Page 1 of 1

Resource allocation discussion

Posted: Thu Dec 23, 1999 3:28 pm
by "Slobodan P. Simonovic"
I am following discussion of the resource allocation problem with great
interest. I maybe missing something, but it looks to me that this problem
can be very easily solved using optimization (for example Linear
Programming). Why is the solution sought in the realm of SD? Is there any
aspect of the resource allocation problem that calls for SD approach in
particular? The original message by Steve Roderick that started this
discussion is asking for an approach for resource allocation (time is his
resource).

Cheers
Slobodan P. Simonovic
Professor and Director
Natural Resources Institute and
Department of Civil and Geological Engineering
University of Manitoba
Winnipeg, MB, Canada, R3T 2N2
phone (204)474-8375 fax (204)261-0038
e-mail slobodan_simonovic@umanitoba.ca
url http://www.fids.umanitoba.ca/~simon

Resource allocation discussion

Posted: Mon Dec 27, 1999 2:16 pm
by "Jim Hines"
Slobodan asks why all the hoopla over resource allocation problems that can
be solved easily using optimization methods. "Why," he wonders, " is the
solution sought in the realm of SD?"

Most people in the discussion are not interested in resource allocation per
se, but rather in a broader organizational problem such as what to do about
dropping morale or, how to improve lower-than-acceptable project
performance. There are two kinds of people with this focus: Those who
believe that the allocation currently going on in the organization is not a
source of problems and those who think it might be.

Those who think resource allocation is not itself a part of the problem are
looking for a simple formulation that will allocate some resource without
having the allocation scheme itself cause "phantom" problems in the
simulation. These folks want something that will work robustly within
their larger system dynamics model. Hence, they are trying to solve it in
the realm of SD.

The other folks are interested in representing a (potentially) harmful
allocation scheme in current use in the organization. They want to model
the essential way that managers actually allocate time or other resources.
Clearly an LP is not appropriate here (unless the managers actually use an
LP). These modelers want a simple formulation that captures the essence of
resource allocation, making the errors that real managers make, but not
making errors that real managers dont make.

If the people who are looking for an error-free formulation succeed, then
the folks who are interested in flawed formulations could go to town first
converting the error-free formulation to a behavioral representation and
then figuring out the ways that the good formulation can go astray. And,
conversely, if the folks who are looking for flawed formulations succeed in
finding a flexible formulation that can represent a wide variety of flaws,
the people looking for a non-error prone formulation can go to town figuring
out what the conditions are that reduce problems.

All this would be very similar to the way we represent supply-line control,
where we can either formulate a trouble free supply line, or we can
formulate one that will generate common problems.

Regards,
Jim Hines
jhines@mit.edu