The Test requires gradual disturbances, Errata

[From Rick Marken (2003.02.18.0830)]

Bruce Nevin (2003.03.18 10:54 EST)

2. The Pattern of rings and arcs in the Gather/Crowd programs is a
byproduct of a number of individuals each controlling some variable(s) at
about the same values. The pattern itself is not controlled by any
individual. Yet it is a controlled outcome.

Actually, these are not controlled outcomes.

Disturb a ring somehow and you
simultaneously disturb one or more individual's control of proximity. So in
an interesting sense the rings and arcs are collectively controlled.

This is not how you demonstrate that the rings are controlled. You disturb the
rings themselves and see if they are protected from the disturbance. They are not.
Ergo, they are not controlled variables.

Best

Rick

PS. In my previous post the sentence:

If we didn't have the goal of keeping the cursor on target in attacking task we
wouldn't act by move the mouse in exact opposition to the disturbance wave form.

should obviously have read:

If we didn't have the goal of keeping the cursor on target in a tracking task we
wouldn't act by moving the mouse in exact opposition to the disturbance wave form.

···

--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Rick Marken (2003.03.18.1310)]

Bruce Nevin (2003.03.18 13:31 EST)

Rick Marken (2003.02.18.0830)--
>This is not how you demonstrate that the rings are controlled. You disturb the
>rings themselves and see if they are protected from the disturbance. They
>are not. Ergo, they are not controlled variables.

I agree that they are not controlled by any individual agent.

Please specify what you are doing to move the individual agents out of the
ring or arc such that they can easily overcome what you are doing by
actions that are natural in the course of what they are doing.

This disturbance - whatever it is - is also a disturbance to their control
of proximity to the common "target" of all the individuals in the ring or
arc (and probably their control of proximity to one another). How do you
distinguish disturbance to the individual's CV from disturbance of the
collective outcome of their control?

I don't think you have to make this distinction. What you see in the CROWD demo is
a variable in the world, in this case the 2D spatial arrangement of a set of
agents as seen from "above". You are saying that this variable is being
maintained (controlled) in the state "ring". How this control is exerted --
collectively or individually -- is not really relevant until we've tested to
determine whether or not this variable is actually under control. If it is, then
the variable will be maintained in a ring, protected from disturbances.

One way to disturb the ring shape is to mix "fake" agents in with the real agents.
The fake agents aren't controlling distance but they will disturb the positions of
the the real agents, which are controlling their proximity to all objects around
them. If, no matter where the fake agents were located among the real agents, the
real agents (or all the agents, depending on how you defined the purported
controlled variable) always end up in a ring, then there would be evidence that
the ring is, indeed, controlled (and then we would have to figure out how it is
controlled). But we wouldn't find that the ring is controlled (if we did this
test) because we already know that the rings and other 2D patterns we see in the
CROWD demo are not controlled. Indeed, that's one of the main points of the demo,
I think: eye-catching "social patterns" like those seen in the CROWD demo may be
uncontrolled side effects of the fact that individual agents are controlling
perceptions like proximity.

Best

Rick

···

At 08:32 AM 3/18/2003, Richard Marken wrote:

--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bruce Nevin (2003.03.18 13:31 EST)]

Rick Marken (2003.02.18.0830)--

This is not how you demonstrate that the rings are controlled. You disturb the
rings themselves and see if they are protected from the disturbance. They are not. Ergo, they are not controlled variables.

I agree that they are not controlled by any individual agent.

Please specify what you are doing to move the individual agents out of the ring or arc such that they can easily overcome what you are doing by actions that are natural in the course of what they are doing.

This disturbance - whatever it is - is also a disturbance to their control of proximity to the common "target" of all the individuals in the ring or arc (and probably their control of proximity to one another). How do you distinguish disturbance to the individual's CV from disturbance of the collective outcome of their control?

         /Bruce Nevin

···

At 08:32 AM 3/18/2003, Richard Marken wrote:

[From Bill Powers (2003.03.18.2107 MST)]

Bruce Nevin (2003.03.18 13:31 EST)--

>This disturbance - whatever it is - is also a disturbance to their control
>of proximity to the common "target" of all the individuals in the ring or
>arc (and probably their control of proximity to one another). How do you
>distinguish disturbance to the individual's CV from disturbance of the
>collective outcome of their control?

You're on the right track here. The method would be to establish for each
of the follower agents that it is not controlling for the size of the ring
or its position around the perimeter of the ring.

This can be done directly for each agent in turn by seizing all the other
agents as well as the guru and moving them to make a new circle with the
tested relationship (distance and direction from tested agent to guru)
remaining the same. The agent being tested would do nothing to change the
size of the circle, since its reference conditions are still satisfied.
This would prove to be the case for all agents in turn.

More elegantly, with proper analytical tools it would be possible to
measure the resistances of all the agents at the same time to multiple
disturbances established by moving just the passive obstacles in various
patterns. The solution of the n equations thus determined would show that
in each agent there is maximum sensitivity to disturbances of direction and
distance to the guru, maximum sensitivity to the sum of proximity errors to
left and right, and no sensitivity either to changes in the diameter of the
ring or the position around the periphery of the ring.

As in every case of investigating the properties of control systems, it is
essential that disturbances be manipulated (or happen naturally) while
quantitative data are taken. This can be difficult in testing for control
of social variables, but the principle remains the same. It's up to the
experimenter's ingenuity to apply it.

Best,

Bill P.