In a message dated 8/11/00 7:31:18 AM Eastern Daylight Time,
Gustavo.SalaEspiell@petroken-pesa.com.ar writes:
<< I need the help of the community to continue explaining this person (I
used Stermans new book explanation of soft variables and Forresters System
Dynamics and the Lesson of 35 years article on sources of information) the
differences between the mentioned scene in the film and the SD point of view
quantifying soft data.
>>
Gustavo -
Great question about quantifying soft variables, and I think, one of the
underlying questions about modeling. It sets the tension between intuitive
(feel the poem) and analytic (explain and quantify) thinking.
The only set of reasoning that I have found to be effective in an
organization setting runs something like this.
Yes, individuals with experience can understand what is happening through
intuition. But it is very challenging to make informed group decisions
without the ability for each person to explicitly described what they think
happening and why. The modeling process is a why for groups to do this.
Framing it this way says that qauntification is not necessary for a single
person making decision (particularly a short term decision) but it is a good
tool for group decision making. This then does not challenge the individual,
but asks them to be clear for the same of the group decision making....
I had never make the connection to dead poets society, but I think it is a
terrific question to ask of system dynamics, and one that still bothers me to
some extent!
don seville
-------------------------------------
Don Seville
85 Brattle Street
Arlington, MA 02474
781 648 3563
781 658 2010 (fax)
dseville@aol.com
Quantifying soft data
-
Arlen Wolpert
- Junior Member
- Posts: 12
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Gustavo,
For the past 16 years I have used soft variables
and system dynamics to model and simulate
consciousness during a 10-hour religious experience.
It works, because it has helped me to gain some
comprehension of the experience.
Ill grant that a system dynamics analysis can never
do full justice to a poetic idea or a religious
experience. Rather, my approach is to use system
dynamics as a bootstrapping operation to get some
structure for the experience. This allows my mind
to think at a subtler and deeper level about the
experience. From this I advance toward the valuable
insights I am searching for. It is the way an engineer
approaches a problem.
Arlen Wolpert
Independent Scholar
411 Franklin Street, Apt. 1008
Cambridge, MA 02139 USA
Telephone: (617)547-6994
email: awolpert@world.std.com
url: http://world.std.com/~awolpert
For the past 16 years I have used soft variables
and system dynamics to model and simulate
consciousness during a 10-hour religious experience.
It works, because it has helped me to gain some
comprehension of the experience.
Ill grant that a system dynamics analysis can never
do full justice to a poetic idea or a religious
experience. Rather, my approach is to use system
dynamics as a bootstrapping operation to get some
structure for the experience. This allows my mind
to think at a subtler and deeper level about the
experience. From this I advance toward the valuable
insights I am searching for. It is the way an engineer
approaches a problem.
Arlen Wolpert
Independent Scholar
411 Franklin Street, Apt. 1008
Cambridge, MA 02139 USA
Telephone: (617)547-6994
email: awolpert@world.std.com
url: http://world.std.com/~awolpert
-
ularoch@attglobal.net
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
hi,
using soft data in business modelling.
apart from marketig our presently restricted experience is: use hard data
ÿand models only and treat soft arguments in negociating with the hard
ÿsimulated evidence.
yours sincerely
ulrich la roche
la roche consulting
heilighüsli 18, CH-8053 Zuerich,
fax +411 382 1349
From: ularoch@attglobal.net
using soft data in business modelling.
apart from marketig our presently restricted experience is: use hard data
ÿand models only and treat soft arguments in negociating with the hard
ÿsimulated evidence.
yours sincerely
ulrich la roche
la roche consulting
heilighüsli 18, CH-8053 Zuerich,
fax +411 382 1349
From: ularoch@attglobal.net
-
Bruce Campbell
- Junior Member
- Posts: 9
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Sala,
In my limited experience, the best way to go about this is to involve
key stakeholders in the modelling process. This is sometimes very
difficult - they usually try to find reasons why they dont need to be
there.
Emphasise to participants that what they are trying to achieve is an
understanding of why the system behaves in a given way. They are not
trying to predict what will happen in the future. They are, however,
trying to gain an understanding of how the system is likely to behave in
the future.
Just limiting, and defining, the problem can take some time and cause a
great deal of "discussion". I dont know how much experience youve had,
but make sure you are modelling the problem, not the system. Trust me,
you will only ever model the system once. It is an experience you will
never wish to repeat. It is very important to define the problem and
stick to it.
During the group modelling sessions take the time to get participants to
agree on which are the few important variables - those that actually
affect the problem you are modelling. These can be either hard or soft
variables. This, again, is not easy. Everyone will want to include their
pet variables.
Quantifying, not measuring, soft models is often quite easy. For
example, the range of possible values could be somehwere between zero
and one. Get the person with the most direct experience with the
particular soft variable to give his idea of where, on the continuum,
the value lies. Emphasise that ignoring a soft variable, because you
cant accurately measure it, assumes that its value is zero - about the
only value it cant have (a marvellous observation by Forrester 1961,
Industrial Dynamics, p. 57).
My last recommendation - try to choose problems that have no, or very
few, soft variables to start with. That way you can get some points on
the board with your company with relatively little dissension. Then
tackle the more difficult problems as the company gains confidence in
SD.
Bruce Campbell
--
Bruce Campbell
Joint Research Centre for Advanced Systems Engineering
Division of Information and Communication Sciences
Macquarie University 2109
Australia
E-mail: Bruce.Campbell@mq.edu.au
Ph: +61 2 9850 9107
Fax: +61 2 9850 9102
In my limited experience, the best way to go about this is to involve
key stakeholders in the modelling process. This is sometimes very
difficult - they usually try to find reasons why they dont need to be
there.
Emphasise to participants that what they are trying to achieve is an
understanding of why the system behaves in a given way. They are not
trying to predict what will happen in the future. They are, however,
trying to gain an understanding of how the system is likely to behave in
the future.
Just limiting, and defining, the problem can take some time and cause a
great deal of "discussion". I dont know how much experience youve had,
but make sure you are modelling the problem, not the system. Trust me,
you will only ever model the system once. It is an experience you will
never wish to repeat. It is very important to define the problem and
stick to it.
During the group modelling sessions take the time to get participants to
agree on which are the few important variables - those that actually
affect the problem you are modelling. These can be either hard or soft
variables. This, again, is not easy. Everyone will want to include their
pet variables.
Quantifying, not measuring, soft models is often quite easy. For
example, the range of possible values could be somehwere between zero
and one. Get the person with the most direct experience with the
particular soft variable to give his idea of where, on the continuum,
the value lies. Emphasise that ignoring a soft variable, because you
cant accurately measure it, assumes that its value is zero - about the
only value it cant have (a marvellous observation by Forrester 1961,
Industrial Dynamics, p. 57).
My last recommendation - try to choose problems that have no, or very
few, soft variables to start with. That way you can get some points on
the board with your company with relatively little dissension. Then
tackle the more difficult problems as the company gains confidence in
SD.
Bruce Campbell
--
Bruce Campbell
Joint Research Centre for Advanced Systems Engineering
Division of Information and Communication Sciences
Macquarie University 2109
Australia
E-mail: Bruce.Campbell@mq.edu.au
Ph: +61 2 9850 9107
Fax: +61 2 9850 9102
-
"Guenther Ossimitz"
- Junior Member
- Posts: 7
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Dear Mr. SalaEspiell,
thank you for your very kind mail. The question you arise is really worth
to be considered seriously. In my humble opinion "hard" System Dynamics
(SD) is essentially a quantitative methodology.
Nevertheless SDers use their tool sometimes in pretty a qualitative manner.
This might be criticised and your comparision of the scene in the "Dead
poets society" film (which is really excellent IMHO) is not so far fetched.
But please let us be fair: when our students get marks from 1 - 5
(in Austria) or Ice Dancers get notes of 5.7 or 5.9 for their performance
or a 105 score at an intelligence test: isnt that a very similar kind of thing?
The scene in the film gives a certain possibility of discussing the importance
of a poem by trying to evaluate ist "perfection" or "importance". Just asking:
"When should such a measure be zero?" or "What does one unit mean?" opens a
field of interesting investigation.
I think the following: SD works best were actual stocks and flows are involved,
as in production lines, delivery chains, stocks of workforce, capital or whatever.
In these fields quantification is quite natural. When quantification becomes
problematic, the SD method becomes questionable, too. One can still use it,
but one should be aware of the weak numerical basis of the numbers.
Greetings
GO
---
Dr. Guenther Ossimitz
University of Klagenfurt, Univ.str. 65
A-9020 Klagenfurt, Austria
guenther.ossimitz@uni-klu.ac.at
thank you for your very kind mail. The question you arise is really worth
to be considered seriously. In my humble opinion "hard" System Dynamics
(SD) is essentially a quantitative methodology.
Nevertheless SDers use their tool sometimes in pretty a qualitative manner.
This might be criticised and your comparision of the scene in the "Dead
poets society" film (which is really excellent IMHO) is not so far fetched.
But please let us be fair: when our students get marks from 1 - 5
(in Austria) or Ice Dancers get notes of 5.7 or 5.9 for their performance
or a 105 score at an intelligence test: isnt that a very similar kind of thing?
The scene in the film gives a certain possibility of discussing the importance
of a poem by trying to evaluate ist "perfection" or "importance". Just asking:
"When should such a measure be zero?" or "What does one unit mean?" opens a
field of interesting investigation.
I think the following: SD works best were actual stocks and flows are involved,
as in production lines, delivery chains, stocks of workforce, capital or whatever.
In these fields quantification is quite natural. When quantification becomes
problematic, the SD method becomes questionable, too. One can still use it,
but one should be aware of the weak numerical basis of the numbers.
Greetings
GO
---
Dr. Guenther Ossimitz
University of Klagenfurt, Univ.str. 65
A-9020 Klagenfurt, Austria
guenther.ossimitz@uni-klu.ac.at
-
"geoff coyle"
- Senior Member
- Posts: 94
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
>using soft data in business modelling.
>apart from marketig our presently restricted experience is: use hard data
>and models only and treat soft arguments in negociating with the hard
>simulated evidence.
This is absolutely the right advice. By persisting in quantifying what
cannot be quantified we risk producing models which are misleading and maybe
even fundamentally wrong. Forrester, back in 1961, said
In the proper formulation of model the variables should correspond to those
in the real system and should represent the concepts, social pressures and
sources of information which control the actual decisions. Often our
attempts to do that are no more than guess-work.
Regards,
Geoff
Professor R G Coyle,
Consultant in System Dynamics and Strategic Modelling,
Telephone +44 (0) 1793 782817, Fax ... 783188
email geoff.coyle@btinternet.com
>apart from marketig our presently restricted experience is: use hard data
>and models only and treat soft arguments in negociating with the hard
>simulated evidence.
This is absolutely the right advice. By persisting in quantifying what
cannot be quantified we risk producing models which are misleading and maybe
even fundamentally wrong. Forrester, back in 1961, said
In the proper formulation of model the variables should correspond to those
in the real system and should represent the concepts, social pressures and
sources of information which control the actual decisions. Often our
attempts to do that are no more than guess-work.
Regards,
Geoff
Professor R G Coyle,
Consultant in System Dynamics and Strategic Modelling,
Telephone +44 (0) 1793 782817, Fax ... 783188
email geoff.coyle@btinternet.com
-
Norman_Marshall@mckinsey.com
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Ive always subscribed more to the views set out by Jay and George: soft
variables are far too important to leave out, even if they have to be
estimated. However, a recent experience has made me concede that a
pragmatist might in some circumstances be wise to leave them out - at least
to start with.
I recently been involved in an assignment where a stock flow model of the
HR chain was a key analytical tool. We were aware that on the one hand soft
variables were important drivers of the system; on the other we knew that
attempting to put them in would get us into a syndication exercise of
religious intensity and geological timescales. So instead we built a much
simpler model that showed what would happen if the momentum of hiring and
attrition continued under some fixed assumptions. No soft variables
(except, as Jay points out, what were already buried in the assumptions on
attrition and hiring rates), and almost no feedback. The clients looking at
the output then had to do some mental simulation of their own along the
lines of "Well, if the ratio of middle managers to front line staff got way
out there, then wed see some unhappy campers, and then ...". They were
then into discussing how they could influence the way the system got back
into equilibrium, rather than watching it happen. The model was also very
illuminating in terms of how long it would take various initiatives to
change the shape of the HR chain. As a result, the debate got shifted onto
what seems to us so far to have been an altogether more productive footing,
and we have agreement to incorporate the model as a standard corporate
analytical tool and to collect data at a greater level of granularity.
Going forward I can see that we may later be able to incorporate soft data
and more complex feedback structures. I have not changed my point of view
on the ultimate power of soft variables to illuminate a systems true
dynamics. However, for the time being I thing weve got a better result
than we might have done if wed tried to take the high road.
Norman Marshall
McKinsey & Company
From: Norman_Marshall@mckinsey.com
variables are far too important to leave out, even if they have to be
estimated. However, a recent experience has made me concede that a
pragmatist might in some circumstances be wise to leave them out - at least
to start with.
I recently been involved in an assignment where a stock flow model of the
HR chain was a key analytical tool. We were aware that on the one hand soft
variables were important drivers of the system; on the other we knew that
attempting to put them in would get us into a syndication exercise of
religious intensity and geological timescales. So instead we built a much
simpler model that showed what would happen if the momentum of hiring and
attrition continued under some fixed assumptions. No soft variables
(except, as Jay points out, what were already buried in the assumptions on
attrition and hiring rates), and almost no feedback. The clients looking at
the output then had to do some mental simulation of their own along the
lines of "Well, if the ratio of middle managers to front line staff got way
out there, then wed see some unhappy campers, and then ...". They were
then into discussing how they could influence the way the system got back
into equilibrium, rather than watching it happen. The model was also very
illuminating in terms of how long it would take various initiatives to
change the shape of the HR chain. As a result, the debate got shifted onto
what seems to us so far to have been an altogether more productive footing,
and we have agreement to incorporate the model as a standard corporate
analytical tool and to collect data at a greater level of granularity.
Going forward I can see that we may later be able to incorporate soft data
and more complex feedback structures. I have not changed my point of view
on the ultimate power of soft variables to illuminate a systems true
dynamics. However, for the time being I thing weve got a better result
than we might have done if wed tried to take the high road.
Norman Marshall
McKinsey & Company
From: Norman_Marshall@mckinsey.com
-
"SALA ESPIELL, Luis Gustavo"
- Newbie
- Posts: 1
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
I am working to introduce System Dynamics in the company I work for.
During one of the first meetings, a physician said that the quantification that System Dynamics did modelling Human Behaviour reminded him of the film "The dead poets society".
In this film Robin Williams, a literature teacher, used a xy graph in the book "Understanding Poetry" to quantify the greatness of a poem. He rated the poem perfection in the x axe and the importance of the poem in the y axe. The area of the rectangle represents the greatness of the poem. In this way the books author qualifies Shakespeare, Byron, etc.
After the diagram presentation, the literature teacher asked the pupils to tear these pages out their book, because this quantification for him was garbage.
I need the help of the community to continue explaining this person (I used Stermans new book explanation of soft variables and Forresters System Dynamics and the Lesson of 35 years article on sources of information) the differences between the mentioned scene in the film and the SD point of view quantifying soft data.
Thank you in advance for your help
L. Gustavo Sala Espiell
salaesp@netverk.com.ar
Gonnet. Buenos Aires
Argentina
During one of the first meetings, a physician said that the quantification that System Dynamics did modelling Human Behaviour reminded him of the film "The dead poets society".
In this film Robin Williams, a literature teacher, used a xy graph in the book "Understanding Poetry" to quantify the greatness of a poem. He rated the poem perfection in the x axe and the importance of the poem in the y axe. The area of the rectangle represents the greatness of the poem. In this way the books author qualifies Shakespeare, Byron, etc.
After the diagram presentation, the literature teacher asked the pupils to tear these pages out their book, because this quantification for him was garbage.
I need the help of the community to continue explaining this person (I used Stermans new book explanation of soft variables and Forresters System Dynamics and the Lesson of 35 years article on sources of information) the differences between the mentioned scene in the film and the SD point of view quantifying soft data.
Thank you in advance for your help
L. Gustavo Sala Espiell
salaesp@netverk.com.ar
Gonnet. Buenos Aires
Argentina
-
"Raymond T. Joseph, PE"
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
What fun, thanks for the exercise.
The are the same, the are opposites.
They are the same it that the teacher required the students to feel the
impact of the material. As Richard Hamming pointed out, the value of a
model is not in the data it produces but in the experience and understanding
one can derive from it - that is the learning, the internalization to allow
the information to become innate and produce feeling.
They are opposite in that the teacher shunned a quantification - a metric,
especially where that metric may be used to compare qualities. The SD
approach is to quantify, identify a rational basis for comparison.
Ray
From: "Raymond T. Joseph, PE" <ray.joseph@powerwaresolutions.com>
The are the same, the are opposites.
They are the same it that the teacher required the students to feel the
impact of the material. As Richard Hamming pointed out, the value of a
model is not in the data it produces but in the experience and understanding
one can derive from it - that is the learning, the internalization to allow
the information to become innate and produce feeling.
They are opposite in that the teacher shunned a quantification - a metric,
especially where that metric may be used to compare qualities. The SD
approach is to quantify, identify a rational basis for comparison.
Ray
From: "Raymond T. Joseph, PE" <ray.joseph@powerwaresolutions.com>
-
Alexander Leus
- Junior Member
- Posts: 11
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Sala,
I need a little more information from you. What is the purpose of your
dynamic model? Are you modeling the work flow of the various
organizations versus the customer requirements etc. to enhance the work
flow process? Does the work flow process include Doctors, Nurses, Admin
people etc.? What are the boundaries of the dynamic model?
The Doctor that gave you the feedback, I think has a certain perspective
or mental model of the situation. Do you understand his/her mental
model or perspective of the situation being modeled.
Sincerely,
Alex Leus
leusa@tds.net
I need a little more information from you. What is the purpose of your
dynamic model? Are you modeling the work flow of the various
organizations versus the customer requirements etc. to enhance the work
flow process? Does the work flow process include Doctors, Nurses, Admin
people etc.? What are the boundaries of the dynamic model?
The Doctor that gave you the feedback, I think has a certain perspective
or mental model of the situation. Do you understand his/her mental
model or perspective of the situation being modeled.
Sincerely,
Alex Leus
leusa@tds.net
-
Jim Thompson
- Junior Member
- Posts: 8
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
L. Gustavo Sala Espiell wrote, "I need the help of the community to continue
explaining....the differences between the mentioned scene in the film [Dead
Poets Society] and the SD point of view quantifying soft data."
It sounds like youre on slippery ground. Any model variable is like a dab
of paint from a painters pallette. An observer may say he doesnt not like
the color red. But most people cant see individual, subtly applied colors.
When the colors are pla ced by the artist, each seems to lose its identity and
become part of a whole picture. And only a rude or foolish painter would point
out the red in his painting or make it too apparent.
So unless your colleague takes up system dynamics modeling, s/he will be unlikely
to be able to understand why any one variable is in a model. But if your model
provides desirable results, s/he will want more.
Nanu-nanu,
Jim Thompson
jim.thompson@cigna.com
explaining....the differences between the mentioned scene in the film [Dead
Poets Society] and the SD point of view quantifying soft data."
It sounds like youre on slippery ground. Any model variable is like a dab
of paint from a painters pallette. An observer may say he doesnt not like
the color red. But most people cant see individual, subtly applied colors.
When the colors are pla ced by the artist, each seems to lose its identity and
become part of a whole picture. And only a rude or foolish painter would point
out the red in his painting or make it too apparent.
So unless your colleague takes up system dynamics modeling, s/he will be unlikely
to be able to understand why any one variable is in a model. But if your model
provides desirable results, s/he will want more.
Nanu-nanu,
Jim Thompson
jim.thompson@cigna.com
-
George Richardson
- Junior Member
- Posts: 7
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Geoff says:
>By persisting in quantifying what
>cannot be quantified we risk producing models which are misleading and maybe
>even fundamentally wrong.
But if such "soft" variables are crucial to the policy dynamics of
the system, then the only absolutely sure way we can be wrong is to
follow Geoffs advice and not try to quantify them!
(Besides, John Sterman says all models are wrong, so "correctness"
looks like an impossible criterion.)
Moreover, if such "soft" variables are indeed crucial to the policy
choices management faces, then management WILL make their best
guesses about them, no matter what we do. Since formal models give
us opportunities to test our "soft" guesses against real data and to
explore model sensitivities to such approximations, management is
much better off making quantitative stabs at "what cannot be
quantified" than charging ahead with only intuition to ground the
policy choices.
This line of thinking seems so self-evident to me, Id need my friend
Geoff and those who think like him on this issue to show me vividly
how its flawed.
Further, this line of thinking is embedded in the system dynamics
approach from the earliest days by Forrester and others and is
especially illuminated most recently by him in his chapter "Policies,
Decisions, and Information Sources for Modeling," in Morecroft &
Stermanss Modeling for Learning Organizations. So Id need my
friend Geoff to explain in what sense what he is talking about is
actually system dynamics.(!)
...George
--
------------------------------------------------------------------------
George P. Richardson G.P.Richardson@Albany.edu
Chair, Dept. of Public Administration and Policy 518-442-5258
Rockefeller College of Public Affairs and Policy 518-442-5298
University at Albany, Albany, NY 12222 http://www.albany.edu/~gpr
------------------------------------------------------------------------
>By persisting in quantifying what
>cannot be quantified we risk producing models which are misleading and maybe
>even fundamentally wrong.
But if such "soft" variables are crucial to the policy dynamics of
the system, then the only absolutely sure way we can be wrong is to
follow Geoffs advice and not try to quantify them!
(Besides, John Sterman says all models are wrong, so "correctness"
looks like an impossible criterion.)
Moreover, if such "soft" variables are indeed crucial to the policy
choices management faces, then management WILL make their best
guesses about them, no matter what we do. Since formal models give
us opportunities to test our "soft" guesses against real data and to
explore model sensitivities to such approximations, management is
much better off making quantitative stabs at "what cannot be
quantified" than charging ahead with only intuition to ground the
policy choices.
This line of thinking seems so self-evident to me, Id need my friend
Geoff and those who think like him on this issue to show me vividly
how its flawed.
Further, this line of thinking is embedded in the system dynamics
approach from the earliest days by Forrester and others and is
especially illuminated most recently by him in his chapter "Policies,
Decisions, and Information Sources for Modeling," in Morecroft &
Stermanss Modeling for Learning Organizations. So Id need my
friend Geoff to explain in what sense what he is talking about is
actually system dynamics.(!)
...George
--
------------------------------------------------------------------------
George P. Richardson G.P.Richardson@Albany.edu
Chair, Dept. of Public Administration and Policy 518-442-5258
Rockefeller College of Public Affairs and Policy 518-442-5298
University at Albany, Albany, NY 12222 http://www.albany.edu/~gpr
------------------------------------------------------------------------
-
"Jim Hines"
- Senior Member
- Posts: 88
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
The difference between the scene you describe and the system dynamics view
lies in the reason for quantifying soft variables. In the scene, the reason
for quantifying the literary works is to compare the various works of
Shakespeare, Byron, etc. The purpose of quantifying in system dynamics is
so that a computer can simulate the dynamics of some structure of which the
soft variables are a part.
The quantification of literary works in the scene is **not** very useful,
because it doesnt help anyone understand the differences between
Shapespeare to Byron. The quantification in SD models **IS** useful because
it lets the computer work out approximations of dynamics we are interested
in.
It would be better if we could program our computers in naturally, fuzzy
language and have them display the dynamics in a natural fuzzy way, too.
But, our computers dont operate that way. So, we do the best we can,
quantifying soft variables. The real question is: Is it useful? And the
answer will depend on whether the resulting simulated dynamics lead to some
insight you didnt have before.
Regards,
Jim Hines
From: "Jim Hines" <jhines@MIT.EDU>
lies in the reason for quantifying soft variables. In the scene, the reason
for quantifying the literary works is to compare the various works of
Shakespeare, Byron, etc. The purpose of quantifying in system dynamics is
so that a computer can simulate the dynamics of some structure of which the
soft variables are a part.
The quantification of literary works in the scene is **not** very useful,
because it doesnt help anyone understand the differences between
Shapespeare to Byron. The quantification in SD models **IS** useful because
it lets the computer work out approximations of dynamics we are interested
in.
It would be better if we could program our computers in naturally, fuzzy
language and have them display the dynamics in a natural fuzzy way, too.
But, our computers dont operate that way. So, we do the best we can,
quantifying soft variables. The real question is: Is it useful? And the
answer will depend on whether the resulting simulated dynamics lead to some
insight you didnt have before.
Regards,
Jim Hines
From: "Jim Hines" <jhines@MIT.EDU>
-
"Raymond T. Joseph, PE"
- Junior Member
- Posts: 2
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
One of the base intents of an SD approach is to understand a system. Hard
data may make it (seem) easy to do this. Mathematically it is not
necessary. Only a relationship needs to be defined: The red button
produces a peanut, the blue button produces a walnut. This example produces
no scale to quantify a mapping of color to type of nut - just a
relationship. Where there are many such relationships, a model may become
very useful to understand the system. There is hard data here, but no
ordering.
My body does not have a thermometer for me to read how cold I feel. But
when I feel a little cold, I can turn up the rooms thermostat to help me
feel more comfortable. I dont really care what setting the thermostat is
on if I want it warmer, I just turn it up. Here we have an ordered
relationship: cold, colder, . . . up, up further, . . . This is sufficient
data to define a relationship. In fact it is sufficient to define a dynamic
system. The domain and range sets are fuzzy, no hard data, but the
relationships are well defined. This provides sufficient information that
one could build a controller where I may state my feeling and the system
responds by making the room more comfortable.
Thus we can take some soft data and turn it into value. Mathematics
provides a nice tool set to help define the requirements and methods.
Hard data may be easier to work with but that doesnt restrict the viability
of using soft data to drive knowledge.
Ray
From: "Raymond T. Joseph, PE" <ray.joseph@powerwaresolutions.com>
data may make it (seem) easy to do this. Mathematically it is not
necessary. Only a relationship needs to be defined: The red button
produces a peanut, the blue button produces a walnut. This example produces
no scale to quantify a mapping of color to type of nut - just a
relationship. Where there are many such relationships, a model may become
very useful to understand the system. There is hard data here, but no
ordering.
My body does not have a thermometer for me to read how cold I feel. But
when I feel a little cold, I can turn up the rooms thermostat to help me
feel more comfortable. I dont really care what setting the thermostat is
on if I want it warmer, I just turn it up. Here we have an ordered
relationship: cold, colder, . . . up, up further, . . . This is sufficient
data to define a relationship. In fact it is sufficient to define a dynamic
system. The domain and range sets are fuzzy, no hard data, but the
relationships are well defined. This provides sufficient information that
one could build a controller where I may state my feeling and the system
responds by making the room more comfortable.
Thus we can take some soft data and turn it into value. Mathematics
provides a nice tool set to help define the requirements and methods.
Hard data may be easier to work with but that doesnt restrict the viability
of using soft data to drive knowledge.
Ray
From: "Raymond T. Joseph, PE" <ray.joseph@powerwaresolutions.com>
-
Steve Shervais
- Newbie
- Posts: 1
- Joined: Fri Mar 29, 2002 3:39 am
-
"Jay W. Forrester"
- Senior Member
- Posts: 63
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
We have seen a diversity of views on how to use and even whether or not to
use "soft" variables in system dynamics models.
Guenther Ossimitz wrote," SD works best were actual stocks and flows are
involved, as in production lines, delivery chains, stocks of workforce,
capital or whatever. In these fields quantification is quite natural."
Geoff Coyle added, " By persisting in quantifying what cannot be quantified
we risk producing models which are misleading and maybe even fundamentally
wrong."
And on the other side, George Richardson replied, "if such "soft" variables
are indeed crucial to the policy choices management faces, then management
WILL make their best guesses about them, no matter what we do. ... to
explore model sensitivities to such approximations, management is much
better off making quantitative stabs at "what cannot be quantified" than
charging ahead with only intuition to ground the policy choices."
I believe that the maximum reward from system dynamics comes from working
on the most serious problems facing an organization. Seldom does the
difference between corporate success and corporate failure depend on how
inventories are managed. Even when inventories are the critical issue, the
solution is not to be found in flow of parts into and out of inventories
but rather in the upper level decision making that determines adequacy of
inventories, purchasing of capital plant, and finance.
In my own modeling experience, very often only ten percent of the variables
are of the hard variety like inventories and employees and the other ninety
percent are such matters as pressures from the finance people on holding
down inventories, or the marketing people exerting influence to lower
prices until there are not sufficient resources to provide the quality and
prompt delivery that customers want.
Even in the area of so-called "hard" variables like inventories, one will
still be using soft concepts. As I look at Coyles book, "System Dynamics
Modeling," on page 212, in a diagram about inventories, I see some assumed
constants that are actually semi-soft variables behind which, in real
systems, are an array of soft variables--averaging period for order rate,
weeks of average orders, and time to correct order backlog. Where do such
parameters come from? They are usually influenced by managerial pressures
from outside the inventory system, are subject to pressures from finances,
and complaints from marketing and customers.
Are there any concepts that can not be quantified? Anything that can be
discussed in a better or worse, or a more or less, sense can be quantified.
Consider the integrity of a person or organization. Integrity is
considered very important in many situations. I assume it lies in what
most people would consider soft variables. If it is important to the
problem at hand, then to leave it out because it is soft is to say that it
has no influence, which would be the worst possible assumption. If one
lists ten people or corporations that are well known to a dozen people, I
expect that there will be a rather similar ranking of the ten by the dozen
judges. Not identical, but similar. Choose an arbitrary scale, like zero
to ten. Then assign a low integrity candidate at three and a high
integrity example as eight. Now there is a scale. The challenge then
becomes deciding how to use integrity in decision making on such matters as
bidding on contracts, which is soft but important. Taking these steps into
soft variables always sharpens peoples thinking, and the resulting model
usually reveals important but surprising behavior.
---------------------------------------------------------
From: "Jay W. Forrester" <jforestr@MIT.EDU>
Jay W. Forrester
Professor of Management
Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
use "soft" variables in system dynamics models.
Guenther Ossimitz wrote," SD works best were actual stocks and flows are
involved, as in production lines, delivery chains, stocks of workforce,
capital or whatever. In these fields quantification is quite natural."
Geoff Coyle added, " By persisting in quantifying what cannot be quantified
we risk producing models which are misleading and maybe even fundamentally
wrong."
And on the other side, George Richardson replied, "if such "soft" variables
are indeed crucial to the policy choices management faces, then management
WILL make their best guesses about them, no matter what we do. ... to
explore model sensitivities to such approximations, management is much
better off making quantitative stabs at "what cannot be quantified" than
charging ahead with only intuition to ground the policy choices."
I believe that the maximum reward from system dynamics comes from working
on the most serious problems facing an organization. Seldom does the
difference between corporate success and corporate failure depend on how
inventories are managed. Even when inventories are the critical issue, the
solution is not to be found in flow of parts into and out of inventories
but rather in the upper level decision making that determines adequacy of
inventories, purchasing of capital plant, and finance.
In my own modeling experience, very often only ten percent of the variables
are of the hard variety like inventories and employees and the other ninety
percent are such matters as pressures from the finance people on holding
down inventories, or the marketing people exerting influence to lower
prices until there are not sufficient resources to provide the quality and
prompt delivery that customers want.
Even in the area of so-called "hard" variables like inventories, one will
still be using soft concepts. As I look at Coyles book, "System Dynamics
Modeling," on page 212, in a diagram about inventories, I see some assumed
constants that are actually semi-soft variables behind which, in real
systems, are an array of soft variables--averaging period for order rate,
weeks of average orders, and time to correct order backlog. Where do such
parameters come from? They are usually influenced by managerial pressures
from outside the inventory system, are subject to pressures from finances,
and complaints from marketing and customers.
Are there any concepts that can not be quantified? Anything that can be
discussed in a better or worse, or a more or less, sense can be quantified.
Consider the integrity of a person or organization. Integrity is
considered very important in many situations. I assume it lies in what
most people would consider soft variables. If it is important to the
problem at hand, then to leave it out because it is soft is to say that it
has no influence, which would be the worst possible assumption. If one
lists ten people or corporations that are well known to a dozen people, I
expect that there will be a rather similar ranking of the ten by the dozen
judges. Not identical, but similar. Choose an arbitrary scale, like zero
to ten. Then assign a low integrity candidate at three and a high
integrity example as eight. Now there is a scale. The challenge then
becomes deciding how to use integrity in decision making on such matters as
bidding on contracts, which is soft but important. Taking these steps into
soft variables always sharpens peoples thinking, and the resulting model
usually reveals important but surprising behavior.
---------------------------------------------------------
From: "Jay W. Forrester" <jforestr@MIT.EDU>
Jay W. Forrester
Professor of Management
Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
-
marcelo somenson
- Newbie
- Posts: 1
- Joined: Fri Mar 29, 2002 3:39 am
Quantifying soft data
Respect the question of the soft variables, I think that the physician that
have questioned the modeling with soft variables, must re-consider the advances
of epistemiology in the last decades. I am refering to the "objetivity" of the
science: As Forrester said, the soft variables are in the 90% of the models. I
say, where there are no softs variables?
Since the models are made by humans, they reflect the human thougth, equal than
science do in all the disciplines (including the physics.) It is a more
"objetive" model (more "hard") the one that can recognice the subjetivity
underlying, than the one that ignore it or deny it.
The question is not the use or not of soft variables, but the identification of
the accurately in the use of them, the utility of modeling them, and so on.
It´s important, in any case, the modeler indicate in an explicit form what are
the concepts and the values that underly the variables used. In much cases, the
soft variables can be a "cuantification" of expertise, collective or
institucional values.
By the other side, the models are usually used in decision making processes,
wich are always defined by persons (one o more, grouped or not), wich will at
last use "soft" variables in making the final choise. (no matter how hard the
model ofered by the expert should be).
Lic. Marcelo Somenson
Environmental Consultant
Argentina
from: somen@netverk.com.ar
have questioned the modeling with soft variables, must re-consider the advances
of epistemiology in the last decades. I am refering to the "objetivity" of the
science: As Forrester said, the soft variables are in the 90% of the models. I
say, where there are no softs variables?
Since the models are made by humans, they reflect the human thougth, equal than
science do in all the disciplines (including the physics.) It is a more
"objetive" model (more "hard") the one that can recognice the subjetivity
underlying, than the one that ignore it or deny it.
The question is not the use or not of soft variables, but the identification of
the accurately in the use of them, the utility of modeling them, and so on.
It´s important, in any case, the modeler indicate in an explicit form what are
the concepts and the values that underly the variables used. In much cases, the
soft variables can be a "cuantification" of expertise, collective or
institucional values.
By the other side, the models are usually used in decision making processes,
wich are always defined by persons (one o more, grouped or not), wich will at
last use "soft" variables in making the final choise. (no matter how hard the
model ofered by the expert should be).
Lic. Marcelo Somenson
Environmental Consultant
Argentina
from: somen@netverk.com.ar