Khalid Saeed wrote:
> A feedback loop must,
> however, have at least one stock in its circular path that decouples flows,
> creates integration and delays.
This is unquestionably true for any system dynamics (SD) feedback loop.
But: IMHO also "non-SD feedback loops" do exist. Let me give an example
from Paul Watzlawicks classic "Pragmatics of Human Communication":
A couple might have the following problem: She always criticises her
husband, because he keeps off her. He keeps off her, because she always
criticises him.
This is unquestionably a reinforcing feedback loop, which actually is
not so seldom among couples - says Watzlawick. Now, where are the
"stocks" and the "flows" in this loop? Of course this situation can be
modeled using the stock-and-flow technology - that is not my point. My
point is that this feedback loop does exist without any stocks and
flows "creating" the dynamic structure. Watzlawick had most probably no
idea about stocks and flows; nevertheless he was able to identify and
describe that feedback loop as a feedback loop.
Another, even more compelling argument for the existence of non-SD
feedback loops is George P. Richardsons great book "Feedback Thought",
which gives a fascinating overview over the emergence of the
feedback concept (and thus feedback loops) in different fields.
Reading that book makes clear that most of the feedback thought originally
was _not_ based on the stock-flow-technology.
This doesnt diminish the power of the SD approach a a very versatile and
mighty tool for describing and analyzing feedback structures; I just think
it is fair to acknowledge that we have feedback concepts beyond SD, too.
Guenther Ossimitz
--
Dr. Guenther OSSIMITZ
University of Klagenfurt
A-9020 Klagenfurt, Univ.str. 65 Austria/Europe
mail: ossimitz@bigfoot.com
home: http://go.just.to/ossimitz
Does feedback need to involve levels
-
Guenther Ossimitz
- Junior Member
- Posts: 17
- Joined: Fri Mar 29, 2002 3:39 am
-
"William Steinhurst"
- Member
- Posts: 21
- Joined: Fri Mar 29, 2002 3:39 am
Does feedback need to involve levels
One needs to distinguish between the types of models being solved.
In many dynamic systems (those involving a time variable), it is
posited that the state of the system evolves in accordance with a
system of differential equations. In such systems of equations,
typically, the partial derivatives of the state variables depend on the
instantaneous and (sometimes) simultaneous values of state
variables or their rates of change. If the system of differential
equations in question can be solved directly, well and good. But that
is often impossible, so we turn to numerical solution via difference
equations. This is what Dynamo, Stella, and other SD modeling
programs do. In this approach the value of each state variable at
time t+Dt is computed as a function of the values of the state
variables at time t, their rates of change between times t-Dt and t, or
simlar higher order backward differences. (Im using capital Dt to
represent a finite "delta t" for lack of greek letters in ASCII.) The
solution for each state variable at time t must be computed this way
(using values at time t to compute values at time t+Dt) because the
values of the other state variables are not known until AFTER the
time step is computed.
Thus, the necessity for banning instantaneous feedback comes not
from SD, but from the mathematics of difference equations and their
numerical solution. IMHO the SD dictum requiring each loop to
include a stock is heuristic that ensures we consistently do forward
solutions of the difference equations in a model. Forcing ourselves
to seek out and identify a stock in every loop may also have
conceptual and logical benefits, but leaving those possible benefits
aside, without that rule, the solutions of the model equations are
prey to hidden defects.
William Steinhurst, Dir. for Regulated Utility Planning
Department of Public Service
112 State St, Drawer 20, Montpelier VT 05620
802 828 4006 Fax 802 828 2342 tty/ttd 1 800 734 8390
wsteinhu@psd.state.vt.us http://www.state.vt.us/psd
In many dynamic systems (those involving a time variable), it is
posited that the state of the system evolves in accordance with a
system of differential equations. In such systems of equations,
typically, the partial derivatives of the state variables depend on the
instantaneous and (sometimes) simultaneous values of state
variables or their rates of change. If the system of differential
equations in question can be solved directly, well and good. But that
is often impossible, so we turn to numerical solution via difference
equations. This is what Dynamo, Stella, and other SD modeling
programs do. In this approach the value of each state variable at
time t+Dt is computed as a function of the values of the state
variables at time t, their rates of change between times t-Dt and t, or
simlar higher order backward differences. (Im using capital Dt to
represent a finite "delta t" for lack of greek letters in ASCII.) The
solution for each state variable at time t must be computed this way
(using values at time t to compute values at time t+Dt) because the
values of the other state variables are not known until AFTER the
time step is computed.
Thus, the necessity for banning instantaneous feedback comes not
from SD, but from the mathematics of difference equations and their
numerical solution. IMHO the SD dictum requiring each loop to
include a stock is heuristic that ensures we consistently do forward
solutions of the difference equations in a model. Forcing ourselves
to seek out and identify a stock in every loop may also have
conceptual and logical benefits, but leaving those possible benefits
aside, without that rule, the solutions of the model equations are
prey to hidden defects.
William Steinhurst, Dir. for Regulated Utility Planning
Department of Public Service
112 State St, Drawer 20, Montpelier VT 05620
802 828 4006 Fax 802 828 2342 tty/ttd 1 800 734 8390
wsteinhu@psd.state.vt.us http://www.state.vt.us/psd
-
shelley friedman
- Newbie
- Posts: 1
- Joined: Fri Mar 29, 2002 3:39 am
Does feedback need to involve levels
It may be that I am only a novice, but the repsonse by Dr. Ossimitz points
to a flaw in CLDs in general. The implied stock is the husband wife
relationship. In most cases, it is either love of the spouse or
significant other. When the wife is critical, the stock of love falls. It
could even be
seen as a table function. The level of new love compared to the
original
level of love effecting the response of the husband. Unfortunately, the
CLD does not always show the stock as explicit.
Shelly Friedman
From: shelley friedman <sfriedma@rh.edu>
to a flaw in CLDs in general. The implied stock is the husband wife
relationship. In most cases, it is either love of the spouse or
significant other. When the wife is critical, the stock of love falls. It
could even be
seen as a table function. The level of new love compared to the
original
level of love effecting the response of the husband. Unfortunately, the
CLD does not always show the stock as explicit.
Shelly Friedman
From: shelley friedman <sfriedma@rh.edu>
-
"George Backus"
- Member
- Posts: 33
- Joined: Fri Mar 29, 2002 3:39 am
Does feedback need to involve levels
Guenther Ossimitz felt that an inter-personal argument had no intervening
levels. As SD theory claims, it does. Many of us who must attempt to solve
client problems, live in an adversarial environment. Whatever we recommend
will change the system conditions and affect some member of that system.
These affected members will produce counter-responses -- often in the form
of derogatory
arguments during executive staff meetings.... If the topic is the lay-off
of staff or the removal of divisions, the "discussions" can be quite emotive
and the consequences very serious.
Novice staff in these discussions have a difficult time. Those who have an
even more severe reaction are those who, for lack of a more direct word,
suffer personality traits dominated by personal insecurity -- they do not
have what is often defined as coping skills. Those in the meeting who are
"battle hardened" have experienced and survived worse situations, they can
step back and view the problem, *cope* with it, and develop solutions
alternatives. They have been in worse situations and the current crisis is
a "piece of cake" compared to some others they have experienced.
The same is true of a spousal argument. The "crisis" developed over time.
The coping skills of both parties are/were inadequate to the initial
conditions and each sortie of banter reduces each participants coping
skills further. A counselor (or some intervener) would allow them to step
back and view the problem. This act alone helps them increase their ability
to cope. If nurtured, over time, the coping skills improve. If a crisis
occurs that is too large for the existing LEVEL of coping skills, then the
coping skill is overwhelmed and a setback occurs -- just like when we work
on a model and hit so many problems that we feel we are useless as an SD
analyst. Yet, when we finally do succeed, we now are more capable of
dealing with even more complex or critical issues. This success and the
increase in coping skills is like a successful resolution of a issue with
the subject couple.
Causality is a sequence of events where the preceding event has an effect,
causing the next event. The sequence by definition takes time. It can NEVER
be simultaneous. Because any portion of reality of interest is a system,
there are, by definition, parts that interact. These interactions are always
feedback loops. They are always causal. They can never be simultaneous. We
can not define history until after the events happen.
(Some may object here ala Quantum Mechanics. Only the Copenhagen
Interpretation avoids causality. There are several competing
interpretations - Pilot wave, Stochastic QM, Zero-point fields, etc.) that
give identical answers in most areas. Therefore, while the Copenhagen
interpretation is mathematically accurate, the most complete (being the
oldest) and therefore the most useful, it is still not clear that a causal
interpretation does not exist. But most importantly, at the macro level,
even QM leads to causality -- the real macro-world we experience and
describe via
classical physics -- and non-simultaneous feedback loops.)
Bottom line is that if you think the (sub) system is a circular loop without
a level, then you need to rethink the logic.
G
George Backus: George_Backus@ENERGY2020.com
levels. As SD theory claims, it does. Many of us who must attempt to solve
client problems, live in an adversarial environment. Whatever we recommend
will change the system conditions and affect some member of that system.
These affected members will produce counter-responses -- often in the form
of derogatory
arguments during executive staff meetings.... If the topic is the lay-off
of staff or the removal of divisions, the "discussions" can be quite emotive
and the consequences very serious.
Novice staff in these discussions have a difficult time. Those who have an
even more severe reaction are those who, for lack of a more direct word,
suffer personality traits dominated by personal insecurity -- they do not
have what is often defined as coping skills. Those in the meeting who are
"battle hardened" have experienced and survived worse situations, they can
step back and view the problem, *cope* with it, and develop solutions
alternatives. They have been in worse situations and the current crisis is
a "piece of cake" compared to some others they have experienced.
The same is true of a spousal argument. The "crisis" developed over time.
The coping skills of both parties are/were inadequate to the initial
conditions and each sortie of banter reduces each participants coping
skills further. A counselor (or some intervener) would allow them to step
back and view the problem. This act alone helps them increase their ability
to cope. If nurtured, over time, the coping skills improve. If a crisis
occurs that is too large for the existing LEVEL of coping skills, then the
coping skill is overwhelmed and a setback occurs -- just like when we work
on a model and hit so many problems that we feel we are useless as an SD
analyst. Yet, when we finally do succeed, we now are more capable of
dealing with even more complex or critical issues. This success and the
increase in coping skills is like a successful resolution of a issue with
the subject couple.
Causality is a sequence of events where the preceding event has an effect,
causing the next event. The sequence by definition takes time. It can NEVER
be simultaneous. Because any portion of reality of interest is a system,
there are, by definition, parts that interact. These interactions are always
feedback loops. They are always causal. They can never be simultaneous. We
can not define history until after the events happen.
(Some may object here ala Quantum Mechanics. Only the Copenhagen
Interpretation avoids causality. There are several competing
interpretations - Pilot wave, Stochastic QM, Zero-point fields, etc.) that
give identical answers in most areas. Therefore, while the Copenhagen
interpretation is mathematically accurate, the most complete (being the
oldest) and therefore the most useful, it is still not clear that a causal
interpretation does not exist. But most importantly, at the macro level,
even QM leads to causality -- the real macro-world we experience and
describe via
classical physics -- and non-simultaneous feedback loops.)
Bottom line is that if you think the (sub) system is a circular loop without
a level, then you need to rethink the logic.
G
George Backus: George_Backus@ENERGY2020.com
-
"Hines, Chip"
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
Does feedback need to involve levels
IMHO,I think that Watzlawicks example does include stocks and flows.
Perhaps they arent physical, but the buildup of resentment until a critical
action occurs is certainly a flow into a stock that results in action when a
threshold is achieved. Likewise the stock of "willingness to keep on
her(?)" is depleted by the wifes actions. I think resentment (and the
like) is certainly a stock, and not just an artifical representation to
enable us to model the system.
chip
From: "Hines, Chip" <Chip.Hines@fema.gov>
Perhaps they arent physical, but the buildup of resentment until a critical
action occurs is certainly a flow into a stock that results in action when a
threshold is achieved. Likewise the stock of "willingness to keep on
her(?)" is depleted by the wifes actions. I think resentment (and the
like) is certainly a stock, and not just an artifical representation to
enable us to model the system.
chip
From: "Hines, Chip" <Chip.Hines@fema.gov>
-
Bill Braun
- Senior Member
- Posts: 73
- Joined: Fri Mar 29, 2002 3:39 am
Does feedback need to involve levels
Hi Guenther,
Your example represents the challenge I had "connecting the dots" between
causal loop diagrams and SD models. Your description is quite accurate if
the dynamic you illustrated is represented in a CL diagram. However, the
ease of CL diagrams (opinion looming on the horizon) can obscure some of
the "hidden" dynamics.
The stock in your example is the current state (i.e., level) of the
husbands dominant perception of his wife. That is the stock from which
feedback contributes to his behavior (the decision to act). The wifes
criticism is the flow into the stock.
In terms of intervention, if the flow into the stock changed (a change in
the wifes behavior) then the level of the stock would correspondingly
change, similarly altering the feedback, leading to changed behavior.
Another way of thinking it, if the system came to rest (no criticism, no
laying off), would the husbands instinctive perception of his wife change.
I would argue, No, he would still be left with the dominant impression of
his last encounter (the level in the stock), which reinforced his
perception of his wife.
Best regards,
Bill Braun
From: Bill Braun <medprac@hlthsys.com>
The Health Systems Group
- Physician Leadership Training
- Simulation Modeling for Healthcare
http://www.hlthsys.com
Your example represents the challenge I had "connecting the dots" between
causal loop diagrams and SD models. Your description is quite accurate if
the dynamic you illustrated is represented in a CL diagram. However, the
ease of CL diagrams (opinion looming on the horizon) can obscure some of
the "hidden" dynamics.
The stock in your example is the current state (i.e., level) of the
husbands dominant perception of his wife. That is the stock from which
feedback contributes to his behavior (the decision to act). The wifes
criticism is the flow into the stock.
In terms of intervention, if the flow into the stock changed (a change in
the wifes behavior) then the level of the stock would correspondingly
change, similarly altering the feedback, leading to changed behavior.
Another way of thinking it, if the system came to rest (no criticism, no
laying off), would the husbands instinctive perception of his wife change.
I would argue, No, he would still be left with the dominant impression of
his last encounter (the level in the stock), which reinforced his
perception of his wife.
Best regards,
Bill Braun
From: Bill Braun <medprac@hlthsys.com>
The Health Systems Group
- Physician Leadership Training
- Simulation Modeling for Healthcare
http://www.hlthsys.com