Mgmt of Technology Course: SD & "fun"

[jm62004@Jetson.UH.EDU (Jaideep)]

A

Hello Iqbal,

Here are some thoughts.

1. One could teach an SD based course with examples from production, market,
distribution and financing questions in the management of technology for
the first two months or so, and then go on to optimization questions for
the next 2. There are three ways to go here:

a) Do optimization based on SD simulations. Do a large number of
manual or automated simulations to figure out how a particular measure
(costs/profits etc.) holds up. At the optimal value (found by, say, looking
at a graphical output), stop. This is simplistic, will not work for more
complex problems, is cumbersome and does not guarantee optimality.

b) In one course I had taken with Prof. Bruce Hannon, he taught
optimal control using Stella itself. This was neat except I will have to
dig my notes as to how it was done (he may want to comment here...). This
was in the later part of the course where students could grasp what optimal
control meant and had some clue as to what is involved with the ODE behind
Stella models.

c) Assuming the 400-level course is a graduate level management course
(and not undergrad junior/senior, who may lack math sophistication and/or
hate math and/or may not need this stuff), students have a working
knowledge of ODE (ordinary differential equations), you may want to use
other packages which could use Stella-generated ODE for optimal control.
Id prefer this approach because it leads to global/guaranteed optimality
and is applicable to "realistic" problems. To extend it further, if you are
looking for multiplayer optimization in dynamic situations, you are in
dynamic game territory (my joy and love;-), as this was part of my Ph. D.
work). Again, you will need a package which can solve "nonlinear two-point
boundary-value problems", after you have obtained costate equations, for
which you may want a symbolic solver such as Mathematica or Maple. Again,
this may itself require a full semesters worth of extra teaching, so may
not be appropriate. I am including all this, for the sake of (my idea of)
completeness, as you had asked about optimality and SD.

The above brings me to another point. It is an extension of the debate on
SD theory we had on this list a few weeks/months back. For some real-world
problems/applications, I maintained that some understanding of ODE is
required as that is the basis for SD. (I had the honour of a response by
Prof Forrester and others, indicating that a lot of practical experience is
needed to be good at SD, just as in other professions, and just a
theoretical knowledge of ODE will not make a good system dynamicist - I
agree). My problem was: we still need a good grounding in some theory to be
good at applications, and the latter could take even decades to perfect.
The problem with teaching SD sometimes is: we try to make it a lot of
"fun", and that takes away the seriousness of the theory part. I heard
Professor John Lienhard today morning on UH Radio, and he expresses my
sentiments much better. Please see
(http://www.uh.edu/campus/kuhf/engines/engines.htm)

I feel strongly about SD teaching in a similar way, as many times the
emphasis is on a teacher to make SD "fun" without a corresponding effort on
the part of students to work hard and learn the principles behind SD. The
tragedy is more prevalent while teaching unprepared, unmotivated
undergrads, though some graduate students also suffer from the illusion
that if the class wasnt "fun", or if they did not understand the rudiments
in *one* lecture, somehow the teacher was not up to snuff and they have
ZERO responsibility. I may also add I am not talking here of MIT-type engg
students, to which this wont apply. The majority is non-MIT, non-engg type
to whom we may want to promote/teach SD. (A similar argument applies to
many other areas, where commercialization/selling/facade is more important
than the content, but that is a different topic...)

I know Prof. Forrester, and others, are involved with making SD accessible
to the young. Please comment from your experience. How does one do it,
without diluting the science and giving explanations ("friction" not
"momentum") such as mentioned above. I am perplexed.

Sorry for the long (partly tangential) reply, but this has concerned me
ever since I taught some grad/undergrad courses. How should I tailor my
teaching and what should be the responsibility of students so that good
learning and teaching occurs?

Regards and thanks

Jaideep
jm62004@Jetson.UH.EDU

====================================================================
Jaideep Mukherjee, Ph. D.
Research Associate
Industrial Engineering
E210, D3, Cullen College of Engineering
University of Houston
Houston, TX 77204-4812

http://www.uh.edu/~jm62004/jm97.html
Ph: 713 743 4181; Fax: 713 743 4190

====================================================================

[ijamal@gov.edmonton.ab.ca (Iqbal]

Jaideep: Thanks

I was going to use SD as an organizing paradym first (to help the students
understand the overall dynamics
elationships in the field of study) and then explore quantatitive solutions
using SD and optimization approaches. Does that sound reasonable? What problems should I expect/foresee?
I had planned to have students do field projects to help solidify the approaches covered.
The course is a senior undergraduate with some MBA students.

Your comment on SD being "fun" is well taken. Unfortunately, being an practitioner (in the municpal field)
most decisions are made in fuzzy systems, with fuzzy data, and limited time horizons. As such, the level of rigor/
resolution that you speak often goes far beyond that required to make "reasonable near-optimal" decisions.
I find that the major effort is in trying to convince decision-makers of other perspectives (to their own) that
will provide them a greater insight into their operations. I believe that the "breakthrough" in thinking is more
important than the rigor/optimality of the solution. And so Id be willing to give up a bit on optimality (it still has to
be the "right" answer) for elegance in "breakthrough" (defined for the business decision-maker). For all the modeling
that we did on the infrastructure question, the greatest benefit were the "a-has" we got from our clients. At that point
they were more receptive to the possibilities, and then the results.
Comments? Cheers

Iqbal
ijamal@gov.edmonton.ab.ca