Locus of Reference Signal

[From Fred Nickols (2005.02.14.0723 EST)] --

I have a question about a comment Bill P made in the Prediction thread (see
below):

>That is what Flach said, and what Forrester and Eberlein said, too.
>I disagree with that idea. I think that reference signals are inside
>the head, not outside in the environment.

I've got some experience working with the SD folks, including Eberlein and
others, including a fellow John Sterman said was the best SD modeler of all.

Those folks are usually grappling with what I call "contrived" systems
(e.g., the dynamics of a business or perhaps some portion of it (e.g., a
process). In contrived systems, the "reference signals" are definitely in
the environment (although they doubtless started out in someone's head).

PCT folks are dealing with what I call "natural" systems (e.g, people). In
natural systems, the reference signals are definitely in the head (and
perhaps in other parts of the body as well. Do muscles have "memory"?)

In any event, as I've watched the PCT/SD dialogue, it's struck me that the
two tackle very different kinds of systems. The control mechanism in a
contrived (i.e., man-made) system is likely to be different from the control
mechanism in a natural system.

More to the point, the SD view of things seems to work for the systems with
which they deal but wouldn't work for individual human beings. I don't know
if the PCT view would work for the same kinds of systems that SD folks
tackle.

Just a thought...

Regards,

Fred Nickols
nickols@att.net

[From Bill Powers (2005.02.14.0707 MST)]

Fred Nickols (2005.02.14.0723 EST) --

Those [SD] folks are usually grappling with what I call "contrived" systems
(e.g., the dynamics of a business or perhaps some portion of it (e.g., a
process). In contrived systems, the "reference signals" are definitely in
the environment (although they doubtless started out in someone's head).

PCT folks are dealing with what I call "natural" systems (e.g, people). In
natural systems, the reference signals are definitely in the head (and
perhaps in other parts of the body as well. Do muscles have "memory"?)

In any event, as I've watched the PCT/SD dialogue, it's struck me that the
two tackle very different kinds of systems. The control mechanism in a
contrived (i.e., man-made) system is likely to be different from the control
mechanism in a natural system.

Actually, if you look at their "contrived" systems you'll see that the loop gain is quite low and disturbances have a large effect (that's been true in the models I've seen, though it may not be true in all). The plots of behavior just sort of wander up and down, the way a plot of barometer readings or stock market prices might do. There is little evidence of anything being controlled to stay at a specific level or to repeat a specific pattern.

The reason is precisely the lack of a controlling agent or agents in these models -- a person with a specific goal acting with high gain to maintain or achieve it. The model I sent Bob had such agents in it, which is the reason that it achieved the goal so quickly and accurately.

That was a case in which the goal was achievable. There are other contrived systems that are not so easy to control, such as the stock market. In those cases there is no human controller who knows what to do to make the (potential) controlled variables behave in a desired way or stay at a desired level, so disturbances have large effects and the plots of behavior wander in an uncontrolled way. Also, when you have interactions among contrived systems, the human controllers at least in a market economy are normally in direct conflict; if my ad campaign works, your sales drop and you have to spend your profits on counteractive ads. And again, control gets worse because a lot of output is used just to cancel the other guy's output.

All the goals in contrived systems are human goals, of course: they were specified by human controllers and are carried out, if possible, by human controllers (though perhaps not the same ones). The design of the system might include control-system features, such as what action to take when certain kinds of errors arise, but where many people are involved, "what to do" is subject to continual reinterpretation as the individuals carry out their own ideas of what is wrong and how to correct it. So individual human controllers always get into the act. Often, in my experience, that is all that makes the so-called business plan work.

Best,

Bill P.

From [Marc Abrams (2005.02.14.1207)]

Fred, I think every ‘system’ that involves people will by necessity be ‘natural’, with some having more or less ‘contrived’ components in them.

In a message dated 2/14/2005 7:40:12 A.M. Eastern Standard Time, nickols@WORLDNET.ATT.NET writes:

[From Fred Nickols (2005.02.14.0723 EST)] –

I have a question about a comment Bill P made in the Prediction thread (see
below):

That is what Flach said, and what Forrester and Eberlein said, too.
I disagree with that idea. I think that reference signals are inside
the head, not outside in the environment.

I’ve got some experience working with the SD folks, including Eberlein and
others, including a fellow John Sterman said was the best SD modeler of all.
This of course is Sterman’s opinion. Others feel differently. Homer is certainly an all-star, but crowning anyone as the ‘best’ in anything is dubious at best

In any event, as I’ve watched the PCT/SD dialogue, it’s struck me that the
two tackle very different kinds of systems. The control mechanism in a
contrived (i.e., man-made) system is likely to be different from the control
mechanism in a natural system.
If you have been following the argument then you should know that SD and PCT are NOT different systems. One, SD is a method of modeling systems. PCT is a theory about a system.

These are quite different things.

More to the point, the SD view of things seems to work for the systems with
which they deal but wouldn’t work for individual human beings. I don’t know
if the PCT view would work for the same kinds of systems that SD folks
tackle.
They each provide a ‘view’ of something different. PCT is trying to show a view of human behavior. SD is a method for modeling ‘continuous’ rather than ‘discrete’ systems.

I hope this helps clarify.

regards,

Marc