many different forms, as many as the number of
lineal causes (rules) for behavior that one can imagine. On the other hand,
if one wants to build a model that controls its own perceptions, the
choices are severely constrained; the model must be a model of a control
system. There were no control systems in the article in Physical Review
Letters.

Later,

Tom