>
> I am slowly moving toward your point of view. I still cannot resist the
> apparent ease with which students grasp CLDs and their transition to SDMs
> seems relatively painless. However, I am now wondering if going straight to
> SDMs, though it may be slower on the front end, will produce richer results
> on the back end.
My personal experience is this: after having learnt SD in a system dynamics
"modeling" manner, without ever touching CLDs, (from the great teacher Bruce
Hannon on this list), I decided once to use causal-loop diagrams first in my
own teaching, against my better judgment. That was a very bad experience -
it is very hard to teach good SD AFTER the flakiness and lack of rigor of
CLDs (at least it was for me - it will be good to hear from others of their
experiences). Students would try to make SD models as copies of CLDs, would
make bad distinctions between stocks and flows, will forget the high-school
physics principles of dimensional consistency, will try to draw whatever
conclusions they wanted to from the causal-loop diagrams, and then blame
teachers for not doing consistent teaching because some things, using SD
models, had to be re-explained because the corresponding CLDs were
insufficient or plain wrong, etc. That was the start of my thinking about
the tremendous importance of SD modeling BEFORE any CLDs, and the huge
importance of being data-centric instead of insight-centric (CLDs, like bad
use of statistics, can be used to convey anything, whereas it is very
difficult to do that with SD models with good data. If data is considered
not that important, then models may be twisted to convey anything too -
hence my remarks in the previous posts; please dont read this as a plea for
unnecessary details). Used properly, they of course can be great
communication and organizing tools. People with heavy math/engineering
training may want to use more CLDs and people with heavy social science
training may want to use more rigorous SD modeling in their approaches -
just to bring the best of the "other" in their "own" research and thinking.
Else engineers are just too "hard" and sociologists are just too "soft", if
you get my drift.
There is an article somewhere reflecting on the problems with CLDs (I
believe it is GPRs). Bad for me I read it after the above experiment.
Best
Jaideep
jaideep@optimlator.com
http://www.optimlator.com/
Causal loop and Stock and Flow Diagramming
-
"Jaideep Mukherjee"
- Junior Member
- Posts: 15
- Joined: Fri Mar 29, 2002 3:39 am
-
Stephen Wehrenberg
- Member
- Posts: 29
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
Phil Odence wrote:
> My take was that Kim was saying feedback IS as basic as integration, AND
> integration (I prefer accumulation) is as basic as feedback. Stock flow maps
> give us both. CLDs convey part only part of the story.
I think (no pun intended) I agree with the above. I have thought a lot about
the psychology of CLDs vs SF methods. I tend to operate such that in any case
where it is possible that a discussion will lead to modeling I start with an SF
conceptualization. However, I also use CLDs a lot in the course of normal
conversation and day to day crises and issues. Sometimes I guess wrong and
start with a CLD only to find that I need to shift folks to SF ... I just
address the differences, the reasons, the strengths and weaknesses of each, and
start over.
I suppose Im suggesting that there is a place for both views ... but Id be
curious to see how everyone else feels about this. Im a psychologist and
wrestled with the nature-nurture argument all through graduate school ... only
to finally discover that both camps were right in their context -- or more
accurately, I found not a compromise, but a theory that transcended the argument
and encompassed both positions non-competitively. (Since you ask, social
learning theory and social behaviorism ... I think the APA has renamed me as a
"social cognitivist" now, but I dont recall voting on that issue! I believe
the emphasis on reciprocal interaction explains in part my appreciation for
system dynamics. First career in servomechanisms helped too!)
Do YOU ever use CLDs, and if so, under what conditions?
Steve
--
Stephen B. Wehrenberg, Ph.D.
HR Capability Development, US Coast Guard
Administrative Sciences, The George Washington University
wstephen@erols.com
> My take was that Kim was saying feedback IS as basic as integration, AND
> integration (I prefer accumulation) is as basic as feedback. Stock flow maps
> give us both. CLDs convey part only part of the story.
I think (no pun intended) I agree with the above. I have thought a lot about
the psychology of CLDs vs SF methods. I tend to operate such that in any case
where it is possible that a discussion will lead to modeling I start with an SF
conceptualization. However, I also use CLDs a lot in the course of normal
conversation and day to day crises and issues. Sometimes I guess wrong and
start with a CLD only to find that I need to shift folks to SF ... I just
address the differences, the reasons, the strengths and weaknesses of each, and
start over.
I suppose Im suggesting that there is a place for both views ... but Id be
curious to see how everyone else feels about this. Im a psychologist and
wrestled with the nature-nurture argument all through graduate school ... only
to finally discover that both camps were right in their context -- or more
accurately, I found not a compromise, but a theory that transcended the argument
and encompassed both positions non-competitively. (Since you ask, social
learning theory and social behaviorism ... I think the APA has renamed me as a
"social cognitivist" now, but I dont recall voting on that issue! I believe
the emphasis on reciprocal interaction explains in part my appreciation for
system dynamics. First career in servomechanisms helped too!)
Do YOU ever use CLDs, and if so, under what conditions?
Steve
--
Stephen B. Wehrenberg, Ph.D.
HR Capability Development, US Coast Guard
Administrative Sciences, The George Washington University
wstephen@erols.com
-
"Jay W. Forrester"
- Senior Member
- Posts: 63
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
I have seen students waste much time trying to define a system in
causal-loop diagrams.
In my own work, I have never used causal loops to start a project.
Instead, I decide what system levels (integrations) are involved in the
behavior of interest. Then establish the flows to and from those levels,
and finally work on how the levels in the system control all of the flows.
I have used causal loops after a model is complete for quickly explaining
some of the structural concepts to an audience that wants a 20-minute
introduction to the nature of the system structure and who will be doing
nothing with the model or the ideas after the lecture in finished.
In this discussion thread, there seems to be some confusion about the
relative importance of integration and feedback loops. They go together.
A feedback loop without integration is, at the best, a simultaneous
equation without dynamic behavior. Integration without a feedback loop
means that the flows into and out of the level are not controlled by the
level, which will almost always lead to unrealistic behavior.
Has everyone paused to think about why the closed-form solutions to
differential equations are always in the form of exponentials, either real
exponentials or sine functions? These are the only mathematical forms that
can circulate around a feedback loop and retain their characteristic
time-shape. Differential equations describe feedback loops, but are a
confusing backward way of doing so. The concept of a derivitive is a
figment of mathematicians imaginations. No where in nature or human
affairs is a derivitive taken. Nature only integrates. Starting dynamics
from the viewpoint of integration and feedback loops puts the ideas within
the reach and experience of K-12 students. Any child that can fill a water
glass or take toys away from a playmate knows what accumulation (or
integration) is. Dynamics approached from the differential equation
viewpoint means that the ideas must be postponed until reaching a higher
level of mathematical complexity to which most people never arrive.
Looking at dynamics through differential equations produces in many
students a reversed sense of causality. I have had students argue that
saying water flows out of the faucet because the water in the glass is
rising is just at reasonable as saying that the water is rising because
water is flowing into the glass. Of course, the full feedback loop runs
from faucet to flow to rising water level to eye to hand to control of the
faucet. The flow causes the water to rise. The water level, not the rate
of rise, controls the faucet. Anyone who wants to argue that the rate of
rise of the water is to be taken into account must then introduce the
additional integration that permits such a perception.
---------------------------------------------------------
Jay W. Forrester <jforestr@MIT.EDU>
Professor of Management
Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
tel: 617-253-1571
fax: 617-258-9405
causal-loop diagrams.
In my own work, I have never used causal loops to start a project.
Instead, I decide what system levels (integrations) are involved in the
behavior of interest. Then establish the flows to and from those levels,
and finally work on how the levels in the system control all of the flows.
I have used causal loops after a model is complete for quickly explaining
some of the structural concepts to an audience that wants a 20-minute
introduction to the nature of the system structure and who will be doing
nothing with the model or the ideas after the lecture in finished.
In this discussion thread, there seems to be some confusion about the
relative importance of integration and feedback loops. They go together.
A feedback loop without integration is, at the best, a simultaneous
equation without dynamic behavior. Integration without a feedback loop
means that the flows into and out of the level are not controlled by the
level, which will almost always lead to unrealistic behavior.
Has everyone paused to think about why the closed-form solutions to
differential equations are always in the form of exponentials, either real
exponentials or sine functions? These are the only mathematical forms that
can circulate around a feedback loop and retain their characteristic
time-shape. Differential equations describe feedback loops, but are a
confusing backward way of doing so. The concept of a derivitive is a
figment of mathematicians imaginations. No where in nature or human
affairs is a derivitive taken. Nature only integrates. Starting dynamics
from the viewpoint of integration and feedback loops puts the ideas within
the reach and experience of K-12 students. Any child that can fill a water
glass or take toys away from a playmate knows what accumulation (or
integration) is. Dynamics approached from the differential equation
viewpoint means that the ideas must be postponed until reaching a higher
level of mathematical complexity to which most people never arrive.
Looking at dynamics through differential equations produces in many
students a reversed sense of causality. I have had students argue that
saying water flows out of the faucet because the water in the glass is
rising is just at reasonable as saying that the water is rising because
water is flowing into the glass. Of course, the full feedback loop runs
from faucet to flow to rising water level to eye to hand to control of the
faucet. The flow causes the water to rise. The water level, not the rate
of rise, controls the faucet. Anyone who wants to argue that the rate of
rise of the water is to be taken into account must then introduce the
additional integration that permits such a perception.
---------------------------------------------------------
Jay W. Forrester <jforestr@MIT.EDU>
Professor of Management
Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
tel: 617-253-1571
fax: 617-258-9405
-
Bryan.A.James@gd-is.com
- Junior Member
- Posts: 3
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
Jay Forrester writes:
>>> The concept of a derivitive is a figment of mathematicians imaginations.
>>> No where in nature or human affairs is a derivitive taken. Nature only integrates.
What!? I guess I should throw out most of what I learned about electromagnetics
in my geophysical career. Im going to go ponder Maxwells Equations and see if
Del cross ... and Del dot ... mean something different than what I learned.
I think we should be careful about such broad sweeping statements. There are many
useful ways to view the world and each way is very useful in its context and, when
you get right down to it, they are all constructs of the human mind. To single one
out as inherently the right way to think about something and declare that any
other is inherently wrong suggests a one tool fits all problems perspective.
I dont believe that is going to help SD in the long run.
Bryan James
From: Bryan.A.James@gd-is.com
General Dynamics Information Systems
Denver, CO
>>> The concept of a derivitive is a figment of mathematicians imaginations.
>>> No where in nature or human affairs is a derivitive taken. Nature only integrates.
What!? I guess I should throw out most of what I learned about electromagnetics
in my geophysical career. Im going to go ponder Maxwells Equations and see if
Del cross ... and Del dot ... mean something different than what I learned.
I think we should be careful about such broad sweeping statements. There are many
useful ways to view the world and each way is very useful in its context and, when
you get right down to it, they are all constructs of the human mind. To single one
out as inherently the right way to think about something and declare that any
other is inherently wrong suggests a one tool fits all problems perspective.
I dont believe that is going to help SD in the long run.
Bryan James
From: Bryan.A.James@gd-is.com
General Dynamics Information Systems
Denver, CO
-
"Jim Hines"
- Senior Member
- Posts: 88
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
Stephen Wehrenberg asks us
>> Do YOU ever use CLDs, and if so, under what conditions?
Answering for me: Yes. In consulting, I usually start with causal loop
diagrams before going on to stock and flows. The exception is when I see
immediately a very clear and important stock and flow structure (the iThink
folks might call this a "main chain") in which case, I might dive into the
stock and flow right from the start.
In teaching the SD applications course here at MIT, we encourage students to
start with causal loop diagrams. One reason for this is that students who
start with stocks and flows often never complete any important feedback
loops.
Other reasons to start with causal loop diagrams include:
1. CLDs are usually more dramatic and hence capture the interest of
students and clients alike (its good to start with a bang).
2. Causal loop diagrams lead to insights on their own more frequently than
stock and flow diagrams do. (Note, I am distinguishing between stock and
flow diagrams and the simulation model).
3. Causal loops are easy to develop at a relatively high level of
abstraction - this means that they can provide an overview of the system you
are modeling, before getting down to the nitty gritty.
4. Causal loop diagrams are fuzzier, so they can be drawn even if you are
not yet clear on every single concept (this is a common state at the
beginning of the project).
5. Causal loop diagrams are cheap relative to simulation modeling (and cheap
relative to an equation-level stock and flow diagram). This means you can
ore quickly get a comprehensive feel for the problem area. And
inexpensively generate some initial insights.
Regards,
Jim Hines
jhines@mit.edu
>> Do YOU ever use CLDs, and if so, under what conditions?
Answering for me: Yes. In consulting, I usually start with causal loop
diagrams before going on to stock and flows. The exception is when I see
immediately a very clear and important stock and flow structure (the iThink
folks might call this a "main chain") in which case, I might dive into the
stock and flow right from the start.
In teaching the SD applications course here at MIT, we encourage students to
start with causal loop diagrams. One reason for this is that students who
start with stocks and flows often never complete any important feedback
loops.
Other reasons to start with causal loop diagrams include:
1. CLDs are usually more dramatic and hence capture the interest of
students and clients alike (its good to start with a bang).
2. Causal loop diagrams lead to insights on their own more frequently than
stock and flow diagrams do. (Note, I am distinguishing between stock and
flow diagrams and the simulation model).
3. Causal loops are easy to develop at a relatively high level of
abstraction - this means that they can provide an overview of the system you
are modeling, before getting down to the nitty gritty.
4. Causal loop diagrams are fuzzier, so they can be drawn even if you are
not yet clear on every single concept (this is a common state at the
beginning of the project).
5. Causal loop diagrams are cheap relative to simulation modeling (and cheap
relative to an equation-level stock and flow diagram). This means you can
ore quickly get a comprehensive feel for the problem area. And
inexpensively generate some initial insights.
Regards,
Jim Hines
jhines@mit.edu
-
Tom Fiddaman
- Senior Member
- Posts: 55
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
I think Bryans referring to Maxwells third and fourth equations, which
relate electric and magnetic fields to the rates of change of magnetic or
electric fluxes. This shows up most often when talking about inductors,
which appear in simple circuit equations with a derivative of current, e.g.
V = i*R + di/dt*L (i = current, R = resistance, L = inductance, V = voltage)
This is certainly very convenient, particularly for solving with Laplace
transforms, but Ive always suspected that there was a more intuitive but
possibly less convenient integral explanation of the same behavior. The
units for voltage (work per unit charge) and the fact that energy is
accumulated in a magnetic field kind of suggest this to me, but Ive never
actually looked into it.
Circuits aside, I live by Jays comment. Sometimes its convenient to
assume that a flow is instantaneously perceived, and I occasionally create
equations that take differences over finite time intervals to mimic
accounting outputs like revenue growth, but these work by accumulation. I
cant think of a single example of an instantaneous derivative in a
business or public policy model that wasnt a formulation error or couldnt
be replaced by a better integrative alternative.
I think I once heard Jay say something to the effect that one should speak
in black and white because people listen in shades of gray. This may ruffle
some feathers on the borders of the field, but repeating the mantra "nature
only integrates" sure does help novices avoid some obvious mistakes.
Regards,
Tom
****************************************************
Thomas Fiddaman, Ph.D.
Ventana Systems http://www.vensim.com
8105 SE Nelson Road Tel (253) 851-0124
Olalla, WA 98359 Fax (253) 851-0125
Tom@Vensim.com http://home.earthlink.net/~tomfid
****************************************************
relate electric and magnetic fields to the rates of change of magnetic or
electric fluxes. This shows up most often when talking about inductors,
which appear in simple circuit equations with a derivative of current, e.g.
V = i*R + di/dt*L (i = current, R = resistance, L = inductance, V = voltage)
This is certainly very convenient, particularly for solving with Laplace
transforms, but Ive always suspected that there was a more intuitive but
possibly less convenient integral explanation of the same behavior. The
units for voltage (work per unit charge) and the fact that energy is
accumulated in a magnetic field kind of suggest this to me, but Ive never
actually looked into it.
Circuits aside, I live by Jays comment. Sometimes its convenient to
assume that a flow is instantaneously perceived, and I occasionally create
equations that take differences over finite time intervals to mimic
accounting outputs like revenue growth, but these work by accumulation. I
cant think of a single example of an instantaneous derivative in a
business or public policy model that wasnt a formulation error or couldnt
be replaced by a better integrative alternative.
I think I once heard Jay say something to the effect that one should speak
in black and white because people listen in shades of gray. This may ruffle
some feathers on the borders of the field, but repeating the mantra "nature
only integrates" sure does help novices avoid some obvious mistakes.
Regards,
Tom
****************************************************
Thomas Fiddaman, Ph.D.
Ventana Systems http://www.vensim.com
8105 SE Nelson Road Tel (253) 851-0124
Olalla, WA 98359 Fax (253) 851-0125
Tom@Vensim.com http://home.earthlink.net/~tomfid
****************************************************
-
"Raymond T. Joseph"
- Junior Member
- Posts: 11
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
Since I like electronic circuits, Ill take this path.
Capacitors and inductors can be modeled with either differential or integral
relationships. One chooses the relationship according to the intent. If
the intent is to model the circuit, the differential relationship is usually
chosen so the system can be cast in a state space format. Since the state
space format is just a system of differential equations, the simulation is
just an integration.
Numerically, integration is stable. If one were to chose a system of
equations requiring differentiation instead, the equations may still be
solvable, but numerically, there may be significant obstacles.
As modelers, it is our choice how to setup a problem. Casting everything as
differential equations allows us to apply a common solution method which is
robust. But it is our choice, independent of how nature formulates the
problem and executes its solution.
Ray
From: "Raymond T. Joseph" <rjoseph@wt.net>
Capacitors and inductors can be modeled with either differential or integral
relationships. One chooses the relationship according to the intent. If
the intent is to model the circuit, the differential relationship is usually
chosen so the system can be cast in a state space format. Since the state
space format is just a system of differential equations, the simulation is
just an integration.
Numerically, integration is stable. If one were to chose a system of
equations requiring differentiation instead, the equations may still be
solvable, but numerically, there may be significant obstacles.
As modelers, it is our choice how to setup a problem. Casting everything as
differential equations allows us to apply a common solution method which is
robust. But it is our choice, independent of how nature formulates the
problem and executes its solution.
Ray
From: "Raymond T. Joseph" <rjoseph@wt.net>
-
"Peter Heffron"
- Junior Member
- Posts: 3
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
I really like what Raymond is saying. I only question, "As modelers, it is
our choice how to set up a problem."
Every modeler has a different frame-of-reference. Every modeler is under
different pressures. Most modelers probably arent exclusively modelers. As
Einstein said, "Everything is relative." As Buddha said, "Life is illusion."
Thus (especially) as modelers we instinctively know that (a) since we are
part of the system we are modeling and (b) since no matter how hard we try,
we are mostly subjective beings--our so-called "choice" in setting up a
problem is dicey at best.
This is not meant as holier-than-thou invective. Raymond sparked these
reflections, and I sincerely appreciate his observations for that reason. I
dont have a real answer or alternative, except perhaps that as modelers we
might:
1. Recognize that none of us know the best way to set up a problem
2. Reference and imitate to the degree possible, as Raymond points to,
problem-solving methods that have proven themselves in electronics,
mathematics, and, I would submit--physics, ecology, and other disciplines.
3. As insurance against our own prejudices, involve people in setting up a
problem who we know have different perspectives from our own
4. Set up problems more along the lines of "fuzzy logic" --where "true" is
in the range 0 to 1 (Bart Kosko, et.al.) rather than "absolute
truths" --where true is only 0 or 1 (the operative terms being "0 TO 1"
versus "0 OR 1.")
Special thanks to Raymond,
-Peter
Peter Heffron
95 West Naauao Street
Hilo, Hawaii 96720 USA
E-Mail: heffron@hialoha.net
Telephone: 808-934-0527
our choice how to set up a problem."
Every modeler has a different frame-of-reference. Every modeler is under
different pressures. Most modelers probably arent exclusively modelers. As
Einstein said, "Everything is relative." As Buddha said, "Life is illusion."
Thus (especially) as modelers we instinctively know that (a) since we are
part of the system we are modeling and (b) since no matter how hard we try,
we are mostly subjective beings--our so-called "choice" in setting up a
problem is dicey at best.
This is not meant as holier-than-thou invective. Raymond sparked these
reflections, and I sincerely appreciate his observations for that reason. I
dont have a real answer or alternative, except perhaps that as modelers we
might:
1. Recognize that none of us know the best way to set up a problem
2. Reference and imitate to the degree possible, as Raymond points to,
problem-solving methods that have proven themselves in electronics,
mathematics, and, I would submit--physics, ecology, and other disciplines.
3. As insurance against our own prejudices, involve people in setting up a
problem who we know have different perspectives from our own
4. Set up problems more along the lines of "fuzzy logic" --where "true" is
in the range 0 to 1 (Bart Kosko, et.al.) rather than "absolute
truths" --where true is only 0 or 1 (the operative terms being "0 TO 1"
versus "0 OR 1.")
Special thanks to Raymond,
-Peter
Peter Heffron
95 West Naauao Street
Hilo, Hawaii 96720 USA
E-Mail: heffron@hialoha.net
Telephone: 808-934-0527
-
"Guenther Ossimitz"
- Junior Member
- Posts: 7
- Joined: Fri Mar 29, 2002 3:39 am
Causal loop and Stock and Flow Diagramming
Dear SD-Community,
This fascinating thread brings up some very interesting questions:
1) What is the relation between the idea of feedback and the idea of accumulation?
2) How are these concepts related to the issues of causal loop diagrams (CLD) and
the system dynamics modeling (SDM) method?
3) What is the relation between the idea of accumulation and mathematical integration?
Just a few spontaneous ideas concerning these questions:
A) When looking through George Richardsons excellent book "Feedback Thought",
one can find that the notion of feedback appears in many contexts where
the idea of accumulation is not so central or completely unimportant:
e.g. feedback in techical devices like noise-reducting amplifiers or
James Watts classical homeostat.
B) For SDM the idea of accumulation is essential, whereas the idea of feedback
is just optional. Most SD models have feedback loops; but is not necessary
for a SD model to contain a feedback loop. There are a number of
constellations, where the accumulat ion idea is much more important than
the feedback idea: e.g. a promotional system in an organization. Just imagine
an university faculty with assistant professors, associate professors and full
professors. If promotion is just a matter of qualification and does not
depend upon the number of members in the higher groups, we might have no
feedback loop in such a system. Another example where accumulation concept
is much more important than the feedback idea are sequences of geologic
formations (as in Peter Vescuscos chapter 19 of the classical STELLA
Academic User Guide of 1987.)
C) For CLD technology neither feedback loops nor accumulation is essential.
(BTW: the word "loop" in CLDs is a bit misleading, since this technology
works for situations without any closed loops, too. In German CLDs are called
"Wirkungsdiagramme" [=inf luential diagrams], which prevents this problem.)
It is easily possible to identify closed loops in CLDs. But it is almost
impossible to make the idea of accumulation clear in the causal loop diagramming
technique. On the other hand it is not so easy to see causal loops in a system
dynamics flow diagram in any case. Just imagine a simple radioactive decay
system with a single stock of a radioactive substance that is depleted over
time (i.e. there is just one outflow radiation wich depends upon the amount
of the stock.) It is not trivial to "see" the feedback loop in this
flow diagram, because the (double-lined) arrow indicating physical outflow
of the radiation from the radioactive stock has the opposite direction than
the logical influence from the radiation to the stock.
D) For SDM quantification of all system variables is essential,
for CLDs it is not. CLDs might contain quantified variables,
but they dont necessarily need so.
E) The concept of accumulation necessarily implies at least the possibility
of quantification, whereas the notion of feedback does not. Feedback might
very well appear in quantitative settings, but quantification is not essential
for feedback. To imagine
some accumulation without (at least implicit) quantification seems impossible
to me.
F) The points already given imply that wherever quantification is problematic, so
is the concept of accumulation and the SDM method. I think that the opposite is
true, too: wherever we have a situation which can be quantified naturally with
variables varying over time, the SDM method is a hot candidate when choosing
a modeling technique.
G) This brings me to the bottom line: I would categorize both the feedback idea
and the CLD sketching technique into the qualitative realm, whereas the idea of
accumulation and the SDM method both belong to the quantitative realm. As a
consequence of this I would think as a rule of thumb that the CLD method might
be be the most versatile diagramming technique if the idea of feedback is the
focus of attention, whereas the SDM method is clearly favorable over CLDs
wherever the concept of accumulation plays the most important role.
H) Just a last point: I think that the mathematical concept of integration
is more general than the concept of accumulation, as it appears in SDM.
Accumulation is an integration over time, whereas mathematical integration
is possible over any kind of measurable sets: e.g. integrals over
areas are quite different from SD stock variables. Nevertheless the SD
accumulation concept is a very important protoype of mathematical
integration; it just does not cover the full scope of mathematical integration, as
it is understood in measurement theory and subsequent integration theory.
Guenther Ossimitz
---
Dr. Guenther Ossimitz
From: "Guenther Ossimitz" <guenther.ossimitz@uni-klu.ac.at>
University of Klagenfurt, Univ.str. 65
A-9020 Klagenfurt, Austria
guenther.ossimitz@uni-klu.ac.at
http://go.just.to/go
This fascinating thread brings up some very interesting questions:
1) What is the relation between the idea of feedback and the idea of accumulation?
2) How are these concepts related to the issues of causal loop diagrams (CLD) and
the system dynamics modeling (SDM) method?
3) What is the relation between the idea of accumulation and mathematical integration?
Just a few spontaneous ideas concerning these questions:
A) When looking through George Richardsons excellent book "Feedback Thought",
one can find that the notion of feedback appears in many contexts where
the idea of accumulation is not so central or completely unimportant:
e.g. feedback in techical devices like noise-reducting amplifiers or
James Watts classical homeostat.
B) For SDM the idea of accumulation is essential, whereas the idea of feedback
is just optional. Most SD models have feedback loops; but is not necessary
for a SD model to contain a feedback loop. There are a number of
constellations, where the accumulat ion idea is much more important than
the feedback idea: e.g. a promotional system in an organization. Just imagine
an university faculty with assistant professors, associate professors and full
professors. If promotion is just a matter of qualification and does not
depend upon the number of members in the higher groups, we might have no
feedback loop in such a system. Another example where accumulation concept
is much more important than the feedback idea are sequences of geologic
formations (as in Peter Vescuscos chapter 19 of the classical STELLA
Academic User Guide of 1987.)
C) For CLD technology neither feedback loops nor accumulation is essential.
(BTW: the word "loop" in CLDs is a bit misleading, since this technology
works for situations without any closed loops, too. In German CLDs are called
"Wirkungsdiagramme" [=inf luential diagrams], which prevents this problem.)
It is easily possible to identify closed loops in CLDs. But it is almost
impossible to make the idea of accumulation clear in the causal loop diagramming
technique. On the other hand it is not so easy to see causal loops in a system
dynamics flow diagram in any case. Just imagine a simple radioactive decay
system with a single stock of a radioactive substance that is depleted over
time (i.e. there is just one outflow radiation wich depends upon the amount
of the stock.) It is not trivial to "see" the feedback loop in this
flow diagram, because the (double-lined) arrow indicating physical outflow
of the radiation from the radioactive stock has the opposite direction than
the logical influence from the radiation to the stock.
D) For SDM quantification of all system variables is essential,
for CLDs it is not. CLDs might contain quantified variables,
but they dont necessarily need so.
E) The concept of accumulation necessarily implies at least the possibility
of quantification, whereas the notion of feedback does not. Feedback might
very well appear in quantitative settings, but quantification is not essential
for feedback. To imagine
some accumulation without (at least implicit) quantification seems impossible
to me.
F) The points already given imply that wherever quantification is problematic, so
is the concept of accumulation and the SDM method. I think that the opposite is
true, too: wherever we have a situation which can be quantified naturally with
variables varying over time, the SDM method is a hot candidate when choosing
a modeling technique.
G) This brings me to the bottom line: I would categorize both the feedback idea
and the CLD sketching technique into the qualitative realm, whereas the idea of
accumulation and the SDM method both belong to the quantitative realm. As a
consequence of this I would think as a rule of thumb that the CLD method might
be be the most versatile diagramming technique if the idea of feedback is the
focus of attention, whereas the SDM method is clearly favorable over CLDs
wherever the concept of accumulation plays the most important role.
H) Just a last point: I think that the mathematical concept of integration
is more general than the concept of accumulation, as it appears in SDM.
Accumulation is an integration over time, whereas mathematical integration
is possible over any kind of measurable sets: e.g. integrals over
areas are quite different from SD stock variables. Nevertheless the SD
accumulation concept is a very important protoype of mathematical
integration; it just does not cover the full scope of mathematical integration, as
it is understood in measurement theory and subsequent integration theory.
Guenther Ossimitz
---
Dr. Guenther Ossimitz
From: "Guenther Ossimitz" <guenther.ossimitz@uni-klu.ac.at>
University of Klagenfurt, Univ.str. 65
A-9020 Klagenfurt, Austria
guenther.ossimitz@uni-klu.ac.at
http://go.just.to/go