Professional Documents
Culture Documents
Modeling
mal study of existing models. Little [1970] lyzed. In contrast to teaching such ready-
discussed how the then emerging technol- made models, I, like Morris, am concerned
ogy of interactive computing could be em- with the processes of discovery and elabo-
ployed to develop models that managers ration that are essential parts of modeling
would be likely to use. Hugh Miser and and of model development. I am inter-
Ed Quade had much to say on the subject ested in the ways people build and use
in their magnum opus on craft issues in models, rather than in the details of indi-
systems analysis [Miser and Quade 1988]. vidual, ready-made models. Modeling ac-
Hodges [1991] argued that even bad mod- tivity is at the technical heart of OR/MS
els may be used in satisfactory ways, even practice.
if those models fall short of their creator’s Models in OR/MS have two main uses:
original intentions. Pat Rivett has written (1) People use models to explore the pos-
much on the subject of modeling, and sible consequences of actions before they
Rivett [1994] provided a number of exam- take them. An example of this is the evalu-
ples to illustrate a general approach to ation of weapons systems in defense
model building. Powell [1995] discussed [Cannella, Sohn, and Pate 1997].
how modeling skills may be taught to Boothroyd [1978] called this “reflection be-
MBA students and suggested six key mod- fore action.” Used in this way, a model is
eling heuristics for this purpose. a convenient world in which one can at-
I intend to complement the issues that tempt things without the possible dire
these and others have raised. I also hope consequences of action in the real world.
that I may stimulate other people to con- In this sense, models become tools for
sider their own key aspects of modeling in thinking. This thinking might relate to
OR. In discussing these principles, I will one-time events, of which an example
rely on my own background in discrete might be a particular capital-investment
simulation. However, I believe the princi- decision. Or the thinking might concern
ples I discuss are relevant for most forms occasional events, such as pricing reviews.
of mathematical and computer modeling Alternatively, the thinking might concern
in OR/MS. Some of the material in this routine events, as in weekly production
paper is based on chapter 4 of my book planning. We also use models as tools for
[Pidd 1996]. thinking when we try to understand a
Models and Modeling complex system, even if we contemplate
I will focus on modeling as a verb or as no immediate action.
an activity and not models as nouns or (2) The other broad use for models in
subjects. Morris [1967] wrote that “the OR/MS is as part of embedded computer
teaching of modeling is not the same as systems for routine decision support. This
the teaching of models.” By models he use is very common in the management of
meant approaches and methodologies, logistics and supply chains [Arntzen et al.
such as those of linear programming or 1995]. When used in this way, the models
queuing theory, that present ready-made form an essential and automatic part of
models by which situations may be ana- the management and control of an organi-
zation. This does not mean that there will plored and manipulated. Checkland’s sec-
never be human intervention, but it does ond type of model includes the conceptual
mean that the model or modeling system models used in soft systems methodology
is usually intended to get things right. (SSM). These models are intended to em-
I discuss models as tools for thinking, body the elements that should be present
that is, as ad-hoc exploratory devices for in an idealization of a system under scru-
reflection before action. tiny [Checkland 1981]. Checkland’s
“should be” comes from systems theory
Models are no substitute for and its notion that any viable system will
thought and deliberation. include a number of essential activities,
such as those needed for control by feed-
With this in mind, I define a model in back. Thus a conceptual model, as used in
OR/MS as follows: A model is an external SSM, serves as a basis for debate about the
and explicit representation of part of real- differences between the situation as it is
ity as seen by the people who wish to use now and the situation as it might become.
that model to understand, to change, to In this sense, Checkland’s conceptual
manage, and to control that part of reality model clearly need not be a representation
in some way or other. of reality. I focus here on the first type of
This definition has a number of impor- model, those that are would-be representa-
tant features. First, the model is external tions of the real world.
and explicit rather than existing as a set of My definition’s third feature is the as-
mental constructs that are not accessible to sumption that no model as used in
other people. This means that the model OR/MS will be a complete representation
can be examined, can be challenged, and of reality. If it were, then it would be as
can be written in a logical language, such complicated, as expensive, as hard to ma-
as that of mathematics or of computer pro- nipulate, and as disastrous when things go
gramming, and it may even serve as a wrong as reality itself. Instead, the repre-
form of organizational memory when a sentation is partial, and the partiality is
model-building team is in action. governed by people’s intended use of the
Second, the model is a representation of model. They want it to be fit for some pur-
the real world. Checkland [1995] distin- pose. These people want to use the model
guished between two types of model. First to understand, to change, to manage, and
are those intended to be representations of to control that part of reality in some way.
the real world, and these are, I suspect, Thus, the modeling is goal-oriented. This
typical of those employed in OR/MS. This does not, however, imply that the partici-
does not mean that the models are com- pants necessarily agree about the intended
plete or as complex as the world they rep- use of the model.
resent. It does, however, mean that they The difficulty with assuming that mod-
are intended to represent certain aspects of els are representations of even part of real-
the real world. They then become surro- ity is that people may differ in what they
gate forms of reality that can be safely ex- regard as reality or may disagree over
what part of reality to model. To appreci- must be able to match the full variety of
ate this, one need only realize that models the system that it is controlling if it is to be
are always simple, whereas realities are al- of much use. Thus, a device to control the
ways complex. An old mathematical joke temperature of a domestic furnace must be
about numbers helps to illustrate this. able to detect and act when the furnace
Complex numbers have two parts—the gets too cold as well as when it is too hot.
real and the imaginary. A system being Ashby took this common-sense notion and
modeled is like a complex number, with developed it mathematically.
real and imaginary parts. The problem is Does this mean that a model must be as
that different people see things in different complicated as the system it represents?
ways; one person’s reality looks like an- Thankfully, no. Were the answer yes,
other’s imagination, and vice versa. Thus a modeling would be uneconomic, since a
complex system may be impossible to model would take as long to build as the
model in all its complexity. system it represented—and it would be
With these ideas in mind, I consider six just as expensive to develop and control.
simple principles of modeling: The model alone need not satisfy
(1) Model simple; think complicated. Ashby’s principle; rather the system that
(2) Be parsimonious; start small and add. comprises the model and the user(s) must
(3) Divide and conquer; avoid match this variety. That is, in systems
megamodels. terms, the model:user(s) system displays
(4) Use metaphors, analogies, and emergent behavior that must match that of
similarities. the system being modeled. This notion ex-
(5) Do not fall in love with data. plains this first principle of modeling:
(6) Model building may feel like mud- model simple; think complicated. Models
dling through. are no substitute for thought and delibera-
My own technical expertise in discrete tion; they are part of a process of reflection
simulation is bound to color my choice of before action. That process must have req-
principles, but I believe they are generally uisite variety, not the model alone. Little
applicable. [1970] made a similar argument for sim-
Principle 1: Model Simple; Think plicity when considering how managers
Complicated actually use models. He argued that peo-
My definition of a model suggests that ple use models as a means of decision
models are simple representations of com- support and that the model:user system is
plex things. How then can be they be ade- a form of man:machine system more pow-
quate devices to support reflection before erful than the human or the model alone.
action? A particular problem is the need This principle means that a relatively
for variety noted by Ashby [1956], of simple model can support complicated
which one statement is that “variety must analysis. A model intended as a tool for
match variety.” Ashby’s principle of requi- thinking needs to be adequate for the task,
site variety stems from tenets of control and it must be skillfully used if the task is
theory, which state that a control system to be done well. It implies a mind-set dif-
ferent from that occasionally evident chain saw: used properly it is a very use-
among OR/MS people. For example, at a ful tool; used without training it can cause
conference I attended, a plenary speaker, a considerable damage, some of it to the
distinguished and senior manufacturing user. The opposite is also true: a powerless
manager of one of the world’s most suc- but easily manipulated model (one that
cessful companies, told the audience that does not fit its intended purpose) is like a
he wanted a model to simulate all his chain saw with no cutting chain. It makes
company’s worldwide factories at almost a lot of noise but isn’t of much use.
any level of detail. As he was our distin-
guished guest, it would have been rude to “To a point, Lord Copper, to a
disagree. The appropriate response, point!”
though, is that given by Evelyn Waugh’s
character Boot to his boss, Lord Copper, in A simple model does not have to be a
the novel Scoop. Whenever Lord Copper small model. In mathematical terms, sim-
expressed some outrageous view, Boot’s plicity can be regarded as the close rela-
polite response was, “To a point, Lord tive of elegance. Knowing when to sim-
Copper, to a point!” No model can or plify requires considerable understanding
should do everything. of the system being modeled and of the
Why is simplicity desirable? Little [1970] tools of modeling. In wrestling with theo-
argued that models should be transparent logical problems, William of Ockham, a
(that is, simple to understand, at least in British monk who lived around 1200 AD,
outline form) and should be easy to ma- argued that “a plurality (of reasons)
nipulate and control. Transparency is de- should not be posited without necessity.”
sirable because successful OR/MS practice This principle, “Ockham’s razor,” is an ar-
depends on trust between consultant and gument for simplicity of explanation, the
client. Trust is easier to establish if the cli- parallel of simplicity in representation.
ent can appreciate the overall workings of Simplicity can never be an end in itself.
the model and understand what the model Sometimes complex models are needed,
can and cannot do and why. However, especially when they are to be part of em-
this notion of transparency does not imply bedded real-time systems. Most modern
that a management science model must be jetliners are at least partially automated
limited by the technical prowess of the and include avionics systems that control
people sponsoring the work. the aircraft. These systems and their mod-
Ease of manipulation is a goal that is els must meet Ashby’s principle of requi-
much easier to satisfy now than it was for site variety, otherwise automatic landing
Little in 1970. We take interactive com- and flight would be impossible with any
puter facilities for granted, and their use, degree of safety. However, Little’s notion
especially with visual interactive ap- of transparency still holds good: the pilot
proaches, can make decision models very should understand how the model of the
easy to manipulate. This brings a danger flight surfaces is intended to perform,
in its train. A powerful model is like a even if he or she does not understand the
nut to crack. We draw an analogy when something that may be currently infeasible
we point out the agreement or correspon- as a way of directing attention toward the
dence in certain respects between two dif- important features of the issues under con-
ferent things. In OR/MS modeling, we sideration. An example might be some-
might deliberately draw an analogy be- thing like, “OK, let’s suppose that we have
tween, say, the demands on an emergency some way of instantaneously making any
service and demands in a queuing system. blood type available. What then would be
The two are not the same, though both the important features of a blood-
may be regarded as having servers and transfusion service?”
customers, and they have enough similari- It seems likely that analogies are most
ties that we can transfer learning from one useful in the initial stages of modeling
to another. This use of analogy is close to [Morris 1976]. They can be used to illumi-
the notion of model enhancement by asso- nate the development of the initial simple
ciation [Morris 1967]. models suggested by the principle of
Synectics, developed at Arthur D. Little parsimony.
and described by Evans [1991], is a more Principle 5: Do Not Fall in Love with
general attempt to use analogies by en- Data
couraging participants to draw different The availability of friendly computer
types of associations in tackling an issue. packages has produced a generation of
Synectics suggests three analogies relevant people who are hooked on data, data
to OR/MS: junkies. I am not arguing that data and its
(1) In personal analogy, participants imag- analysis are unimportant and can be ig-
ine themselves inside the systems being nored. Rather, I observe that many stu-
modeled. This is a very common approach dents seem to imagine that modeling is
in discrete simulation in which modelers impossible without data, preferably lots of
may try to imagine the states through it. They treat data as a form of Linus’
which important system entities pass. For blanket. They assume that, because a
example, they might imagine themselves model is a representation of some system,
as cars finding their way through a con- examination of data from that system will
gested road network via a sequence of reveal all that they need to construct the
junctions. model. Such an assumption may be a mis-
(2) Direct analogy is the type drawn take, even though exploratory data analy-
above between an emergency service and sis is very useful, and I would not wish to
a straightforward queuing system. The return to the days before modern com-
analogy need not be perfect (indeed it puter software appeared for this purpose.
could be argued that it can never be per- The slapdash use of exploratory data anal-
fect), but the idea is to transfer lessons ysis can never replace careful thought and
learned in one sphere to another. analysis.
(3) Fantasies, which are not really analo- Modeling should drive any data collec-
gies, permit participants to stretch the tion and not the other way around. The
boundaries and imagine the occurrence of modeler should think about the type of
model that might be needed before at- answers form the preliminary data, which
tempting large-scale data collection. is the appetizer before the main course. I
Conway et al. [1995, p. xx] gave some very also call these the six idiot questions, since
sound advice about the use of data in the they can be asked only in the early stages
discrete simulation of manufacturing sys- of a project. Ask them too late, and your
tems, including the following assertion: client thinks you are an idiot.
The verse also illustrates the important
You should resist the temptation, or the in-
structions, to collect real data with all the inge-
point that preliminary data are not just
nuity you can muster. The collection process is quantitative; the qualitative is important
much less valuable than you might think. in framing the issues to address. This is
not to say that a preliminary quantitative
Although this is an extreme view, it is a
analysis is a waste of time, but numbers
useful counter to the familiar greeting of
are not enough.
the naive modeler to his client, “Take me
Beware of data brought on a plate—order
to your data.” Only after thinking about
à la carte, not table d’hôte
the model can the analyst know what type
With modern information systems,
of data to collect.
many organizations are awash with data,
Go for an appetizer before the main
but it is worth remembering that old ad-
course
age of IS: information is data plus inter-
It can be very helpful to recognize that a
pretation. When taking data straight from
modeling exercise may require two types
a corporate IS, one risks misinterpreting it.
of data, both of which may be qualitative
In this way, MIS information becomes
as well as quantitative. Preliminary data,
misinformation. In today’s global organi-
collected early in a modeling project, form
zations, many people work in virtual
part of the problem structuring, during
teams and this issue becomes particularly
which the modeler frames and names the
important. Just like liquids to be pumped,
important issues. The modeler uses pre-
data put into an MIS are usually filtered
liminary data to get an idea of what type
and cleaned up. This filtering and cleaning
of model is needed.
may have such an effect that the data are
Being British, I often remind my
of little use in analysis and modeling.
OR/MS students of Rudyard Kipling’s
As an example, suppose a hospital
verse from the Just So Stories:
wished to reduce its waiting lists and was
I keep six honest working men proposing to model some aspects of its
(They taught me all I knew); operation in order to decide whether to do
Their names are What and Why and When so. One might think that the best way to
obtain data about patient referrals to the
And How and Where and Who.
hospital would be to examine the referral
Much loved by industrial engineers un- database that contains information about
der the heading of critical examination, actual referrals. However, this data might
the verse suggests six questions that are be very misleading for a number of rea-
important in structuring problems. Their sons. First, doctors who refer patients to
the hospital may have a choice of hospitals analysis dies of starvation and it is too late
and may know how long their waiting for force feeding. The other reason for
lists are. They may refer their patients to such apparent arrogance may be, para-
hospitals with short lists. Slightly more doxically, that we feel inadequate in the
subtly, they may refer them to hospitals arena in which we are invited to work.
that they believe have short lists. Further, Thus, we shy away from engagement and
the waiting lists are a symptom of a prob- end up without enough data of the type
lem, not the problem itself. As doctors can we need.
tell you, they give symptomatic or pallia-
tive treatment only when there are no Numbers are not enough.
other options; if possible they treat the un-
derlying causes. Thus, it may be better to An equivalent of gluttony is seen as an
treat the waiting lists as outputs from a obsession with data, especially in huge
model rather than as inputs. quantities. Sad to say, modern computer
Eat healthily—avoid anorexia and systems may encourage this. For example,
gluttony it is possible to work through huge data
People on a starvation or subsistence sets even if they are geographically dis-
diet must find it perverse that so many persed. Because this can be a fascinating
people in modern Western society have thing to do, it may become something we
eating disorders. Two such disorders are wish to do. But the real question to ask is,
anorexia, a form of deliberate self- “Is it necessary?” It may be better to work
starvation, usually because of a false body with a single detailed data set and then
image and low self-esteem, and gluttony take samples from others to check that the
in which excessive food intake leads to detailed set is representative. Careful plan-
obesity. Without stretching this food anal- ning and use of data are as important in
ogy too far, there are clear parallels in the OR/MS as are exercise and a balanced
collection and use of data for OR/MS diet to us as people. It may even be a
modeling. good idea to have a data fast, much as
The equivalent of anorexia occurs when suggested in the earlier quote from
analysts or groups decide that they know Conway et al. [1995].
best. Although I am convinced that the Other foods are available—if you ask
model should drive the data collection In collecting and analyzing data, one
and not vice versa, sometimes this can be must remember that any data are just
dangerous. We can be tempted to disen- samples from some population. When
gage from the world in which our clients asked for a forecast, cynical economists
and customers work. We may feel that we advise a golden rule, “Give a number or a
know best: “We’ve done so much work in date, but never both!” The data used in
this area that we can pretty easily translate modeling should always be dated. When
previous projects into something workable we use data to build or to test a model,
here.” Sadly, this false image of our own that data will have been collected at a par-
prowess may be exposed only when the ticular time and over a certain period. If
we say the data are believed to be repre- modeling in our minds. The reality is very
sentative, we imply that the data are rep- different. As a parallel, consider the pro-
resentative of a larger population that dis- cess of PhD research and the final thesis. It
plays some regularity through time. We seems unlikely that most such theses de-
do not expect someone to surprise us later scribe how the student actually spent his
with drastically different data. Neverthe- or her time during the years of research.
less this may happen, and it is always a Instead, the student presents a summary
risk, especially when we use the results of within a strictly defined logical frame-
a model to extrapolate into the future. The work. This presents a problem for many
future may simply differ from the past, PhD students, since they have to place ra-
and the population from which the data tional constructions on a process that was
sample comes may behave differently in probably shot through with intuition,
the future. hope, and despair. A pretense that model-
Data also form a set of observations. ing is a rational and linear process may
The modeler must realize that data are create similar problems, especially for
samples of what he or she might obtain newcomers.
given enough time and other resources. Willemain [1994, 1995] set out to find
Any observation process is subject to er- how expert modelers actually work and
rors of different types. It is reasonable to found that they rarely operated like Mr.
be skeptical about the resulting data, espe- Spock in Star Trek. His work was in two
cially if it is readily available. In a persua- parts, both with experienced OR/MS prac-
sive example, Morse [1986] described the titioners and academics. In the first stage,
early operations research efforts in war- he interviewed them to ask them to give
time, including this quote from a pilot accounts of how they went about their
who had been asked to provide reports af- modeling work. Interestingly enough,
ter each mission. “Hell, I didn’t think any- their accounts had much in common, and
one ever read those damned reports!” On they claimed the following. They [Wille-
hearing this, Morse and his team chose to main 1994, p. 214] “develop their models,
collect their own data. not in one burst, but over an extended pe-
Principle 6: Modeling May Feel Like riod of time marked by heavy client con-
Muddling Through tact.” Also, they are “guided by analogies,
Muddling through is a very British con- drawing and doodling, they develop more
cept that seems to go with the idea of the than one alternative model, in each case
gifted amateur. It is also a fairly accurate starting small and adding.” Many of their
description of how it feels to be build claimed approaches follow the principles I
models in OR/MS. It is tempting to see have discussed.
model building as a linear process in In the second stage of his research,
which we move from step 1 to step 2 to Willemain placed the modelers individu-
step 3 and so on. We might concede that ally in a room, set a modeling task, and
some people may need an extra step 2a, asked them to try to develop a suitable
but we often carry this simple view of model, all the time talking through what
they were doing. There can be no certainty relief from justifications of why OR/MS
that their talking accurately reflected their students must know about the various
thinking; however, analysis of their talk- linear-optimization algorithms. Other peo-
ing revealed some interesting things. ple will probably have their own lists that
The tapes show that the modelers de- differ from mine and the others I’ve men-
voted about 60 percent of their time to tioned. I would like to hear from them. I
model structure, that is, what we would hope we can embody these and other
regard as the core of model building. They principles in the courses we teach. Stu-
divided 30 percent of the time equally be- dents need to realize that learning the
tween problem context and model assess- skills of modeling may be more important
ment. They devoted just 10 percent to than learning about models.
model realization and almost none to This paper is also a plea for some seri-
implementation. ous research about how people go about
In the study, the modelers had just 60 their modeling. Tom Willemain made a
minutes to work on a problem, so it is not start in this area. With Bob Woolley, I did
surprising that they devoted so little time some small-scale empirical work in the
to model realization or implementation. late 1970s, but much has changed since
But what is significant is that they spent then, both in the organizational world and
so much time thinking about problem con- in the tools available to support model de-
text and model assessment. What is also velopment. Smith [1989] proposed a pre-
very important is that the time they spent scriptive framework, based on cognitive
on the three major concerns was scattered psychology as well as on OR/MS, which
through the modeling session. They kept makes some sense of modeling as an activ-
picking up a concern for a while, drop- ity. Perhaps someone could take some of
ping it, and then returning to it. Presum- these ideas and conduct more empirical
ably this pattern would be even more research to see if there are better ways of
marked were it possible to follow how doing things.
they operate over a much longer period in
References
their real work. Muddling through is quite Ackoff, R. L. and Sasieni, M. W. 1968, Funda-
a good description of how these experi- mentals of Operations Research, John Wiley and
enced people went about their work. Sons, New York.
Arntzen, B. C.; Brown, G. G.; Harrison, T. P.;
An Invitation and Trafton, L. L. 1995, “Global supply chain
Like any book, this “rough guide” must management and Digital Equipment Corpo-
come to an end. Like all authors of such ration,” Interfaces, Vol. 25, No. 1, pp. 69–93.
tourist guides, I can describe only the Ashby, R. 1956, An Introduction to Cybernetics,
Chapman and Hall, London.
main features of the land being visited. Axelrod, R. 1976, Structure of Decision: The Cog-
Most such authors exhort their readers to nitive Maps of Political Elites, Princeton Uni-
send in comments and corrections. I do versity Press, Princeton, New Jersey.
Boothroyd, H. A. 1978, Articulate Intervention,
too.
Taylor and Francis and the Operational Re-
Many writers have proposed principles search Society, London.
of modeling, all of which are a welcome Brookes, F. P. 1975, The Mythical Man-Month:
Essays on Software Engineering, Addison Wes- Rivett, B. H. P. 1994, The Craft of Decision Model-
ley, Reading, Massachusetts. ling, John Wiley and Sons, Ltd, Chichester,
Cannella, D. A.; Sohn, S. Y.; and Pate, C. 1997, England.
“Evaluating new weapon systems technol- Rosenhead, J. V., ed. 1989, Rational Analysis for
ogy: The Javelin versus the Dragon,” Inter- a Problematic World, John Wiley and Sons,
faces, Vol. 27, No. 3, pp. 29–38. Ltd, Chichester, England.
Checkland, P. B. 1981, Systems Thinking, Systems Saaty, T. L. 1998, “Reflections and projection’s
Practice, John Wiley and Sons, Ltd, Chiches- on creativity in operations research and man-
ter, England. agement science: A pressing need for a shift-
Checkland, P. B. 1995, “Model validation in soft ing paradigm,” Operations Research, Vol. 46,
systems practice,” Systems Research, Vol. 12, No. 1, pp. 9–16.
No. 1, pp. 47–54. Smith, G. F. 1988, “Towards a heuristic theory
Conway, R.; Maxwell, W. L.; McClain, J. O.; of problem structuring,” Management Science,
and Worowa, S. L. 1995, User’s Guide to Vol. 34, No. 12, pp. 1489–1506.
XCELL Plus: Factory Modeling System, Release Smith, G. F. 1989, “Defining managerial prob-
4.0., Scientific Press, Palo Alto, California. lems: A framework for prescriptive theoriz-
Eden, C. L.; Jones, S.; and Sims, D. 1983, Mess- ing,” Management Science, Vol. 35, No. 8, pp.
ing About in Problems, Pergamon Press, Ox- 963–981.
ford, England. Willemain, T. R. 1994, “Insights on modeling
Evans, J. R. 1991, Creative Problem Solving in the from a dozen experts,” Operations Research,
Decision and Management Sciences, South Vol. 42, No. 2, pp. 213–222.
Western Publishing Company, Cincinnati, Willemain, T. R. 1995, “Model formulation:
Ohio. What experts think about and when,” Opera-
Hodges, J. S. 1991, “Six (or so) things you can tions Research, Vol. 43, No. 6, pp. 916–932.
do with a bad model,” Operations Research, Zeigler, B. P. 1976, Theory of Modeling and Simu-
Vol. 39, No. 3, pp. 355–365. lation, John Wiley and Sons, Inc., New York.
Little, J. D. C. 1970, “Managers and models: Zeigler, B. P. 1984, Multi-facetted Modeling and
The concept of a decision calculus,” Manage- Discrete Event Simulation, Academic Press,
ment Science, Vol. 16, No. 8, pp. B466–485. New York.
Miser, H. J. and Quade, E. S. 1988, Handbook of
Systems Analysis: Craft Issues and Procedural
Choices, John Wiley and Sons Ltd, Chichester,
England.
Morris, W. T. 1967, “On the art of modeling,”
Management Science, Vol. 13, No. 12, pp.
B707–717.
Morse, P. M. 1986, “The beginning of opera-
tions research in the United States,” Opera-
tions Research, Vol. 34, No. 1, pp. 10–17.
Pidd, M. 1996, Tools for Thinking: Modelling in
Management Science, John Wiley and Sons,
Ltd, Chichester, England.
Pidd, M. and Woolley, R. N. 1980, “Four views
on problem structuring,” Interfaces, Vol. 10,
No. 1, pp. 51–54.
Powell, S. G. 1995, “The teacher’s forum: Six
key modeling heuristics,” Interfaces, Vol. 25,
No. 4, pp. 114–125.
Raiffa, H. 1982, “Policy analysis: A checklist of
concerns,” PP-82-2. International Institute for
Applied Systems Analysis, Laxenburg,
Austria.