Ye Gods! Rules!

CSGnet Archive CSG1995
1

[From Bruce Abbott (950902.1215 EST)]

Tom Bourbon (950902.0115) --

Tom, your reply arrived just after I sent out my reply to Bill Powers' post
on this topic, so I will limit my comments to new ground.

I believe an important point is involved here, one that it easily
overlooked. The physicists' rule is a surrogate for them. When their
program runs, it (the program) applies the rule in a way that makes the
rule a stand-in for the physicists. It is as though, on every program step
(time interval), they look at each particle, one at a time, and for each
one they calculate the average direction of movement of all other particles
within radius r of the particle. Then they arbitrarily align the direction
of movement of each particle with the average direction of movement of its
immediate neighbors, within radius r.

It is as though, on every program step (time interval), every particle is
affected by the motions of the particles around it within a certain radius.
If you (the physicist) make this assumption, it can be shown (via the
computer model) that the particles will behave in a certain way. If you
(the control-systems theorist) make other assumptions, it can be shown (via
the computer model) that the particles will behave in a certain way. Both
models embody the rules of the modeler and apply them stepwise to each particle.

The only real difference in the two models is that in the ferromagnetic
model, the mechanism through which the physicists' rule emerges is left
unspecified; it is just assumed rather than emerging from the properties of
the mechanism. In this sense the physicists' model is merely descriptive.
It says that _if_ the particles behave in this way, then certain predictable
consequences follow. The control model with its more detailed specification
of mechanism may or may not confirm that the particles will behave in
accordance with the physicists' "as-if" rule.

Why do they use that rule? Why do they arbitrarily make each particle move
in the same direction as its neighbors? Because they want to see particles
moving in the same direction as their neighbors, and they will keep
tinkering with the direction of each individual particle until they (the
physicists) see particles moving in the same direction. The physicists, in
the guise of their rule, are controlling their own perceptions. They have
observed, or read about, phenomena involving large assemblages of objects
of various kinds and sizes (particles, molecules, bacteria, birds, fish,
automobiles in traffic, etc.), in which the objects "tend" to move in the
same direction. They started with phenomena observed in nature, and they
wanted to reproduce certain features of those observations in simulation;
they wanted to see large numbers of simulated particles go from a state in
which their directions of movement varied in a random manner, to a state in
which they moved in the same direction. They succeeded.

If that's all they were looking for, then their system is a tautology. It
begins by _assuming_ that the particles will tend to move in the same
direction as the average direction of their immediate neighbors; it would
therefore be no surprise to find that the particles then do actually move in
the same direction.

No, what is interesting about their simulation is the following: (a) given
random initial directions and a certain amount of "noise," the particles
tend to self-align until the whole group of them is moving in the same
direction, but _only_ under certain conditions involving the level of noise
(spontaneous changes in direction of individual particles), and (b) just
like a school of fish, the whole group from time to time spontaneously
changes direction en masse. One can expect this sort of behavior from _any_
group of particles _whatsoever_ in which the particles behave at least
approximately as the physicists' rule states. It doesn't matter whether the
rule holds because of external forces acting on the particles or because of
their internal organization as living control systems; so long as it does
hold, the "particles" will behave as the simulation indicates.

If that doesn't interest you, fine. I find it fascinating.

Concerning their success, in so far as it might be useful in the study of
living control systems, I say, "so what?"

So your physicists are to be criticized because they discovered a general
principle which may explain certain aggregrate behaviors of particles
ranging from ferromagnetic atoms to wildabeasts and failed to mention that
control is involved only in the latter, even though this point may be
irrelevant from the perspective of the principle. A bit unfair, don't you
think?

Bill [From Bill Powers (950901.1430 MDT)] and Rick [From Rick Marken
(950901.1400)] have already replied to that line of discussion from Bruce.
I share their conclusions, to the effect that any resemblance between the
movements of iron filings influenced by a magnetic field, and the movements
of automobiles, with drivers inside, in traffic are spurious and trivial.

The physicists' model has nothing to do with iron filings: you are confusing
their discussion of ferromagnetic spin-alignment (which applies to atoms
within a matrix) with your own quotation from James. Also, my traffic
problem applies to another analogy (fluid dynamics), not to iron filings.
And my analogy of traffic to fluid motion is neither spurious nor trivial,
but is in fact powerful when applied to the properly analogous conditions.
I expect that a control-system model will be even more powerful.

Regards,

Bruce

2

[From Bruce Abbott (950905.1525 EST)]

Tom Bourbon (950905.1330) --

Ah, so you HAVEN'T gone into hibernation again! I was hoping to have a
reply from you to my last couple of posts regarding the physicists'
ferromagnetic model of fish behavior, but the thread just seems to have been
dropped. A while back Rick said that if he didn't reply to something I said
it was because there was no disturbance. So I guess you now agree with me,
right? (;->

Regards,

Bruce

3

Bill,

Ah, so you HAVEN'T gone into hibernation again!

Not at all. Just busy preparing to sell a house -- painting, fixing
fences, hauling our kids "stuff" from our garage to theirs in another city
-- exciting things like that. All part of the aftermath of my sudden
"career change."

I was hoping to have a reply from you to my last couple of posts
regarding the physicists' ferromagnetic model of fish behavior, but the
thread just seems to have been dropped.

Not to worry :wink: I logged on this time to print out your most recent
comments. After I read them again, from hard copy, I will prepare a
response. It will probably go out early in the morning -- I am a night
creature who starts warming up at about 10:00 PM and goes on until about
2:30 or 3:00 in the morning.

A while back Rick said that if he didn't reply to something I said it
was because there was no disturbance. So I guess you now agree with me,
right? (;->

You only wish! ;->

See you in the morning.

Tom

4

[From Tom Bourbon (950906.0008)]

A little more on the thread that began when I described a new article by
physicists who believe they have a model for coordinated, cooperative
behavior.

[From Bruce Abbott (950902.1215 EST)]

Tom Bourbon (950902.0115) --

Bruce, I apologize for the delay in responding. There are lots of things to
attend to while we prepare to sell a house, move yet again, and settle
into our new life after my recent "career change."

Before I reply directly to a few of your remarks, let me post an agreement
with you. I'll do it by way of quoting part of your exchange with Bill
Powers on this subject.

ยทยทยท

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

[From Bruce Abbott (950902.1100 EST)]

Bill Powers (950901.1430 MDT) --

    Bruce Abbott (950901.1045 EST)

    Yet in the aggregate, it is possible to learn some important
    lessons about traffic flow if one simply imagines that each car is
    a particle responding to "forces" impressed on it by the particles
    around it (e.g., a repulsive force falling off with the square of
    the distance to the particle ahead, to take a simple case).

The question is whether you want to explain natural phenomena with
analogies or by making models which are intended to show how the
phenomena actually work.

I agree completely. The only point I'm attempting to make here is that
descriptive models can be useful, by emphasizing commonalties among
apparently diverse phenomena.

TB: Agreed. As Bill P., Rick, you, and I have said in several posts, the
physicists created one kind of purely descriptive model. It is not a
generative model for the causes of what it describes. It does nothing more
than re-state some of the phenomena we can observe when, for various
reasons, and for various kinds of objects, many objects move in the same
direction, or nearly the same direction. The most obvious point where my
comments about their model might differ from yours is that I would say such
a model might, or might not, be useful, in that some people who see it will
attach more significance to the common patterns of movement than those
patterns deserve.

Back to [From Bruce Abbott (950902.1215 EST)].

It is as though, on every program step (time interval), every particle is
affected by the motions of the particles around it within a certain
radius. If you (the physicist) make this assumption, it can be shown (via
the computer model) that the particles will behave in a certain way. If
you (the control-systems theorist) make other assumptions, it can be shown
(via the computer model) that the particles will behave in a certain way.
Both models embody the rules of the modeler and apply them stepwise to
each particle.

In their article in Physical Review Letters, the physicists did not
describe a mechanism for any of the various kinds of coordinated movement
they cited. They did not describe or model any characteristic of any
particle. Neither did they describe or model any force or any other
influence that might act on a particle. Instead of modeling particles and
forces, a procedure I would be more inclined to like, their procedure was
to apply a "rule" to each particle. The _rule_ altered the direction of
movement of each particle, bringing it into alignment with the average
direction of its immediate neighbors. The resulting similarity in their
directions of movement was determined by the rule, not by any effect of a
modeled force acting on a modeled particle. _You_ are providing the idea
that the results are analogous to those when forces act on particles. _I_
mistakenly imposed the same interpretation as you, in my earlier posts.

A brief point. The fact that we use computational "steps" in our modeling
is an artifact of the way digital computers work. Given a decent analog
computer, there would be no "steps" in PCT modeling, just good old
continuous analog solutions. What is more, the fact that both the
physicists and we run our models in computers has no bearing whatsoever on
the fact that they run a rule-driven description, and we run a generative
control model.

The only real difference in the two models is that in the ferromagnetic
model, the mechanism through which the physicists' rule emerges is left
unspecified; it is just assumed rather than emerging from the properties
of the mechanism.

Well, if we ignore the fact that one model is descriptive, and the other is
generative, I guess you are right. But even then, your "only real
difference" is immense.

In this sense the physicists' model is merely descriptive. It says that
_if_ the particles behave in this way, then certain predictable
consequences follow.

Yes. Indeed, the particles can do nothing but behave the way they do, with
the results that inevitably follow: A "model" that does nothing more than
to arbitrarily make every particle move in the same direction as its
neighbors can produce no other result than particles moving in the same
directions as their neighbors. (As Mary Powers said in a post, it _does_
set my teeth on edge, calling their rule a "model.")

TB

(The physicists) have observed, or read about, phenomena involving large
assemblages of objects of various kinds and sizes (particles, molecules,
bacteria, birds, fish, automobiles in traffic, etc.), in which the
objects "tend" to move in the same direction. They started with
phenomena observed in nature, and they wanted to reproduce certain
features of those observations in simulation; they wanted to see large
numbers of simulated .particles go from a state in which their directions
of movement varied in a random manner, to a state in which they moved in
the same direction. They succeeded.

If that's all they were looking for, then their system is a tautology. It
begins by _assuming_ that the particles will tend to move in the same
direction as the average direction of their immediate neighbors; it would
therefore be no surprise to find that the particles then do actually move
in the same direction.

Bruce, their system begins, not by _assuming_ that the particles will move
in the same direction, but by _dictating_ that they will. If Ed Ford were
to ask them, "What's the rule?", they would say, "That when the simulation
is over, every particle will tend to move in the same direction as its
neighbors." :wink: As you say, that they do move that way comes as no
surprise.

What you say next puzzles me. You describe what you think is
interesting about their simulation, but the two points you describe are not
in their article. Where did these interesting points come from? Are you
running on pure model-based control? If so, I hope Hans sees the dangers
more clearly, now. :slight_smile:

No, what is interesting about their simulation is the following: (a)
given random initial directions and a certain amount of "noise," the
particles tend to self-align until the whole group of them is moving in
the same direction, but _only_ under certain conditions involving the
level of noise (spontaneous changes in direction of individual particles),

Well, the physicists do describe different degrees of similarity in the
final directions of movement, depending on the _density_ of the particles,
and on the amount of random "noise" in their new directions of travel. In
general, they found that, for any particular magnitude of "noise," the
final degree of similarity in direction of movement was affected strongly
and positively by increasing density of the particles. On the other hand,
the magnitude of the directional "noise" was _inversely_ related to the
final degree of similarity in the directions of movements. The more
densely packed the particles, and the smaller the "jitter" in the
new directions imposed by the god-in-the-rule, the more similar the final
directions of movement.

and (b) just like a school of fish, the whole group from time to time
spontaneously changes direction en masse. One can expect this sort of
behavior from _any_ group of particles _whatsoever_ in which the particles
behave at least approximately as the physicists' rule states. It doesn't
matter whether the rule holds because of external forces acting on the
particles or because of their internal organization as living control
systems; so long as it does hold, the "particles" will behave as the
simulation indicates.

Bruce, after reading this interesting claim, I've scoured the article
from front to back, and back again, several times. I can't find a word
about this result of the simulation. I must be wearing blinders. Can you
tell me where it is? You seem to think it is important.

If that doesn't interest you, fine.

It doesn't, in part because "it" doesn't exist in the article I am
describing. It wouldn't really excite me all that much, even if it did
exist. :slight_smile:

I find it fascinating.

I have no problem with that. There is no accounting for taste. :wink:

Concerning their success, in so far as it might be useful in the study of
living control systems, I say, "so what?"

So your physicists are to be criticized because they discovered a general
principle which may explain certain aggregate behaviors of particles
ranging from ferromagnetic atoms to wildebeests and failed to mention that
control is involved only in the latter, even though this point may be
irrelevant from the perspective of the principle.

How did I give you that idea? However I did it, I recant and apologize.
For one thing, I said elsewhere that purely descriptive "models" like
theirs are apparently of some use in physics, and I can't fault these
authors if that is the case. Also, I don't believe I ever said my
complaint was that they failed to mention control. I don't think I said
that, although it is another messy little detail they overlooked. My
major complaints have to do with sweeping claims that a pure re-statement
of a description "explains" _anything_, be it ferromagnetic spin-alignment,
iron filings rushing to the magnets they so dearly love (I do know the
distinction), or Romeo and Juliet finding themselves in each other's arms
in spite of walls, balconies and ancient familial hatreds. Write it off to
my high-gain, fixed-reference, control system for skepticism, but I do not
like the idea that a rule which _mandates_ that particles will "tend" to
move in the same direction _explains_ the similarity of their directions,
other than that the rule makes it happen.

A bit unfair, don't you think?

Not at all! :slight_smile:

Remember, I agree with you, wholeheartedly, when you say the physicists'
"model" is a description, nothing more. I might have been more favorably
impressed had they worked at a less "abstract" level, and instead used
models of particles and forces, on the one extreme, and control models, on
the other.

Later,

Tom

5

[From Bruce Abbott (950906.1050 EST)]

Tom Bourbon (950906.0008) --

Bruce Abbott (950902.1215 EST)

Bruce, I apologize for the delay in responding. There are lots of things to
attend to while we prepare to sell a house, move yet again, and settle
into our new life after my recent "career change."

No need to apologize; in fact, I'm awfully busy myself (gads!). I just
needed to know the score.

Thanks, Tom, for the cogent, calm, and thoughtful reply. This is the high
level of discourse I've come to expect on CSG-L, certain recent exchanges
not withstanding.

Getting back to the "ferromagnetic model," it would appear that we are in
general agreement on most issues. The physicist's model does little more
than explore the consequences of an imposed rule, without regard to any
physical mechanism that might effectively implement the rule.

In their article in Physical Review Letters, the physicists did not
describe a mechanism for any of the various kinds of coordinated movement
they cited. They did not describe or model any characteristic of any
particle. Neither did they describe or model any force or any other
influence that might act on a particle. Instead of modeling particles and
forces, a procedure I would be more inclined to like, their procedure was
to apply a "rule" to each particle. The _rule_ altered the direction of
movement of each particle, bringing it into alignment with the average
direction of its immediate neighbors. The resulting similarity in their
directions of movement was determined by the rule, not by any effect of a
modeled force acting on a modeled particle. _You_ are providing the idea
that the results are analogous to those when forces act on particles. _I_
mistakenly imposed the same interpretation as you, in my earlier posts.

Unfortunately I have not yet obtained a copy of the article and read it, so
I've been relying on those descriptions and interpretations, and apparently
filling in a few details out of my imagination (see below)! However, it
does seem to me that "applying the rule" to each particle on each iteration
of the program can be viewed as computing an influence as "seen" by each
particle at a given moment in time, rather than as seen by the
all-omnipotent God of the computer. The rule specifies how each particle
will behave under that influence without giving the slightest hint of the
mechanism through which such a functional relationship might emerge.

A brief point. The fact that we use computational "steps" in our modeling
is an artifact of the way digital computers work. Given a decent analog
computer, there would be no "steps" in PCT modeling, just good old
continuous analog solutions. What is more, the fact that both the
physicists and we run our models in computers has no bearing whatsoever on
the fact that they run a rule-driven description, and we run a generative
control model.

I understand and agree. My reason for bringing up the step-wise
computations was merely to illustrate that the physicists' model and a
control model would both work by computing local inputs _as "seen"_ by each
particle. I'm sure both models could be implemented equally well on an
analog computer.

The only real difference in the two models is that in the ferromagnetic
model, the mechanism through which the physicists' rule emerges is left
unspecified; it is just assumed rather than emerging from the properties
of the mechanism.

Well, if we ignore the fact that one model is descriptive, and the other is
generative, I guess you are right. But even then, your "only real
difference" is immense.

I'm talking specifically about the fact that the aggregrate behavior of the
particles under both models emerges from relationships (rules) applied at
each moment in time to each individual particle, from its point of view.
Beyond that similarity, the difference between the two models is, as you
say, immense. The control model would specify potentially real connections
among potentially real mechanisms (sensors, effectors, etc.); to the extent
that such a model mirrors the actual mechanisms at work in, say, real fish,
the aggregrate behavior observed in schooling fish ought to emerge as a
consequence. Powerful stuff!

If that's all they were looking for, then their system is a tautology. It
begins by _assuming_ that the particles will tend to move in the same
direction as the average direction of their immediate neighbors; it would
therefore be no surprise to find that the particles then do actually move
in the same direction.

Bruce, their system begins, not by _assuming_ that the particles will move
in the same direction, but by _dictating_ that they will. If Ed Ford were
to ask them, "What's the rule?", they would say, "That when the simulation
is over, every particle will tend to move in the same direction as its
neighbors." :wink: As you say, that they do move that way comes as no
surprise.

I'm using the word "assuming" as in deductive logic; thus "dictate" and
"assume" are equivalent terms in this context. Deductive systems do nothing
more than reveal the consequences of a set of assumptions, though these may
be far from apparent, in which case the result may indeed be surprising. I
can't help thinking that there must be more here than you've described,
otherwise the results are so obvious that there would be little merit in
publishing them. But maybe not.

What you say next puzzles me. You describe what you think is
interesting about their simulation, but the two points you describe are not
in their article. Where did these interesting points come from? Are you
running on pure model-based control? If so, I hope Hans sees the dangers
more clearly, now. :slight_smile:

I guess I imagined it. For some reason I pictured the behavior of schools
of fish, which show frequent, sudden changes in aggregrate direction. I
imagined that with the right amount of "noise" (random directional change)
added to the directional behavior imposed by the rule that local pockets of
directional change might emerge and rapidly spread to the remaining
particles, resulting in the kind of sudden directional change in the
aggregrate I just described. I would find this aggregrate behavior more
interesting as an implication of the assumptions of this model than the fact
that the fish become directionally aligned over time.

So your physicists are to be criticized because they discovered a general
principle which may explain certain aggregate behaviors of particles
ranging from ferromagnetic atoms to wildebeests and failed to mention that
control is involved only in the latter, even though this point may be
irrelevant from the perspective of the principle.

How did I give you that idea? However I did it, I recant and apologize.
For one thing, I said elsewhere that purely descriptive "models" like
theirs are apparently of some use in physics, and I can't fault these
authors if that is the case. Also, I don't believe I ever said my
complaint was that they failed to mention control. I don't think I said
that, although it is another messy little detail they overlooked. My
major complaints have to do with sweeping claims that a pure re-statement
of a description "explains" _anything_, be it ferromagnetic spin-alignment,
iron filings rushing to the magnets they so dearly love (I do know the
distinction), or Romeo and Juliet finding themselves in each other's arms
in spite of walls, balconies and ancient familial hatreds. Write it off to
my high-gain, fixed-reference, control system for skepticism, but I do not
like the idea that a rule which _mandates_ that particles will "tend" to
move in the same direction _explains_ the similarity of their directions,
other than that the rule makes it happen.

Sorry if I misunderstood your meaning. If there is no more to the
physicists' model than you've described, I'd have to agree with you. But I
do think that there is room for models at different levels of specificity,
as in the traffic example I mentioned. It's a good thing, too, or else we'd
be stuck having to completely model every detail of every sensory receptor
and every effector before we could build even a simple control model
involving a living organism. A good model at a very general level of
abstraction may help to understand certain emergent phenomena and make
useful predictions even though the underlying mechanisms through which the
interactions occur remains unspecified. However, a much deeper
understanding emerges from a generative model of the underlying mechanism.
Where I have problems is when the functional (descriptive) model is asserted
to be the ultimate goal, as appears to be the case in EAB.

Remember, I agree with you, wholeheartedly, when you say the physicists'
"model" is a description, nothing more. I might have been more favorably
impressed had they worked at a less "abstract" level, and instead used
models of particles and forces, on the one extreme, and control models, on
the other.

Me, too. You agree? Hey, I must be dreaming!

Regards,

Bruce

6

[From Tom Bourbon (950908.1530)]

[From Bruce Abbott (950906.1050 EST)]

Tom Bourbon (950906.0008) --

Getting back to the "ferromagnetic model," it would appear that we are in
general agreement on most issues. The physicist's model does little more
than explore the consequences of an imposed rule, without regard to any
physical mechanism that might effectively implement the rule.

Right. Here we go again -- agreeing.

In their article in Physical Review Letters, the physicists did not
describe a mechanism for any of the various kinds of coordinated movement
they cited. They did not describe or model any characteristic of any
particle. Neither did they describe or model any force or any other
influence that might act on a particle. Instead of modeling particles
and forces, a procedure I would be more inclined to like, their procedure
was to apply a "rule" to each particle. The _rule_ altered the direction
of >movement of each particle, bringing it into alignment with the
average direction of its immediate neighbors. (TB: And so on.)

Unfortunately I have not yet obtained a copy of the article and read it,
so I've been relying on those descriptions and interpretations, and
apparently filling in a few details out of my imagination (see below)!

I was hoping a few other people might read the article. For all anyone
knows, I might be hallucinating this theme, or I might have seriously
misunderstood or misrepresented the work of the physicists. I don't
believe any of those things happened, though. The article seems clear
enough. I can even understand their math, so it must be pretty low-level
physics. :slight_smile:

However, it does seem to me that "applying the rule" to each particle on
each iteration of the program can be viewed as computing an influence as
"seen" by each particle at a given moment in time, rather than as seen by
the all-omnipotent God of the computer. The rule specifies how each
particle will behave under that influence without giving the slightest
hint of the mechanism through which such a functional relationship might
emerge.

Perhaps we could think of each application of the rule as equivalent to,
"computing an influence as "seen" by each particle at a given moment in
time, rather than as seen by the all-omnipotent God of the computer," but
were we to do that, what would the particle see? The rule, nothing more.
Beyond that, as Bill Powers said in a post on this tread, it is hard to
imagine a rigorous model of forces, and of properties of particles, that
would lead to an "influence" like the one imposed by the rule. It all
still looks god-like, to me.
. . .

Well, if we ignore the fact that one model is descriptive, and the other
is generative, I guess you are right. But even then, your "only real
difference" is immense.

I'm talking specifically about the fact that the aggregate behavior of
the particles under both models emerges from relationships (rules) applied
at each moment in time to each individual particle, from its point of
view.

Interesting. I don't think of it quite that way. In a thorough model of
particles and forces, perhaps we could speak of all of the modeled forces
that act on the particle at time x as being "applied to" the particle.
However, if we are talking, not about models of particles, but about models
of living systems, then our generative model must include a model of the
living system, as well as of its environment. In that case, at time x, the
behavior of the modeled system results from the present magnitudes of
signals within the system. I have some difficulty imagining how the
present magnitudes of signals _inside_ the system might be described as
relationships _applied to_ the system. I think this is one of the
irreconcilable differences between lineal models of forces applied to
inanimate objects, and generative negative-feedback models of systems
viewed as controllers of some of their own perceptions.

Beyond that similarity, the difference between the two models is, as you
say, immense. The control model would specify potentially real
connections among potentially real mechanisms (sensors, effectors, etc.);
to the extent that such a model mirrors the actual mechanisms at work in,
say, real fish, the aggregate behavior observed in schooling fish ought
to emerge as a consequence. Powerful stuff!

You bet it is! We agree, again.
. . .

I can't help thinking that there must be more here than you've described,
otherwise the results are so obvious that there would be little merit in
publishing them. But maybe not.

Take a look at their article. Perhaps we can explain its publication by
the fact that the results were surprising to other physicists.

I was disappointed by the style of the article. Way back in my freshman
year in college, I was a physics major. From back then, I remember
publications about physics were always written like -- well, like physics.
In contrast with the writing in physics back then, the article I have
described looks a lot like what you can find in any journal in psychology,
or sociology.
. . .

What you say next puzzles me. You describe what you think is
interesting about their simulation, but the two points you describe are
not in their article. Where did these interesting points come from? Are
you >running on pure model-based control? If so, I hope Hans sees the
dangers more clearly, now. :slight_smile:

I guess I imagined it. For some reason I pictured the behavior of schools
of fish, which show frequent, sudden changes in aggregate direction. I
imagined that with the right amount of "noise" (random directional change)
added to the directional behavior imposed by the rule that local pockets
of directional change might emerge and rapidly spread to the remaining
particles, resulting in the kind of sudden directional change in the

aggregate I just described.

Is it fair to say, in other words, that you are more impressed by the
behavior of living things, than by behavior as sterile as that of the
physicists' model (as I have described it here)? :slight_smile:

You imagined something you remember about the behavior of animals who share
a limited space, in this case, the behavior of fish in a school. You
thought the physicists' model ought to account for behavior like that. It
doesn't.

There is always a danger of doing what you did, when we try to work with
highly abstract models, especially with lineal cause-effect models for
descriptions of the superficial appearances of behavior. (Whew!) With their
excessive generality, and their lack of detail, highly abstract models
become like inkblots, onto which we can "project" whichever phenomena we
think they should explain. Sometimes it is difficult for us to recognize
that overly-abstract models explain nothing; we can so easily believe that
they explain many things. "Models" like that are abundant in behavioral
science, social science, cognitive science, neuroscience, and, now, in
physics.

I would find this aggregate behavior more interesting as an implication
of the assumptions of this model than the fact that the fish become
directionally aligned over time.

We agree! You identified a major point that bothered me when I first read
that their impoverished rule was supposed to tell us something important
about collective, coordinated, cooperative behavior. Their model _does
not_ imply the amazing things that interest you and me. It cannot. It
treats _all_ moving objects, from particles to bacteria to fish to people,
as though they were lineal cause-effect systems.
. . .

If there is no more to the physicists' model than you've described, I'd
have to agree with you. But I do think that there is room for models at
different levels of specificity, as in the traffic example I mentioned.

See my comments above.

The physicists "mention" the analysis of traffic, but they do not show how
one might use their model to describe or predict any of the interesting
phenomena we observe in traffic flow. In fact, I can't think of practical
situations where you see large numbers of drivers, all of whom keep their
automobiles traveling at the same arbitrary velocity (each car traveling at
the same unvarying velocity as all of its neighbors), and all of whom
immediately align their cars to travel in the same direction as the average
direction of neighboring cars. Theirs is not a model for phenomena we see
in traffic, or anywhere else that living systems behave.
. . .

You agree? Hey, I must be dreaming!

Me too. With so much agreement going on, this thread is turning into a
nightmare. Bring back the good old days -- bricks, fists, clubs! :slight_smile:

Later,

Tom

7

[From Bruce Abbott (950908.2000 EST)]

Tom Bourbon (950908.1530) --

Bruce Abbott (950906.1050 EST)

Way back in my freshman
year in college, I was a physics major.

You were? So was I! No wonder we've been agreeing so much lately . . .

Is it fair to say, in other words, that you are more impressed by the
behavior of living things, than by behavior as sterile as that of the
physicists' model (as I have described it here)? :slight_smile:

Yep! I'll take a good-ol' slimy, wiggley animal over an elementary particle
any day -- snark over quark, as it were. What I want to know is, why do
physicists call their field one of the "hard" sciences? It's damned _easy_,
I'd say, compared to unraveling the mysteries of living, behaving organisms.
Even then, they haven't got it right: relativity and quantum mechanics are
fundamentally incompatible theories. And now what do they give us? _The
Emperor's New Mind_ and fish-particle physics.

You agree? Hey, I must be dreaming!

Me too. With so much agreement going on, this thread is turning into a
nightmare. Bring back the good old days -- bricks, fists, clubs! :slight_smile:

Well, there's always Rick! (;->

Regards,

Bruce

8

[From Erling Jorgensen (950909.930)]

Tom Bourbon (950908.1530)

...what would the particle see? The rule, nothing more.
Beyond that, as Bill Powers said in a post on this tread, it is hard to
imagine a rigorous model of forces, and of properties of particles, that
would lead to an "influence" like the one imposed by the rule. It all
still looks god-like, to me.

When this first came up, I tried manipulating the CROWD demo, to
see if the (outward) appearance of "flocking behavior" could be
produced by a _generative_ model, where each particle _was_
controlling its own limited inputs, with no perception of the overall
pattern. The most I could tweak into shape was some vague
swirling patterns.

Again, as several of you have suggested already, the individual
particles (at least in the CROWD demo) don't know about, could
care less about, and have no way of perceiving the pattern that emerges
out of their collective motion. That pattern is a Relationship perception
of _mine_, as an outside observer. In other words, I'm observing
_by-products_ of the particles' control of such things as proximity and
direction -- toward their _own_ goals and around nearby disturbances.

Actually, the CROWD demo may not be well suited to seeing something
like "flocking," or a "school of fish." The particles move toward either
a stationary or a moving goal (or both). I did not find an easy way to
keep each particle in continuing (controlled) relation to its neighbours,
other than as passing disturbances that must be navigated around.

What I tried was having each particle "followed" by two other particles.
Lead particles needed to be going slightly faster (i.e. higher gain)
toward their goals. When the traces are left on, the outside observer
can perceive some generalized swirlings, which do not seem to stabilize
into "flocks" or "galaxies" or what-have-you. It sort of looked like
"modern art." :wink: What it would take, I think, is control systems
being set up the way Bruce Abbott (950831.1000 EST) was envisioning them:

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"

It was actually this posting of Bruce's that got me wondering what could
be done with our own generative model. Could the physicists' phenomenon,
which they imposed from the outside through their rule, (note [as you
have], it was not the _particles'_ rule) -- could that be reproduced
_from the inside_ via "autonomous control-system particles"? I suspect it
could with the right package of elementary control systems. BTW, for using
"elementary" components, the CROWD demo is really quite sophisticated!

Back to Tom's posting:

I have some difficulty imagining how the
present magnitudes of signals _inside_ the system might be described as
relationships _applied to_ the system. I think this is one of the
irreconcilable differences between lineal models of forces applied to
inanimate objects, and generative negative-feedback models...

It was this aspect that confused me, too, about Bruce's remarks, and made
me wonder if I was following him correctly. Both of your recent remarks
have cleared up the difficulty. As you indicate, when you agree with
someone, you tend to nod and smile, but there's not too much more to
say...

Their model _does
not_ imply the amazing things that interest you and me. It cannot. It
treats _all_ moving objects, from particles to bacteria to fish to people,
as though they were lineal cause-effect systems.

((( :slight_smile: ))) ...Is that a nodding smile??

All the best,
        Erling