QUERY Who wants to share models

[""Magne Myrtveit"" <magne@myr]

Posted by ""Magne Myrtveit"" <magne@myrtveit.com>

The recent discussion on model interchange formats raised some questions
relating to the usefulness of such a format:

Assume that we already had a way to share models among different SD tools?

- Would it make a difference?

- Would people use it?

Sharing of models is useful only if:

1) Modellers create models that have a value also to other people

2) Modellers want to make such models available to others (share, publish,
sell, ...)

3) The models in question are (more or less) disconnected from a particular
software technology

John Sterman's ""Business Dynamics"" is an example of a modeller who have
created valuable models, and who has made the models available to others on
multiple software platforms.

Are there others who have or would like to share models?

Or is Sterman just an exception to the general rule?

Education is important for the future generations' understanding of systems.
So let's start by looking for models in that field:

1) Are there SD models around that are (or can be) used in teaching?

2) Are the copyright-holders interested in distributing/publishing their
models?

3) Do the copyright-holders see their models as a way to sell software, or
is it conceivable that the models and the accompanying material can be made
available on multiple software platforms (just as Sterman's book)?

I think the availability of good SD models is very limited. (If someone
looks for high-quality SD models in areas such as, teaching, education,
research, or business use; where will he or she find such models?)

If people are interested in creating and distributing useful models, a model
interchange format can be of some value both to the authors and to the
users. (Authors get a larger potential user base. Users get a larger source
of models to draw from).

Best regards,

Magne Myrtveit
Posted by ""Magne Myrtveit"" <magne@myrtveit.com>
posting date Wed, 2 Apr 2008 19:28:12 +0200
_______________________________________________

[""Chip Hines"" <hines.chip@gm]

Posted by ""Chip Hines"" <hines.chip@gmail.com>

Good points. Regarding the mechanism of exchanges, I was involved with
efforts to create ""exchange standards"" in the emergency management and
associated communities, and we had much the same issues. I think that
it turns out to be advantageous to everyone to have the ability to
exchange information, and when vendors realize this many more
opportunities open up to them - they can sell their product based on how
well it meets the users needs, and as their customers get to know and
love their product, they wont have to leave it to use a model developed
elsewhere.

Everybody wins, but its difficult changing the paradigm (now where have
I heard that before?) The approach that was worked out was
collaborative, inclusive, rapid (at least as far as standards making
processes go), and effective. Essentially the standards were about the
information that needed to be shared between systems. Developers only
needed to develop the translation of information in the standard to
their internal structures, and implement the ability to send and
receive. THe good news from their perspective was that they only needed
to do this once, and their software would be able to use the products of
any other compliant system. Users would benefit from the information (or
model) and use it in the format/software that was most familiar to them.

I suspect that it would be good for the whole community to be able to
seamlessly exchange models, and would really help the discipline and the
end users move forward. Developing these standards isnt easy though.
The process was interesting and involved iterative consensus building on
the standard itself then the elements of the standard, using XML. The
standards were born with the acronym Emergency Data eXchange Language,
but we soon came to see that this wasnt accurate, but the more correct
version - Emergency Messaging eXchange Standards (EMXS) never really
replaced it. The draft, facilitated standards were then submitted to an
international standards body (OASIS) and benefitted from the formal
standards making and ""certification"" process.

It was a great example of the federal government facilitating, but not
owning a process that involved users, vendors, government and standards
organizations to produce truely effective and implemented standards.
Unfortunately, in these days of budget issues and politics, its future
is uncertain.

Id be happy to share the experiences we've had if anyone is interested.

With regard to the question of whether people want to exchange models,
my guess is that they would, but there are a lot of barriers to doing so
that make it hard to commit to. This also says a lot for those who do
share already.

chip
Posted by ""Chip Hines"" <hines.chip@gmail.com>
posting date Thu, 3 Apr 2008 09:20:43 -0400
_______________________________________________

[""Bobby Powers"" <bobbypowers]

Posted by ""Bobby Powers"" <bobbypowers@gmail.com>

Hello,

I'm very interested in sharing models, even though I'm a student and
don't have many (yet!) of my own. I think this is kind of a catch-22,
people don't share models because its not easy, and its not easy because
there has not been a big enough demand to make it worthwhile for
commercial vendors to make it happen. Since we're talking about open
models, there is a great parallel to open source software. There are
several sites like SourceForge [1] and Google Code [2] which give free
hosting to open source projects. They provide an amount of webspace, a
system for providing downloads of the software (including different
versions), and ways to back up the source code (version control systems,
which incrementally keep track of changes in the code, allowing you to
see who did what when multiple people are working together, and go back
to earlier versions when stuff doesn't work). They also provide a
common place for people to go to search for open source projects
(SourceForge hosts almost 180 thousand different projects!), to see if
there is something they can adapt to their needs before starting a new
project of their own.

I think what we need is something similar for modelers: a free service
that allows you to quickly add a site for your model, upload it, link to
or upload related publications, and perhaps have a small wiki about it
where you can give people an introduction, and others can add feedback
or share what they've done with it. It would also be neat for the
server to be able to read your model and automatically create some
content: it could simply be webpages with the sketches as pictures and
graphs of the results to start.

Sure not every model people will share will be useful, but if there
isn't a place to exchange models, people won't have nearly the same
ability to collaborate. I would love to work on this with people.

yours,
Bobby Powers

[1] http://sourceforge.net/index.php
[2] http://code.google.com/hosting
Posted by ""Bobby Powers"" <bobbypowers@gmail.com>
posting date Thu, 3 Apr 2008 15:44:58 +0200
_______________________________________________

[Tom Fiddaman <tom@ventanasyst]

Posted by Tom Fiddaman <tom@ventanasystems.com>

At 04:31 AM 4/4/2008, SDMAIL Bobby Powers wrote:
> I think what we need is something similar for modelers: a free service
> that allows you to quickly add a site for your model, upload it, link to
> or upload related publications, and perhaps have a small wiki about it
> where you can give people an introduction, and others can add feedback
> or share what they've done with it.

Funny you should say that ... based on the earlier open source and
translation conversations, I started a model wiki as an experiment. I'm
gradually migrating my existing model library into it, in the hope that
it'll be easier to maintain. My preliminary conclusion is that a wiki
probably isn't where you want to end up; it's a little too unstructured.
However, it's a great way to prototype what could ultimately be of use
to the community. So ... have at it:

http://modelwiki.metasd.com

For now, I've allowed self-registration
and uploads of Vensim, PowerSim, and Stella/ithink files. As long as
things stay under control, I'm happy to host this experiment.

I think there's no reason why you couldn't conduct a collaborative
modeling effort on sourceforge (we're using subversion (SVN)
successfully to manage models). I suspect it would work well for a bunch
of people working on one model, but not so well for building a library
of models and model components.

Tom
Posted by Tom Fiddaman <tom@ventanasystems.com>
posting date Fri, 04 Apr 2008 15:27:17 -0600
_______________________________________________

[<martin@utalca.cl>]

Posted by <martin@utalca.cl>

Hi Magne and everybody,

In the light of the discussion converning ""model or modeling"", wouldn't a
model-sharing technology enable to ""model together"", too?

I mean, it's interesting to be able to distribute models that already exist, but
- as far as I'm concerned with my teaching activities (mainly undergraduate
university), the books by Forrester, Sterman and Morecroft seem to be a very
fine model sharing technology, the books explain a lot and one can either
re-build the models (interesting guided activity) or use it from the companion CD.

However, it's rather hard to model together in a group withount being physicall
together. Wouldn't sharing technology enable ""collaborative modeling for learning""?

Best,

Martin
Posted by martin@utalca.cl
posting date Fri, 04 Apr 2008 08:52:57 -0400
_______________________________________________

[""Magne Myrtveit"" <magne@myr]

Posted by ""Magne Myrtveit"" <magne@myrtveit.com>

Dear Martin,

In order to answer your question, I need to understand what is ment by
""collaborative modeling for learning"".

- what is the purpose of the process?
- what is the product(s) or deliverable(s) that comes out of the process?
- who takes part in the process?
- how is it done?
- who uses the end-result? (if there is one
- ...

[Inline - reply from Martin Schaffernicht <martin@utalca.cl> ]
> Dear Magne,
>
> Thanks Magne for these questions, I'll offer my answers.
>
> In recent years, ""collaborative"" learning has become more and more
> important/mentioned. It refers to when students/pupils work together in
> order to learn something. A typical example of software are shared
> sketchpads, but I've also seen an application that would allow several
> tudents to jointly develop system diagrams (over the Internet).
>
> 1. The /purpose /is allowing students to learn ""better"", drawing upon
> their individually different knowledge resources and to benefit
> from interaction. ""Better"" means that they reach a learning
> outcome more easily or that they achieve depper learning.
> 2. The immedtiate /product /would be a simulation model that has been
> jointly developed by several individuals.
> 3. The /participants /would be the students or pupils and one or
> several lecturers.
> 4. I have only one example for ""/how it is don/e"". With one of my
> courses, we are currently starting to adapt David Wheat's
> ""MacroLab"" model (which deals with the US economy) to the Chilean
> economy. There are 10 sectors or submodels, and there is a group
> of (2-3) students assigned to each sector. Each group has to
> understand the current form of ""their"" sector and build a nev
> version that is addapted to Chile. They cannot to this without
> interacting with the other groups (sectors). Since we work with
> the ""iThink"" software, how can they organize their interactions?
> If each group was able to maintain and share ""their"" sector, but
> their part of the whoe model can be shared and interact with he
> other sector models, they could easily do this. However, in the
> current situation they will need to look at two separated files,
> which doubtlessly rises the cognitive load. (I hope this
> description is understandable.)
> 5. The /end-result/ (at least in the case I described) is not so much
> the resulting model as the resulting mental models (understanding)
> of the students. This may look like a rather volatile end-result,
> and its benefits would still have to be assessed (but I believe
> this can be done).
>
> I admit that in this case, I woud not think so much in a repository for
> finished models. Quite the contrary, future groups of students whould
> have to re-do their own effort in order to construct their mental
> models. However, once you have sharable models that can interact (let's
> say over the Internet), one can also think of a diferent situation.
>
> John Morecroft's textbook starts with a fishery example where readers
> can use the model on the CD to play and dig in. If I could, I'd make my
> students organize into several groups, each representing one fishery
> firm. They'd hve to develop their analysis of the situation and their
> policies in the form of a model, and then the ""game"" or ""simulation""
> would be the reiterated interaction of their models with a separate
> model (that implements the ""physics"" of the fish resources). We would
> then have a very clear statement of policies (the models) and a
> simulation that shows the dynamics consequences. In such a scenario, a
> set of models may exist in a repository, available to be used or customized.
>
> Still, all I'm talking about takes place in educational settings.
>
> I hope you'll find these answers understandable,
> Martin

Dear Martin,

Thank you for very clear answers to my uninformed questions. In return,
let me try to give some feedback from a technological point of view.

The processes you describe involve some manual work that can be made
easier and safer with the right support from modelling software.

Some software requirements that can be derived from the needs you
describe are listed below.

1) Manage libraries of complete or partial models

2) Import complete or partial models into existing (or new) models

3) Divide a model into parts that can be worked on independently by
different people

4) Easy updating of consolidated models when individual pieces are
changed, e.g., in response to updated model libraries or revisions
received from members of the group modelling team,

In the ideal world, the libraries of complete or partial models can hold
models developed in any of the popular SD technologies. Maybe a model
interchange format can contribute towards that vision.

The software features we talk about would be useful not only for
education, but businesses as well. Therefore, Dynaplan has invested some
effort into finding solutions along the lines described here. If you are
not too afraid of running into bugs in pre-release software, you and
your students are welcome to test out Dynaplan Smia alpha, which
implements solutions to the four requirements listed above.

Best regards,
Magne
Posted by ""Magne Myrtveit"" <magne@myrtveit.com>
posting date Fri, 4 Apr 2008 16:12:24 +0200
_______________________________________________

[Bill Braun <bbraun@hlthsys.co]

Posted by Bill Braun <bbraun@hlthsys.com>

SDMAIL Magne Myrtveit wrote:
> In order to answer your question, I need to understand what is ment by
> ""collaborative modeling for learning"".

How a group might collaborate to write a paper is a good example. The
ability to track changes, have multiple versions, and compare documents.
I realize there is a magnitude of complexity well beyond the written
word involved, but it may still serve as an example.

Bill Braun
Posted by Bill Braun <bbraun@hlthsys.com>
posting date Sat, 05 Apr 2008 08:09:55 -0400
_______________________________________________

[Jean-Jacques Laublé <jean-jac]

Posted by Jean-Jacques Laublé <jean-jacques.lauble@wanadoo.fr>

Hi Martin

What would be interesting, is a place where anybody with
a real world problem could submit it and have people working freely
to solve it working together.
The motivation of these people would be to learn.
There could be different graduated sections with different level of expertise and one
should find some way to verify the minimum ability of the members relative to
the level of expertise of the section.
An external person submitting a problem, depending on its difficulty
would be directed on the appropriate section depending on its level
of complexity.
Just an idea.
Regards.
Jean-Jacques Laublé Eurli Allocar
Strasbourg France.
Posted by Jean-Jacques Laublé <jean-jacques.lauble@wanadoo.fr>
posting date Sat, 5 Apr 2008 14:52:01 +0200
_______________________________________________

[<martin@utalca.cl>]

Posted by <martin@utalca.cl>

Dear Bill,

I tried to explain what I meant in a reply to Magne's questions - I reproduce
the answers below, because they will show you what I meant with ""collaborative"".

Beyond these explanations, I find what you write extremely relevant: being able
to compare different versions of a model would make it easier to ""see"" traces of
learning that has been going on.

Here come the answers, I hope you'll find them helpful:

In recent years, ""collaborative"" learning has become more and more
important/mentioned. It refers to when students/pupils work together in order
to learn something. A typical example of software are shared sketchpads, but
I've also seen an application that would allow several tudents to jointly
develop system diagrams (over the Internet).

The purpose is allowing students to learn ""better"", drawing upon their
individually different knowledge resources and to benefit from interaction.
""Better"" means that they reach a learning outcome more easily or that they
achieve depper learning.

The immedtiate product would be a simulation model that has been jointly
developed by several individuals.

The participants would be the students or pupils and one or several lecturers.
I have only one example for ""how it is done"". With one of my courses, we are
currently starting to adapt David Wheat's ""MacroLab"" model (which deals with the
US economy) to the Chilean economy. There are 10 sectors or submodels, and
there is a group of (2-3) students assigned to each sector. Each group has to
understand the current form of ""their"" sector and build a nev version that is
addapted to Chile. They cannot to this without interacting with the other
groups (sectors). Since we work with the ""iThink"" software, how can they
organize their interactions? If each group was able to maintain and share
""their"" sector, but their part of the whoe model can be shared and interact with
he other sector models, they could easily do this. However, in the current
situation they will need to look at two separated files, which doubtlessly rises
the cognitive load. (I hope this description is understandable.)

The end-result (at least in the case I described) is not so much the resulting
model as the resulting mental models (understanding) of the students. This may
look like a rather volatile end-result, and its benefits would still have to be
assessed (but I believe this can be done).

I admit that in this case, I woud not think so much in a repository for finished
models. Quite the contrary, future groups of students whould have to re-do
their own effort in order to construct their mental models. However, once you
have sharable models that can interact (let's say over the Internet), one can
also think of a diferent situation.

John Morecroft's textbook starts with a fishery example where readers can use
the model on the CD to play and dig in. If I could, I'd make my students
organize into several groups, each representing one fishery firm. They'd hve to
develop their analysis of the situation and their policies in the form of a
model, and then the ""game"" or ""simulation"" would be the reiterated interaction
of their models with a separate model (that implements the ""physics"" of the fish
resources). We would then have a very clear statement of policies (the models)
and a simulation that shows the dynamics consequences. In such a scenario, a
set of models may exist in a repository, available to be used or customized.

Still, all I'm talking about takes place in educational settings.

Best greetings,
Martin
Posted by martin@utalca.cl
posting date Sun, 06 Apr 2008 13:35:13 -0400
_______________________________________________

[Kate Thompson <k.thompson@edf]

Posted by Kate Thompson <k.thompson@edfac.usyd.edu.au>

Hi all,

I have to agree with Martin on this one. I use a combination of an
online chat environment and a wiki for my (masters level) students to
collaboratively build models. They have access to a shared whiteboard as
part of the chat environment so that they can synchronously work on
their causal loop diagrams with discussion, but they have to make
changes to their actual models separately, upload to the wiki, the rest
of the group has to download, open, and make appropriate changes etc.
What actually happens is that one or two students end up being the
""modellers"" and upload screenshots instead of the actual models, the
rest of them comment on what they think should be changed... but it
could be better. What would be great would be an online resource that
allowed students to build the models synchronously (and writing all the
changes to a database so that we could ""see"" what they changed and
commented on would be helpful too).

That's my wishlist for the day!
Thanks
Kate
Posted by Kate Thompson <k.thompson@edfac.usyd.edu.au>
posting date Mon, 7 Apr 2008 12:13:19 +1000
_______________________________________________

[Stephen Wehrenberg <stephen.w]

Posted by Stephen Wehrenberg <stephen.wehrenberg@verizon.net>

Jean-Jacques,

That's exactly the kind of thing I was talking about in my original post
on the topic ... more collaborative workspaces and not necessarily for
expert modelers, but for people who don't know anything about SD but do
understand problems.

Steve

Stephen B. Wehrenberg, Ph.D./
/Human Resource Strategy and Capability; Director, Future Force;
and Director of Executive Development, US Coast Guard
Organizational Sciences, The George Washington University
Posted by Stephen Wehrenberg <stephen.wehrenberg@verizon.net>
posting date Sun, 06 Apr 2008 12:07:32 -0400
_______________________________________________

[Bill Braun <bbraun@hlthsys.co]

Posted by Bill Braun <bbraun@hlthsys.com>

Martin offers a perspective on collaboration.

Good points, all. In further thinking about Magne's question, and your
thoughts also, I think it may be useful to distinguish between group
model building and group model construction. There is much good thinking
on the former and I am not familiar with anything that addresses the
latter (which I take to mean more than one modeler at work on the same
model at the same time, face-to-face and/or asynchronously).

Dividing a model into sectors as you suggest is viable. Is that the same
as co-constructing? Is such a thing possible given the complexities of
modeling, especially large projects? If there were a universally
recognized library of pieces and parts it might be. There are many good
examples of that also. I'm not sure of their universal acceptance as
building blocks (nor have I even thought about the desirability of such
a thing let alone the practicality).

Bill Braun
Posted by Bill Braun <bbraun@hlthsys.com>
posting date Sun, 06 Apr 2008 13:53:28 -0400
_______________________________________________

[Bill Braun <bbraun@hlthsys.co]

Posted by Bill Braun <bbraun@hlthsys.com>

At the risk of dredging up the common file format conversation, this
strikes me as a good reason for such a thing. I do get the sense that
the absence of responses from the software writers on the idea of a
common format is an indicator that they either still see proprietary
formats to their advantage, do not anticipate a positive ROI on the
effort involved, or both.

A lot of great modeling was accomplished without the benefit of GUI
based software. Is there some middle ground between the convenience of
GUI and text, and which could be ported from application to application
via text?

Bill Braun
Posted by Bill Braun <bbraun@hlthsys.com>
posting date Mon, 07 Apr 2008 14:41:52 -0400
_______________________________________________

[""Magne Myrtveit"" <magne@myr]

Posted by ""Magne Myrtveit"" <magne@myrtveit.com>

Hello Bill,

On your comment:

""Dividing a model into sectors as you suggest is viable. Is that the same
as co-constructing? Is such a thing possible given the complexities of
modelling, especially large projects?""

Yes, what I have in mind is co-construction, in the sense that several
people work on different pieces of one larger model. The architecture is
around one consolidated model and several pieces. The consolidated model is
updated automatically whenever a piece is revised.

Along the same lines, a model can contain components from an external
library. When the library is revised, the models using the library can be
refreshed with the newest information.

This kind of work is especially useful in large projects. It does require
some work on designing the building blocks and the boundaries of each model
piece and how they communicate.

Best regards,
Magne
Posted by ""Magne Myrtveit"" <magne@myrtveit.com>
posting date Mon, 7 Apr 2008 14:11:33 +0200
_______________________________________________

[Bill Harris <bill_harris@faci]

Posted by Bill Harris <bill_harris@facilitatedsystems.com>

""SDMAIL Bill Braun"" <bbraun@hlthsys.com> writes:
> Dividing a model into sectors as you suggest is viable. Is that the same
> as co-constructing? Is such a thing possible given the complexities of
> modeling, especially large projects? If there were a universally
> recognized library of pieces and parts it might be. There are many good
> examples of that also. I'm not sure of their universal acceptance as
> building blocks (nor have I even thought about the desirability of such
> a thing let alone the practicality).

Bill B.,

Before people get too far into the libraries idea, it might be worth
taking a look at the software engineering side of things. Sure, there
are massive libraries in some cases (STL and GSL come to mind), but the
challenge with such libraries is the amount of information you need to
know and remember (or look up) in order to get work done.

One of the advantages of SD has been that the methodology is simple and
the focus stays on the problem. Barry Richmond once opined here, IIRC,
that he got along just fine almost all of the time with combinations of
five or so standard structures. The Molecules give more structure for
cases where it's advantageous, and the various books and texts give
examples of bigger structures for specific situations.

I conjecture that models any bigger than the molecules would likely be
big enough so that they are not much fun to copy manually, at least if
we consider them as productivity aids * (remember typing in programs for
Apple IIs or C-64s in the 1980s?), so libraries might then become
vendor-specific. In the programming side, that seemed to provide
temptation for vendors to make ""improvements"" that contain features that
lock people into their solutions, which (I think) led to more emphasis
on things such as the STL.

* If we consider them as teaching tools, then I suspect people are
better off studying existing models and building their own rather
than thinking they should be memorizing these new standards.

If it's truly needed, a SML (Standard Model Library, to use an
overloaded acronym) could help, but, in that case, why not simply switch
to Goldsim (I think they're farther along in that direction)? The
effort to create a cross-platform (OS X, Linux, Windows) set of
libraries callable from iThink, Vensim, and Powersim doesn't sound
trivial, and the effort to recall the API for each of the structures in
the library wouldn't be trivial, either.

""SDMAIL Bill Braun"" <bbraun@hlthsys.com> writes:
> How a group might collaborate to write a paper is a good example. The
> ability to track changes, have multiple versions, and compare documents.
> I realize there is a magnitude of complexity well beyond the written
> word involved, but it may still serve as an example.

If you're working with a text-based simulator (versions of Vensim are
likely the most popular, but MCSim, DYSMAP, and DYNAMO qualify), what
about using Google Docs, create the model in a document, work on it
together letting Google Docs manage the interaction, and save it as text
on anyone's computer who wants to run it?


""SDMAIL Kate Thompson"" <k.thompson@edfac.usyd.edu.au> writes:
> I have to agree with Martin on this one. I use a combination of an
> online chat environment and a wiki for my (masters level) students to
> collaboratively build models. They have access to a shared whiteboard as
> part of the chat environment so that they can synchronously work on
> their causal loop diagrams with discussion, but they have to make
> changes to their actual models separately, upload to the wiki, the rest
> of the group has to download, open, and make appropriate changes etc.
> What actually happens is that one or two students end up being the
> ""modellers"" and upload screenshots instead of the actual models, the
> rest of them comment on what they think should be changed... but it

Hi Kate,

That sounds like more reason to have plain text as the interchange
format for models (yes, I have used *nix systems for quite a few years
and been infected with the *nix philosophy). I haven't done exactly
what I'm suggesting, but I imagine that a bit of ingenuity and a bit of
effort would let you modularize one text-based model such that
individuals could work on their parts, they could upload them to a
server, and a script could build the resulting full model, ready to run.

With Vensim .mdl files, I suspect that the toughest problem would be the
part at the end that describes the layout of the sketch. With Vensim
Professional or DSS, that's probably not an issue (unless you want to
see a pretty sketch), but it could be with less expensive versions.
There's something to try out.

With MCSim, the toughest problems are likely sorting equations (it
doesn't do that for you) if you divided the work by sector. If you
divided by section of the model file, that's probably not a problem.

With text-based models, diff and patch might be a help, too. If
everyone starts with the same current gold version, each person makes
their changes, and then each person runs diff to get a patch file. Then
you just upload the patch files to the server. Everyone grabs the patch
files and runs patch locally. I'm not an expert there, and I know you
can get into trouble if there are too many overlapping patches, but, for
smaller groups without too much chaos, that might be the easiest way to
speed up what you're suggesting.

> could be better. What would be great would be an online resource that
> allowed students to build the models synchronously (and writing all the
> changes to a database so that we could ""see"" what they changed and
> commented on would be helpful too).

What about tools such as Vyew, WebHuddle, Elluminate, WebEx, Live
Meeting, HP Virtual Rooms, Adobe Connect, ...? That's the way I've done
it. Perhaps your university even has a license for one of those (and
some have free versions).

Bill
- --
Bill Harris
Posted by Bill Harris <bill_harris@facilitatedsystems.com>
posting date Mon, 07 Apr 2008 08:42:26 -0700
_______________________________________________

[j-d <jaideep@optimlator.com&g]

Posted by j-d <jaideep@optimlator.com>

In reply to Martin's description of modeling of the Chilean economy, I
had done similar work using the Limits to Growth (LTG) model. As I
understand it, LTG was done collaboratively by different authors
working on population (Dana Meadows), agriculture, resources, capital,
and pollution. So there were 5 submodels so to speak. I had extended
the model to North (US, Canada, Europe) and South (China, India,
Brazil etc.) regions of countries to understand the dynamics of their
inter-relations over time as resources get tighter and regions compete
or cooperate in different ways (trying to model what Lester Brown and
the WorldWatch Institute people had been saying for years and which
seems to be coming to passnow , with the rising oil prices and so on).
All of this was in the context of dynamic games (which involve
multi-player optimization and are not just simulations).

My impressions are as follows: understanding even a simple nonlinear
model through simulations is difficult indeed (for example,
population, capital models are pretty complicated - resources model
was almost trivial and made LTG model kind of unbalanced, in my view -
but that is okay - it is still a great piece f modeling work).
Splitting the model into sub-models and really going through the
simulations described in the book Dynamics of Growth (DOG) in a Finite
World built my understanding of the various sub-models (did I say I
was working solo on this) - the submodels had exogenous inputs
(constants or functions) as their ""links"" to other submodels, so one
could basically plug-and-pray the sub-models in stead of the exogenous
inputs. For example, population as an exogenous logistic function may
be replaced by the output from a population model, all with cohorts
and internal dynamics as described in DOG. So you could link
population and agriculture models this way, and build your
understanding of linked pop-ag model using simulations. You could
affect population internal dynamics by agricultural parts linking back
to population model, which is when things become a little hairy. They
become really complicated when 10 sub-models are connected - a clear
understanding of internals of each sub-model plus the effect of
different links between submodels is needed. In a real-world
situation, generally you have domain experts for each submodel and I
agree that a complete understanding of everything by everyone is
difficult if not impossible.

To optimize policies, using optimal control or dynamic game theory,
was very hard for the full linked models (CPU requirements plus the
sheer complexity and opaqueness of the model), so I simplified from 10
to I think 2 or 4 submodels for this kind of analyses. Bottomline
learning for me from this whole exercise was that in complex,
nonlinear systems, even discounting the possibilities of chaotic
systems, optimization and simulation have limitations, and a mental
exercise using many simulations, while trying to optimize one or more
policies, can cause deeper intuition and more effective policies. It
is similar to katas or patterns in martial arts practice, where katas
function as libraries of accumulated wisdom, and they are practiced by
all students (and even masters), similar to simulation exercises we
give our students, and then optimization happens when we tweak the
katas or models to suit our real-world problems. I had called this
practice ""optimlation"" at that time

I hope all of the above makes sense -

Best regards

Jaideep Mukherjee, Ph. D.
Posted by j-d <jaideep@optimlator.com>
posting date Mon, 7 Apr 2008 17:23:20 -0500
_______________________________________________

[""richard dudley"" <richard]

Posted by ""'richard dudley'"" <richard.dudley@attglobal.net>

Modelwiki looks like a good experiment for sharing models (as opposed to
the related question of collaborative model building).

Perhaps Tom might give some pointers regarding how we contribute to this. I
see several places where I can put information, but I have no guidelines.
Is there an instant source of information for non-wiki users to get started?

Richard
Posted by ""'richard dudley'"" <richard.dudley@attglobal.net>
posting date Tue, 8 Apr 2008 16:12:58 +0700
_______________________________________________

[Tom Fiddaman <tom@ventanasyst]

Posted by Tom Fiddaman <tom@ventanasystems.com>

Posted by ""'richard dudley'"" <richard.dudley@attglobal.net>
> Modelwiki looks like a good experiment for sharing models (as opposed to
> the related question of collaborative model building).

Right. One could certainly use a wiki for collaborative model building
as well (e.g., to keep a running discussion, documentation, and file
links), but I'm sure better tools could be found. Probably the
combination of a code management system and a forum would be better (as
most sourceforge projects have). I'll try to get a colleague to share
some experience using subversion (SVN) with Vensim text .mdl format
models. That lets users keep models coordinated and review past changes
(including who made them).

There have also been some papers on this topic in recent SD conferences. See
http://www.systemdynamics.org/conf2005/ ... CLA437.pdf
and
http://www.systemdynamics.org/conferenc ... WEL256.pdf

> Perhaps Tom might give some pointers regarding how we contribute to this. I
> see several places where I can put information, but I have no guidelines.
> Is there an instant source of information for non-wiki users to get
> started?

General wiki editing advice is here
http://en.wikipedia.org/wiki/Wikipedia: ... dit_a_page

What I know about wikis would make a very short book. However, I found
it quite easy to create the pages that now exist at
http://modelwiki.metasd.com

First, you need to register & log in to
do anything besides browse. As I recall, I allowed self registration
when I set things up, but if you have trouble, email me.
- Once registered, you can use the Edit tab on any page to modify it.
You can also browse past changes, see who made them, and undo them if
you don't like them.
- With respect to editing, I found I only needed to know about 5 things:
* you can organize a page with headers, using = Level 1 = , == Level 2
== etc.
* you can create pages by simply including the new page name in an
existing page, e.g. here is a link to a [[new page]] - then following
the link will give you blank space to create the new page
* you can link to an [[existing page]] the same way
* you can link to [[an external site]] easily as well
* upload images as images; upload models and anything else as media;
email me if you need a new file type authorized (vensim, powersim, and
stella/ithink models are possible now)

The trick is for the distributed editing to wind up at a point that's
reasonably organized. My guess is that the wiki will work as a
repository for my existing model library. It might also work very well
as a repository for the molecules, which already have a network
structure of relationships (i.e., each molecule potentially has parents
and children and associates). I suspect that it will eventually be
outgrown as a general model repository though. For that, it would be
advantageous to have a more structured database, enabling search and
sorting. For example, one might wish to restrict one's browsing to only
peer reviewed models in biology having less than 100 equations. However,
without messing around a bit, I can't anticipate all the fields that
would be useful in such a respository, so I think the wiki is a good way
to get started by trying things. For that reason, I was deliberately
vague about what ought to be done with it.

I think quality will be key. Only including peer reviewed models would
be one solution, but rather limiting. Instead, I think it would make
sense to develop the guidelines for submission further, in order to
discourage submission of low-quality models (e.g. models with obvious
formulation errors, or missing and unbalanced units). Also, I would
encourage the use of the wiki text to make comments on models. This may
be another area where a modelbase with a review and rating system would
excel at showcasing the best work.

I realize that making this totally open could lead to a mess, but I
rather doubt that will happen. I'll lock things down if there's an
explosion of spam pages or some other disaster, but otherwise I'll just
watch and contribute.

Tom
Posted by Tom Fiddaman <tom@ventanasystems.com>
posting date Tue, 08 Apr 2008 17:35:45 -0600
_______________________________________________

[Dan Goldner <dan@ventanasyste]

Posted by Dan Goldner <dan@ventanasystems.com>

SDMAIL Tom Fiddaman wrote:
> Right. One could certainly use a wiki for collaborative model building
> as well (e.g., to keep a running discussion, documentation, and file
> links), but I'm sure better tools could be found. Probably the
> combination of a code management system and a forum would be better (as
> most sourceforge projects have). I'll try to get a colleague to share
> some experience using subversion (SVN) with Vensim text .mdl format
> models. That lets users keep models coordinated and review past changes
> (including who made them).

Tom mentioned Subversion, a version control system (subversion.org, or
svnbook.red-bean.com for an introduction). This is more on the topic of
how to collaborate on creating models than on sharing & cataloguing
existing models.

Subversion keeps a history of changes to a model and allows multiple
people to edit it, alerting them when conflicts have occured. My
experiments with it have been very successful, though not without
headaches.

The main benefit has been that multiple people can work on a model in
parallel, and no previous revisions are ever lost. That is, Tom may make
excellent changes in revision 326 that I carelessly destroy in revision
327, but we can always revert to or copy & paste from rev. 326 to fix
the error.

The headaches come from the fact that the system is designed to monitor
changes to lines of programming code. This means if I do something I
don't think of as ""changing the model"" such as nudge a sketch element to
the left, or close the model with a different variable in the workbench,
or put carriage returns in equation right-hand-sides for readability,
subversion considers those to be changes. With practice though these
become easy to avoid or manage when they occur.

Subversion does not do anything to help coordinate who is doing what.
For that conversation you need either a wiki (see Tom's post) or an
issue tracker (e.g. Trac - trac.edgewall.org) or an e-mail list like
this one or regular meetings or something.

Karl Fogel has written excellent guide to managing collaboration among a
disperse group at producingoss.com. It's focused on open source but the
techniques are useful for any group.

Notes:
-You need a server to use Subversion, but if you don't have one, an
excellent hosted service for Subversion and Trac is available for
$7/month from hosted-projects.com.

-Lots of Windows users like to use tortoisesvn as an interface to
subversion: tortoisesvn.tigris.org. Like subversion and trac,
tortoisesvn is free.

Posted by Dan Goldner <dan@ventanasystems.com>
posting date Wed, 09 Apr 2008 08:55:41 -0700
_______________________________________________

[""Zagonel, Aldo A"" <aazagon@]

Posted by ""Zagonel, Aldo A"" <aazagon@sandia.gov>

Tom Fiddaman wrote:
> There have also been some papers on this topic in recent SD conferences.

This same group is submitting another paper on collaborative modeling and
working with ""living"" models to this year's conference, in Athens:

Title:
Working with ""living"" Models: Emergent methodological contributions from
modeling for critical infrastructure protection

Abstract:
Critical infrastructures are increasingly automated and interdependent, subject
to possibly cascading vulnerabilities due to equipment failures, natural
disasters, and terrorist attacks. The government seeks to ensure that
disruptions are infrequent, brief, manageable, and cause the least harm
possible. The system dynamics (SD) approach is particularly promising in
understanding these complex systems, interactions, and issues. Problems in
critical infrastructure protection are being investigated with a collection of
SD models developed expressly for these concerns, including agriculture models.
This paper discusses the technical and social modeling context that makes this
SD modeling effort seem uncommon. It involves a modular approach, a model
reassembling technology, a formal process for testing and evaluation, and a
social process to manage the development and use of ""living"" models.

This paper is less technical. It addresses the need for such an approach, and
the elements needed and social process involved to implement it. If anyone would
like to see the draft that was submitted, please contact me directly at
aazagon@sandia.gov

Aldo Zagonel
Albuquerque, NM
Posted by ""Zagonel, Aldo A"" <aazagon@sandia.gov>
posting date Wed, 9 Apr 2008 09:03:37 -0600
_______________________________________________

[Mabel Fong <may_belle_66@yaho]

Posted by Mabel Fong <may_belle_66@yahoo.com>

Hi:

I note that many posts on this subject have called for some
sort of meta-modeling language to facilitate communication of ideas about complex
dynamic systems. As usual, this has suggested some more primitive issues on
which I would appreciate your thoughts.

First, is it not a general commonplace of complex non-linear
dynamic systems analysis that one does not even claim to have a model until it
has produced the same stimulus/response behaviors when programmed in different
languages and run on different CPUs? Is there something special about SD
training, SD software, or the problems addressed by SDers that obviates such
precautions? Or, is it possible that there are ‘validated’ SD models that might
be called into question if they were subjected to such tests?

Second, are we not attempting to define away or automate out
of existence what might better be accepted as permanent issues? At bottom, we
are conveying ideas about systems that are too complicated for their behaviors
to be elucidated by the unaided human imagination. We therefore examine our
ideas by forming their objective embodiment in analog systems. It seems to me
that, while this is the case, communication from mind to mind must occur in
terms of the unappealing task of reading one another’s program documentation - which
always entails the issue of whether or not the documentation actually conveys
what the program is doing.

It is certainly to be granted that the development of SD will
be greatly aided by more self-documenting, bulletproof, and behaviorally rich
software. But, given the differences among SD models and among the problems
addressed by SD, might not the drive toward the ultimate software package be
counterproductive if carried too far? Is it not best to be satisfied with
duplicating one another’s results with our peculiar programming setups?

Mabel
Posted by Mabel Fong <may_belle_66@yahoo.com>
posting date Sat, 19 Apr 2008 14:38:30 -0700 (PDT)
_______________________________________________

[Bill Rathborne <brathborne@sy]

Posted by Bill Rathborne <brathborne@sympatico.ca>

On 20-Apr-08, at 7:31 AM, SDMAIL Mabel Fong wrote:
> It is certainly to be granted that the development of SD will
> be greatly aided by more self-documenting, bulletproof, and
> behaviorally rich
> software. But, given the differences among SD models and among the
> problems
> addressed by SD, might not the drive toward the ultimate software
> package be
> counterproductive if carried too far? Is it not best to be satisfied with
> duplicating one another’s results with our peculiar programming setups?

It appears to me that the objective is to create an ""XML"" type uniform
language that contains the *content* of a particular model that can be
ported across multiple application programs - that may have particular
features, user interfaces, etc. - but where the results (output graphs)
would be the same in all cases. However, each application program may
have options that allow users to select alternative displays not
available in other application programs. ""KML"" is a specific instance
of an ""XML"" language that is used for Google Earth, Extensible Financial
Reporting Markup Language (XFRML) for business, finance and accounting,
and Chemical Markup Language (CML) is a new approach to managing
molecular information.

""XML's"" are text based files with a readily readable structure that can
aid significantly in understanding the structure of a model created by
others.

Bill
Posted by Bill Rathborne <brathborne@sympatico.ca>
posting date Sun, 20 Apr 2008 09:16:45 -0400
_______________________________________________

[Brian Crowe <brian_crowe@i-wo]

Posted by Brian Crowe <brian_crowe@i-worx.com>

I want to build on a point that Bill Rathborne makes in his recent reply
to Mabel Fong's post. This topic is directly in my core expertise.

On Apr 21, 2008, at 5:38 AM, SDMAIL Bill Rathborne wrote:
> ""XML's"" are text based files with a readily readable structure that can
> aid significantly in understanding the structure of a model created by
> others.

XML is not easily readable - but, it contains words we can read, and
that has a great deal of value. If you have ever read the HTML code
that represents a web page, you see a similar degree of ""readability""
that XML-oriented files convey. It is readable after a little
practice. It doesn't convey the dynamics of a model very well. With
some refinement, it could convey all the details in a model between
machines and applications in a file someone could actually read and make
some sense of, but maybe not very much sense without a lot of practice
reading these files.

To satisfy Mabel's concern that models should work identically on any
compatible and like-missioned software, someone must test the model's
description and certify that it ""conforms"" to the specification of the
description language. This process leads to ""interoperability"" - or,
you might rather use the word ""compatibility"".

XML-based languages are not the only candidates for a model description
language. Further, while XML may be useful to describe the high-level
function of the model, its source, applicable areas of coverage,
advantages and limitations, user communities, conformance, etc., it may
never easily describe the dynamic nature of the model to the human
reader. Describing the model's dynamic behavior in a way that the
reader can understand may require a second and quite different
description language. Nothing says that the XML-based language can't
embed the statements of the second language - from one perspective, you
might call the second description language the ""extension"" of XML that
SD requires. It all depends on your perspective... For example:
1) I have a model, and let me tell you about it - who created it, who
uses it, and what problem it helps solve, etc.; or
2) I have a model, and let me tell you about it - what the model
components are, and how they are connected, and what the causal loops
are, and what the inputs and outputs are, and what controls the model
exposes for simulation, etc.
These are two different conversations with very different levels of
detail to convey.

It may not be necessary to make the SD extensions to XML to even be
human readable - it may be important, or it may not be - we decide as a
society of experts if we want to work on this, and if so, than what the
purpose of a description language is and what its requirements are -
this can be a LOT of work, by the way.

Finally, there is one advantage to starting with an XML-based
description language, and that relates a consideration of who important
beneficiaries are for our work. Owen Ambur has posted here about
StratML. If we can describe at least a finite set of attributes for a
model with a StratML file, the US government can import that file into
their information management systems - other governments and industrial
users may follow suit and also make use of XML-based language files.
With this information in this form, the US Government and any other
interested users can create reports, databases, web pages and other
information dissemination vehicles to tell anyone who is interested what
our work is and how it may be applicable to solve that person's
problem. If we want the US government to be a beneficiary of SD
modeling in the future, it may be very important to start here with
XML-based languages like StratML - if the US government user is not
squarely in the Society's mission, XML considerations are less
important. It is something to talk more about after our new mission is
established.

Brian Crowe
President, TELE-WORX
Posted by Brian Crowe <brian_crowe@i-worx.com>
posting date Mon, 21 Apr 2008 11:21:14 -0500
_______________________________________________

[Tom Fiddaman <tom@ventanasyst]

Posted by Tom Fiddaman <tom@ventanasystems.com>

Mabel Fong raises a good point:
> First, is it not a general commonplace of complex non-linear
> dynamic systems analysis that one does not even claim to have a model
> until it
> has produced the same stimulus/response behaviors when programmed in
> different
> languages and run on different CPUs? Is there something special about SD
> training, SD software, or the problems addressed by SDers that
> obviates such
> precautions? Or, is it possible that there are ‘validated’ SD models
> that might
> be called into question if they were subjected to such tests?

It's certainly a good idea to verify numerical sensitivity of models,
but I doubt that it's commonplace, at least outside the natural
sciences. Verification more commonly focuses on the numerical algorithm,
rather than on models implemented using a given algorithm. One might
argue that most SD software, having been around for a while, is well
verified, though I'm reluctant to take that on faith. Given the
counterintuitive nature of systems under study, it would be easy for
errors to persist, with users rationalizing them away. However, if one
follows good SD practice by exploring the model's sensitivity to it's
time step (DT) and solution method (Euler, Runge-Kutta, ...), most
problems are likely to surface. Most will be due to formulations, not
the underlying software.

There is definitely potential for problems though. I have run across
models that behaved differently in single- and double-precision versions
of Vensim and on different CPUs. In those few cases, the differences
were due to what I would consider to be bad formulations, and thus cast
doubt on the model rather than the software. I also know of one model -
Mosekilde et al.'s model of chaos in rat kidneys - that is good, but
still numerically sensitive. I replicated the model from a paper, using
a different numerical scheme (Vensim's FIND ZERO function) to solve a
simultaneous equation. My replication generates the same behavior modes,
but with chaotic transitions at slightly different points in the
parameter space. The model is at
http://metasd.com/models/index.html#Science

I suspect that most SD problems are not really testing the limits of
numerics. Most models of an organization do not track many periods of an
oscillation, or experience wild shifts in feedback loop gains. Even if a
model is chaotic, generally the horizon of interest is short compared to
the time scale over which numerical errors would cause results to
diverge. In general, and especially with chaotic or PDE systems like a
turbulent fluid, one can't count on such benign behavior, and one
shouldn't count on it even when expected. So, again, one should:
- Vary the model time step (a good rule of thumb is to keep halving it
until the model behavior stops changing). Bear in mind that, if the time
step gets too small, increasing roundoff error offsets gains in
precision. A time step that is a power of 2 minimizes roundoff error.
Increasing the time step may also be a useful test, for robustness if
nothing else.
- Vary the integration method. This may not work well for models with
discrete events. Oscillatory models should generally not be using
Euler's method, which tends to amplify oscillations.
- Explore the model's parameter space extensively
Running the model on different software or hardware provides an
additional test, but of these four, I would expect it to be the least
likely strategy to reveal errors.

Stochastic models pose a particular challenge, because experimental
changes (e.g., in the time step) will also cause changes in the random
draws realized. This makes results non-comparable. My personal
preference is to build models such that all stochastic inputs, and
perhaps also data drivers, can be turned off, so that a clean baseline
is available for numerical and equilibrium experiments.



****************************************************
Tom Fiddaman
Ventana Systems, Inc.
Posted by Tom Fiddaman <tom@ventanasystems.com>
posting date Mon, 21 Apr 2008 17:12:39 -0600
_______________________________________________

[""Owen Ambur"" <Owen.Ambur@ve]

Posted by ""Owen Ambur"" <Owen.Ambur@verizon.net>

Since Brian Crowe was speaking my language (XML), perhaps it is not
surprising that I found his posting to be one of the most lucent I've seen
on the SD listserv. However, I would like to ""extend"" one of his points:

> >If we want the US government to be a beneficiary of SD modeling in the
future, it may >be very important to start here with XML-based languages
like StratML - if the US >government user is not squarely in the Society's
mission, XML considerations are less
> >important.

XML is THE international lingua franca for sharing information across
organizations and systems. Thus, if the SD community truly wants to share
information with others and broaden the usage of SD, rendering SD ""data"" in
XML format will serve that objective. If the SD community chooses NOT to do
so, that result speaks for itself.

The aim of AIIM's StratML Committee is to establish StratML as an
international voluntary consensus standard for potential use by all
organizations worldwide -- not merely U.S. .gov agencies. Organizations who
want to engage others in the pursuit of their goals and objectives will
gladly demonstrate leadership in the use an XML-based standard like StratML
.. or they will belie their avowed intentions.

Owen Ambur
Co-Chair Emeritus, xmlCoP
Posted by ""Owen Ambur"" <Owen.Ambur@verizon.net>
posting date Tue, 22 Apr 2008 11:20:33 -0400
_______________________________________________