> After a period of time the group will become stationary (at least it has so
> far every time that I have done this demo...).
Every time I did it, it became stationary, just as you got.
I tried to modelise that behavior but I coulnt understand why it stops.
If somebody more clever than I...
Cordialement
Jean-Louis Cordonnier
From: "Jean-Louis Cordonnier" <jlcord@wanadoo.fr>
36, rue Lavisse
66000 Perpignan (France)
33 4 68 63 87 04
Fast Demonstration of Complex Behavior
-
"Jean-Louis Cordonnier"
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
-
"Brian (Bo) Newman"
- Junior Member
- Posts: 3
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
> I tried to modelise that behavior but I couldnt understand why it stops.
I would think it stops because the "problem" has a solution and those
conditions have been met. As long as conditions (the rules) dont change,
once the group reaches equilibrium ( becoming "stationary" in this case) it
will stay there. Telling one person to change their focus should only
"bump" the system, and it should again return to quiescence.
If you want to extend this experiment (and have the time, the groups
interest/commitment, and have someone with sufficient background
in complexity theory to lead the de-brief ) try the following:
(1) Repeat the process, but this time, without the rest of the group
knowing, instruct one person to change their focus as soon as they
feel they have satisfied the initial requirement. In other words, as soon
as that person feels they are equal distant from both of their
"targets," they are to drop one of their targets and pick another. And
they are to continue to do so each time they feel they have satisfied the
equidistance rule.
(2) Repeat the process again, but this time have 2 people changing, then 3
people, etc.
(3) Dont pre-select people to continuously change their targets, but
rather allow everyone to change their target if they feel the target is
acting erratic (moving around too much or with no consistent pattern) and
trying to achieve their goal (equidistance from two selected targets) would
be too difficult, or take too long. When the people feel they personally
have achieved the goal, have them raise their hand and keep it up as long
as they feel they continue to satisfy the goal condition.
- Bo Newman -
From: "Brian (Bo) Newman" <bonewman@3-cities.com>
I would think it stops because the "problem" has a solution and those
conditions have been met. As long as conditions (the rules) dont change,
once the group reaches equilibrium ( becoming "stationary" in this case) it
will stay there. Telling one person to change their focus should only
"bump" the system, and it should again return to quiescence.
If you want to extend this experiment (and have the time, the groups
interest/commitment, and have someone with sufficient background
in complexity theory to lead the de-brief ) try the following:
(1) Repeat the process, but this time, without the rest of the group
knowing, instruct one person to change their focus as soon as they
feel they have satisfied the initial requirement. In other words, as soon
as that person feels they are equal distant from both of their
"targets," they are to drop one of their targets and pick another. And
they are to continue to do so each time they feel they have satisfied the
equidistance rule.
(2) Repeat the process again, but this time have 2 people changing, then 3
people, etc.
(3) Dont pre-select people to continuously change their targets, but
rather allow everyone to change their target if they feel the target is
acting erratic (moving around too much or with no consistent pattern) and
trying to achieve their goal (equidistance from two selected targets) would
be too difficult, or take too long. When the people feel they personally
have achieved the goal, have them raise their hand and keep it up as long
as they feel they continue to satisfy the goal condition.
- Bo Newman -
From: "Brian (Bo) Newman" <bonewman@3-cities.com>
-
Bill Braun
- Senior Member
- Posts: 73
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
The group exercise aimed at positioning everyone to be equidistant from two
other people in a group is reminiscent of the "boids" program. The
hypothesis of the program was that cohesion in flocks could be explained by
three rules: 1) match speed with birds in your neighborhood, 2) maintain
equal distance from all birds in your neighborhood and 3) always drive to
the center of the flock in your neighborhood. Neighborhood was defined as
the birds you could see in the immediate area.
There is no rule that says, "form a flock". Initial conditions were a
critical determinant of the direction and speed of the flock.
Some additional variations that might be interesting. A large space is
probably required.
One, have the group divide into four subgroups, each in the corner of a
room. Have one person leave their corner and walk (where and how fast is up
to them). "Release" another person, then another, etc. until everyone is
"free".
Two, when the group stops have someone at the edge of the group reposition
her/himself to the opposite side of the group by walking through the center
of the group. Then by walking around the perimiter of the group.
Three, when the group stops have two people (from anywhere) leave the group
and move in the direction of their choice. See if the group splits, remains
cohesive or passes through stages with same or different results.
Compare/contrast the collective behavior of the group based on different
starting conditions and different "course corrections".
Bill Braun
From: Bill Braun <medprac@hlthsys.com>
other people in a group is reminiscent of the "boids" program. The
hypothesis of the program was that cohesion in flocks could be explained by
three rules: 1) match speed with birds in your neighborhood, 2) maintain
equal distance from all birds in your neighborhood and 3) always drive to
the center of the flock in your neighborhood. Neighborhood was defined as
the birds you could see in the immediate area.
There is no rule that says, "form a flock". Initial conditions were a
critical determinant of the direction and speed of the flock.
Some additional variations that might be interesting. A large space is
probably required.
One, have the group divide into four subgroups, each in the corner of a
room. Have one person leave their corner and walk (where and how fast is up
to them). "Release" another person, then another, etc. until everyone is
"free".
Two, when the group stops have someone at the edge of the group reposition
her/himself to the opposite side of the group by walking through the center
of the group. Then by walking around the perimiter of the group.
Three, when the group stops have two people (from anywhere) leave the group
and move in the direction of their choice. See if the group splits, remains
cohesive or passes through stages with same or different results.
Compare/contrast the collective behavior of the group based on different
starting conditions and different "course corrections".
Bill Braun
From: Bill Braun <medprac@hlthsys.com>
-
"Jim Hines"
- Senior Member
- Posts: 88
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
The people-moving example is wonderful not only because its simple and cool.
Its wonderful because it also highlights a difficulty in agent-based
representations of dynamic systems: Analyzing agent-based representations
is more difficult than analyzing more traditional SD representations. (The
traditional way of representing systems in SD is equation-based, and in
fact, essentially differential equations).
Of course, **Ease** of analysis is not everything. In fact, there are other
representations (e.g. **linear** differential equations) that that would
make analysis even easier for SDers, but which arent powerful enough to
capture the range of dynamics were interested in. Note, though that even
though **ease** of analysis is not everything, analysis itself (i.e.
undertstanding how structure generates behavior) is.
So, my question is: what dynamics do agent-based representations capture
that makes it worth sacrificing the (modest) analytical ease of the
traditional SD way of representing systems.
Regards,
Jim Hines
jhines@mit.edu
Its wonderful because it also highlights a difficulty in agent-based
representations of dynamic systems: Analyzing agent-based representations
is more difficult than analyzing more traditional SD representations. (The
traditional way of representing systems in SD is equation-based, and in
fact, essentially differential equations).
Of course, **Ease** of analysis is not everything. In fact, there are other
representations (e.g. **linear** differential equations) that that would
make analysis even easier for SDers, but which arent powerful enough to
capture the range of dynamics were interested in. Note, though that even
though **ease** of analysis is not everything, analysis itself (i.e.
undertstanding how structure generates behavior) is.
So, my question is: what dynamics do agent-based representations capture
that makes it worth sacrificing the (modest) analytical ease of the
traditional SD way of representing systems.
Regards,
Jim Hines
jhines@mit.edu
-
"Brian (Bo) Newman"
- Junior Member
- Posts: 3
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
Jim,
Interesting question.
Can, "the traditional SD way of representing systems" adequately represent
a complex adaptive system exhibiting the dynamics of emergent behaviors?
Do we really need to "sacrifice" the models we have already developed.
How about separate methods for separate problems? Or hybrid methods and
models that combine features of SD (pick your flavor) and agent-based
(again, of the flavor of your choice)?
Methods and models are but codifications of knowledge we have already
developed. If that knowledge remains valid, why not use all that we know
when addressing a new situation.
Maybe we should ask ourselves, "what would MacGyver* do?"
- Bo Newman -
From: "Brian (Bo) Newman" <bonewman@3-cities.com>
Interesting question.
Can, "the traditional SD way of representing systems" adequately represent
a complex adaptive system exhibiting the dynamics of emergent behaviors?
Do we really need to "sacrifice" the models we have already developed.
How about separate methods for separate problems? Or hybrid methods and
models that combine features of SD (pick your flavor) and agent-based
(again, of the flavor of your choice)?
Methods and models are but codifications of knowledge we have already
developed. If that knowledge remains valid, why not use all that we know
when addressing a new situation.
Maybe we should ask ourselves, "what would MacGyver* do?"
- Bo Newman -
From: "Brian (Bo) Newman" <bonewman@3-cities.com>
-
Niall Palfreyman
- Senior Member
- Posts: 56
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
Jim Hines schrieb:
> So, my question is: what dynamics do agent-based representations capture
> that makes it worth sacrificing the (modest) analytical ease of the
> traditional SD way of representing systems.
Jim, I dont think you can know how good it is to me to see this
question explicitly formulated like this. _This_ is the central thrust
of my interest at the moment - I just wish I had more time to work on
it. I dont know the answer - the following are my thoughts at the
moment.
SD is used to treat not individuals, but aggregates. Thus, in a
simulation of spread of a genetic trait through a population, a MAS
(multi-agent system) simulation would have hundreds (thousands?) of
individuals with assorted traits, and then let them interact, while an
SD simulation would probably dive in with the Hardy-Weinberg equation.
For me the point is that MAS try to explain the aggregate behaviour in
terms of individuals and their random interactions, while SD tends (and
I am _only_ saying "tends") to draw the line of abstraction above the
random level, assuming that the random behaviour can be ignored if we
know the laws governing the overall dynamics of the aggregate.
So my tentative answer to Jims question: the onus currently lies with
MAS to show that there are systems in which qualitatively different
dynamical behaviour emerges from individual interactions, than we would
expect if we just looked at the aggregate. And my guess is that if we
can show this, then it will have to do with chaos and with aggregates of
individuals which are _not_ identical. For instance, if we looked at a
genetic simulation of propagation of N traits through a population, we
could no longer justify treating the individuals as aggregates because
they are all different from each other in one gene or another. The
interactions between the genes might then lead to surprising aggregate
behaviour. I wont know until I do the experiment (dont hold your
breath - my plates full with my new job at the moment!).
Cheers,
Niall.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>
> So, my question is: what dynamics do agent-based representations capture
> that makes it worth sacrificing the (modest) analytical ease of the
> traditional SD way of representing systems.
Jim, I dont think you can know how good it is to me to see this
question explicitly formulated like this. _This_ is the central thrust
of my interest at the moment - I just wish I had more time to work on
it. I dont know the answer - the following are my thoughts at the
moment.
SD is used to treat not individuals, but aggregates. Thus, in a
simulation of spread of a genetic trait through a population, a MAS
(multi-agent system) simulation would have hundreds (thousands?) of
individuals with assorted traits, and then let them interact, while an
SD simulation would probably dive in with the Hardy-Weinberg equation.
For me the point is that MAS try to explain the aggregate behaviour in
terms of individuals and their random interactions, while SD tends (and
I am _only_ saying "tends") to draw the line of abstraction above the
random level, assuming that the random behaviour can be ignored if we
know the laws governing the overall dynamics of the aggregate.
So my tentative answer to Jims question: the onus currently lies with
MAS to show that there are systems in which qualitatively different
dynamical behaviour emerges from individual interactions, than we would
expect if we just looked at the aggregate. And my guess is that if we
can show this, then it will have to do with chaos and with aggregates of
individuals which are _not_ identical. For instance, if we looked at a
genetic simulation of propagation of N traits through a population, we
could no longer justify treating the individuals as aggregates because
they are all different from each other in one gene or another. The
interactions between the genes might then lead to surprising aggregate
behaviour. I wont know until I do the experiment (dont hold your
breath - my plates full with my new job at the moment!).
Cheers,
Niall.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>
-
Niall Palfreyman
- Senior Member
- Posts: 56
- Joined: Fri Mar 29, 2002 3:39 am
Fast Demonstration of Complex Behavior
Hi,
I just wanted to say thank-you for the helpful tips for a demo last
week. In the end I used the two-people-carrying-an-egg demonstration and
it went down very well - I used it to get into the subject of what
complexity is, and what might be needed to handle it. At the end there
was applause, and on the spur of the moment I used Andy Fords
suggestion of asking them to clap again in unison. I then closed with
the comment that they had just successfully proven themselves capable of
handling complexity. The whole thing was a great success due
prinicipally to the demonstrations you all suggested - many thanks
again.
By the way, the demo with the people walking around I shall use at a
3-hour workshop for the Employment Office in Munich in June, so Im also
very much looking forward to trying out that one.
Cheers,
Niall Palfreyman.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>
I just wanted to say thank-you for the helpful tips for a demo last
week. In the end I used the two-people-carrying-an-egg demonstration and
it went down very well - I used it to get into the subject of what
complexity is, and what might be needed to handle it. At the end there
was applause, and on the spur of the moment I used Andy Fords
suggestion of asking them to clap again in unison. I then closed with
the comment that they had just successfully proven themselves capable of
handling complexity. The whole thing was a great success due
prinicipally to the demonstrations you all suggested - many thanks
again.
By the way, the demo with the people walking around I shall use at a
3-hour workshop for the Employment Office in Munich in June, so Im also
very much looking forward to trying out that one.
Cheers,
Niall Palfreyman.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>