-282-
Strategic Microworlds and System Dynamics Modelling 1
John D W Morecroft
Business Policy Group
London Business School
ABSTRACT
In the past ten years, system dynamics has become more accessible to
managers and more applicable to strategic issues. The paper reviews
developments in software, theory, gaming and methods of simulation
analysis that have brought about this change. Together these
developments allow modellers to create computer-based learning
environments (or microworlds) for managers to “play-with" their
knowledge of business and social systems and to debate strategic
change.
INTRODUCTION
In the past ten years there have been several important developments
in system dynamics which make the subject more accessible to managers,
more applicable to strategic issues and more challenging for research.
There have been improvements in the symbols and software used to map _
and model system structure. New ideas have been adopted from
behavioural decision theory, which help to capture managers' knowledge
in computer models. There have been improvements in methods of
simulation analysis that enable model users to gain better insight
into dynamic behaviour. Greater emphasis has been placed on small
models incorporating managers' knowledge and on dialogue between
“mental models" and computer models.
As a result of these developments, system dynamics can now be used,
with a management team, to structure informed debate about strategic
change. In this process, models and computer simulations are an
integral part of management discussion. The paper explores each of
the major developments in more depth in order to show the range of
ideas and concepts that system dynamics now encompasses. The paper
concludes with some thoughts on future research.
SYSTEM DYNAMICS - A MICROWORLD FOR DEBATING STRATEGY
What is a "microworld"? for strategy debate? Figure 1 shows the many
elements in the microworld provided by system dynamics. At the top
left is a problem or issue facing managers which leads to debate and
dialogue. The dialogue results in clarification of the problem or
issue and eventually to recommendations for action. The microworld
contains all the elements included in the discussion. A most
important factor is the managers' own knowledge (or mental model) of
the business or social system. This knowledge provides the raw
material for discussion. In conventional policy-making (by means of
argument) it is the interplay between the knowledge base and the
discussion that produces a consensus for action.
-283-
Business Issue Policymakers’
or Problem Knowledge Information Feedback
Debate and
Dialogue
Action
Theory
symbols and rules for
mapping
Text
Diagrams
Algebra
wupse
Simulations
Behavioural Decision
Theory
guidelines for specifying
information flows
Other information about
Business System THE SYSTEM DYNAMICS
MICROWORLD
Problem Knowledge
( | {/. ) Theory
Debate Map ——~
| —)
Action Information
Figure 1 The Microworld for Policy Debate Provided by System Dynamics
~284—
When modeling and simulation enter the debate, the picture becomes
more complex and the interplay of knowledge, information and
discussion becomes more productive. Managers' knowledge, and other
information about the business (staff reports, financial documents
etc) are converted into text, diagrams, algebra, and simulations.
This process of mapping knowledge and information is guided by the
theory and concepts of system dynamics. The figure shows two main
inputs from theory. The first input, from information feedback
theory, provides symbols for diagramming a system and rules for
mapping. As readers know well, these symbols include "levels",
"flows", "flow regulators" and "converters" to represent physical,
financial and decisionmaking processes. The rules for mapping
include rules for connecting the symbols, guidelines for equation
formulation and guidelines for simulation and analysis. The second
input, from behavioural decision theory provides the modeller with
guidelines for specifying a model's information flows. It helps
modellers to ask the "right" questions of managers and so capture in
diagrams the managers' knowledge of the system's operating structure.
The microworld includes knowledge (K), information (I), theory (T),
maps (M), debate (D) and the interplay of these factors as summarised
in the inset of figure 1.
The scope of policy discussion is potentially greater than can be
achieved by conventional argument. The maps, (text, diagrams, algebra
and simulations) provide managers with a variety of perspectives on
their pooled knowledge. The maps also draw information from reports
and staff. So the interplay of discussion and knowledge is enhanced
through increased variety of representation, more information, and
additional paths of interaction. Moreover, the content of the maps
themselves is guided by information feedback theory and behavioural
decision theory.
Now let us turn to the developments in system dynamics which have made
possible this microworld for debating strategy.
REVIEW OF MAPPING METHODS
One of Forrester's (1961) major contributions to modelling was to
adapt abstract analytical methods from classical control theory into a
flexible form suited to modelling and discussion in the business and
social arena. He created symbols for mapping systems together with
rules for connecting the symbols and converting them to algebra.
The main symbols for mapping are shown in figure 2. I will assume
that all readers are familiar with the symbols, so I will not explain
them further. My point here is simply to note that using these few
symbols one can create a visual representation of an organization
which provides a basis for discussion with a management team.
Moreover, after converting the map or diagram into algebra, one can
use simulation to obtain a visual representation of dynamic behaviour.
System dynamics is a highly graphical subject, whose diagrams and
graphs provide a focal point for discussion and learning in a
management team.
-285~
Decision
Function
ieee =
or
Converter Output: Action or
Information
7
/
a A
’ 1
ot \
? ‘
of ‘
A ‘\
ant \
Input: Information ‘
or Influence
Flow Regulator
2)
-—T_ Level or
——EE_ action
) accumulator
Source Action Flow
Figure 2 Symbols for Mapping
-286-
IMPROVEMENTS IN SOFTWARE:
Until recently it has been cumbersome and time-consuming to create
diagrams and graphs, so the visual power of the subject has been
underutilized. However, the arrival of graphic computers like the
MacIntosh has now made it possible to draw symbols directly onto a
computer screen and to edit diagrams interactively. The modelling and
simulation package STELLA (Richmond et al 1987) provides the modeller
with a menu of symbols for creating a diagram on an electronic
worksheet. The symbols include those shown in figure 2 and several
others that help in organizing and connecting the elements of the
diagram. One can select symbols from the menu, move them onto the
computer screen (a small part of the available electronic worksheet),
connect them and edit them. The software provides a very effective
(and entertaining) medium for capturing managers! knowledge. Better
computer graphics have also made it possible to create visually clear
simulation runs that are much easier to read and quicker to prepare
than the old character plots that were common only five years ago.
NEW CONCEPTS FROM BEHAVIOURAL DECISION THEORY
With the symbols and mapping rules of system dynamics it is possible
to create quite complex networks of decisionmaking processes. But
there are innumerable ways to link the symbols which all obey the
connection rules of feedback systems. However, only some symbol
configurations correspond to realistic decisionmaking structures.
There is a need for modellers to be discriminating in their choice of
information links and influences if they are to produce plausible and
insightful strategy models.
Recently, system dynamics has adopted concepts from behavioural
decision theory that are useful for specifying information links among
decision functions. (Hall 1984, Morecroft 1985, Sterman 1985).
Behavioural decision theory focuses on the information and heuristics
used in real-life decision making. What information receives
attention in organizational decisions? What information is ignored,
and why? What factors condition the quantity and quality cof this
information? Behavioural decision theory concludes (with plenty of
empirical evidence) that people make choices using only a few sources
of information processed with simple rules of thumb. So the network
of information flows in a realistic organization is quite sparse
relative to the network that would exist if each decisionmaker used
information from every source in the system.
Figure 3 shows how behavioural decision theory guides the mapping of
decisionmaking processes. One can see in the figure the standard
feedback representation: decision function - action flow - level -
information - decision function. In addition there are many other
information flows and influences (originating from other levels in the
system) which are shown on the outer boundaries of the decision
function. Only a few of the information flows actually penetrate to
the heart of the decision function where they influence the choices
and actions of the players' (individuals, groups, subunits). The
concentric circles surrounding the decision function represent
organizational and cognitive filters which select or limit the
information made available to decisionmakers at different points in
-287-
Flow Regulator
C3 7 > Level
Action Flow
Decision
Function
People's cognitive limitations
Operating goals, rewards and incentives
Information, measurement and communication systems
Organisational and geographical structure
Tradition, culture, folklore, leadership.
new eS
Figure 3 ihe Behavioural Decision Function - Decision making and Information
Filters
—288-
the system.
There are five filters surrounding a decision function. The first
filter represents people's cognitive limits. People are unable to
process all the information that a business or social system may
present to them. They make their judgements on the basis of a few
dominant sources of information processed according to quite simple
rules of thumb.
The outer filters (2,3,4 and 5) in figure 3 represent the ways in
which an organization conditions the information made available to
decisionmakers. This part of the figure draws particularly on Simon's
Administrative Behavior (1976) which explains how organizations may
display effective decisionmaking despite the cognitive limits of
managers and an over-abundance of information. Simon identifies
organizational processes which are designed to simplify
decisionmaking tasks. All employees make their judgements and
decisions in a "psychological environment" provided by the
organization. The psychological environment limits the range of
factors considered and, in principle, supplies only the relevant
information (a tiny subset of the total information available in the
system) for making the correct decision. The filters show the
components of the “psychological environment" and they also provide a
convenient basis for questioning managers.
Filter number 2 represents the influence of operating goals, rewards
and incentives on information flow. Decisions and actions in business
and social systems depend on the operating goals and rewards faced by
the key players in the system. One can only understand organizational
choice and action relative to these goals and rewards. So, for
example, it is well-known that factory managers who are held
accountable for a specific end-of-year inventory target will
drastically curtail or boost production to meet the target, in
defiance of "rational" cost-minimising scheduling criteria. For these
factory managers, information about the status of inventory easily
penetrates filter number 2. The filter excludes other information on
future expected demand, cost structure and capacity constraints, which
together with information on inventory would be required to set a
rational production schedule.
Filter number 3 represents the influence of information, measurement
and communication systems on information flow. To take another
production example, a "good" production schedule for a microcomputer
manufacturer might require information of the status of inventory in
all retail outlets. If there is no information system capable of
monitoring and reporting retail inventory, then the production
schedule must make do with factory information on the size of the
order backlog, the amount of finished inventory and the recent
shipping rate.
Filter 4 represents the influence of organizational and geographical
structure on information flow. As a decisionmaker, one's position in
an organization (both geographical location and position on the
organizational chart) has a profound influence on the information
sources one is exposed to.
Filter 5 represents the influence of tradition, culture, and
-289-
leadership on information flow. Filter 5 is intangible yet very
important in determining the factors that get the attention of
decisionmakers For example, suppose one is modelling the service
division of a yuter company and wants to understand the quality of
service provided to customers. Quality of service depends on the
speed with which servicemen fix customer problems. The division can
respond quickly if its servicemen receive information promptly from
customers. But the company also needs a "service culture". A
customer problem which is known to serviceman will get attention (i.e.
bring about some action) if the company's "culture" encourages good
service. Aculture for good service may derive from stories which
circulate the company. Such stories underpin the attitudes of
individuals in the service division, and condition the attention they
pay to customer problems (in other words, the weight they give to
information from customers requiring service).
What guidance do these filters provide the modeller? Principally they
help modellers to map the structure of organizational decisionmaking
by forcing them to pay close attention to the information sources that
are actually used by decisionmakers (as opposed to the information
sources that are available or that seem, at a distance, to be the most
"sensible") and to be aware that information deficiencies, bias and
error are commonplace. Also, the filters focus attention on the
modelling of decision processes, not just casual links or influences.
By being aware of the filters, modellers can ask more precise
questions to draw-out managers' knowledge, and to better specify
decision functions. The result is plausible feedback structure that
comes from linking well-specified decision functions.
EMPHASIS ON LEARNING AND DIALOGUE
Increasingly models are viewed as tools for learning-by-simulating,
where learning can involve the use of scenarios and many "what-ifs".
The challenge is to generate a useful dialogue between managers'
mental models of the system and simulation models which embody some of
the critical variables and interactions identified by a management.
team. Workshops and role playing simulation games have proved to be
useful in creating such dialogue.
An effective dialogue comes from a combination of obvious and
"surprise" simulations. The obvious simulations (usually partial
model simulations) build confidence in the model and clarify how it
works. Surprise simulations show unexpected or counterintuitive
dynamic behaviour, and often suggest new interpretations of facts
about the system. In order to use surprise simulations effectively,
model users need to establish in advance the results which they expect
from a model simulation (Mass 1981). Discrepancies can then be
recognised as such when they occur and examined closely to explain
whether they arise from errors in the computer model or errors in
people's mental models.
Partial model simulations are particularly effective for building
understanding of counterintuitive dynamic behaviour (Morecroft 1985,
Sterman 1985). The simulations are designed by cutting feedback loops
in the full model (or by building a deliberately simple, incomplete
-290-
model) in order to isolate a subset of the system's interacting
decision functions. The simplification is carried out in such a way
that simulations correspond to scenarios that managers can easily
identify with. Partial models are then combined and simulated in
logical stages to show how counterintuitive behaviour of the whole
model arises from the coupling of understandable pieces.
Partial model simulations expose the "intended rationality" of
decisionmaking in complex systems. They show that decisions and
actions of players in a system are "sensible" (intendedly rational)
when the feedback setting of the players' decisions is simple.
Dynamic behaviour which arises from "sensible" decisions and actions
is usually intuitively clear, and therefore conducive to dialogue.
USING WORKSHOPS AND ROLE-PLAYING SIMULATION GAMES
It used to be common in policy modelling to develop models containing
several hundred or even several thousand equations. These large
models accurately replicate historical time series and provide good
short-term predictions. Now, smaller models of thirty, forty or fifty
equations are commonly presented to management teams. The purpose of
these models is to prepare people for debate. Much less emphasis is
given to replicating time series.
In order to stimulate debate a model should be presented in a way
that dramatises assumptions and relates them to managers! experience.
The idea of "dramatising" a model has led to the development of
"policy workshops" and has brought renewed interest in role-playing
simulation games. In both cases the modeller (perhaps best thought of
as a facilitator/modeller) creates a "learning environment" for
managers that makes them feel part of the model situation. In
principle, participants come to relate their own experience more
closely to the model than they would in a conventional model
presentation. They also learn more readily the "lessons" about
dynamic behaviour that the model contains.
For example, Kreutzer (1985) has developed a workshop to explore the
dynamics of an arms-race. The workshop builds on a small, 20
equation, dynamic model (Forrester 1985). The model represents in
outline the decisionmaking processes used by two countries, X and Y,
for estimating their opponent's stock of arms, for judging the
adequacy of their own stock of arms and for procuring arms from
industrial military suppliers. The model also includes levels that
represent the stock of existing arms and new arms under development.
The decisionmaking network of the model captures in very interesting
ways the lags, distortions and biases that occur in the transmission
and processing of sensitive military and political information. The
dynamic properties of the model (exponential growth in the stock of
ams of both countries X and Y) arise from the imperfections assumed in
the system's decisionmaking processes.
The workshop immerses participants in the realities of military and
political decisionmaking. They are provided with articles on the
arms-race from magazines like Newsweek and the Economist. They are
presented with charts showing the history of the Soviet - US arms
race. They are given cartoon illustrations from magazines like Punch
-291-
or the New Yorker which portray (in amusing but memorable and usually
realistic ways) the imperfections of military intelligence (for
example, an illustration showing large crates being shipped to Cuba on
anonymous freighters, and two military officers debating the likely
contents of the still-closed crates). All this material activates
participants' mental models of the arms-race and highlights the role
of information processing and information feedback in arms control.
With this preparation, participants are able to relate their knowledge
and experience of arms races to the model and to appreciate the
assumptions that underlie the model's feedback structure and dynamic
behaviour.
The arms-race workshop also uses partial model simulations to show how
the decisionmaking processes that generate an arms-race are quite
“sensible and benign" when the imperfections and biases in information
processing are eliminated. For example, simulations which assume that
decisionmakers in countries X and Y have perfect knowledge of their
opponent's stock of ams (both installed and in development) exhibit
much slower exponential growth, or in some cases, no growth at all.
Role-playing games fulfil a very similar function to workshops by
providing a context of realism and drama to relate managers’ knowledge
to simulation models. In the case of games the drama is provided by
making participants play the role of selected decisionmakers in the
model system.
The production-distribution "hand simulation" game Sterman (1984) is a
good example of a game that promotes learning and policy debate. It
is a board game played by teams of four players. Each player takes a
role as either retailer, wholesaler, distributor or manufacturer in a
vertically integrated manufacturing and supply system (a beer
production and distribution system is usually selected). A player is
responsible for managing inventories and backlogs at one point in the
system (e.g. wholesaler) and for placing orders with the adjacent
player downstream (e.g. distributor) in the supply chain. The
objective of the players is to minimise the team's inventory and
backlog costs in the face of exogenous customer orders. The volume of
orders is not known in advance by any player, and is revealed week-by-
week to only the retailer. The game shows the difficulty of co-
ordinating decisionmaking in a system with imperfect information
processing. Almost all teams that play the game incur inventory and
backlog costs which are much greater than the "theoretical" cost
minimum.
The production-distribution game uses coins, paper and a plastic
printed board. But many new electronic or semi-electronic role-
playing games have been developed such as Meadows’ (1985) STRATEGEM 1,
Flint's (1986) multi-product salesforce game and Sterman and
Meadows'(1985) STRATEGEM 2. The management consulting company
Pugh-Roberts Associates (1986) has developed a role-playing game of
project management.
FUTURE RESEARCH ~- IMPROVING MODEL SUPPORTED "DIALOGUE" AND THE
MAPPING OF POLICYMAKERS' KNOWLEDGE
An important objective of future policy-related research in system
-292-
dynamics is to improve the quality of dialogue and debate between
managers and models. Better dialogue comes from capturing accurately
in maps and models managers' knowledge of the system, and from
strengthening the influence of model generated opinion in policy
debate. Many research paths are open to improve model-supported
dialogue. They include field experiments, behavioural decisionmaking,
game design and mapping technology.
FIELD EXPERIMENTS
Fields experiments are already underway to explore the process for
generating effective policy dialogue. The experiments are taking
place in both large and small business organizations in the United
States and Europe. Researchers and consultants are experimenting with
the content and sequence of model development to better understand
which modeling activities should be conducted during meetings and
which beforehand; to better understand what balance to strike between
qualitative mapping and simulation; and to better understand how to
use partial model simulations and simple scenarios to challenge
managers' intuition.
Researchers and consultants are also experimenting with the
composition of the project team (the mix of managers, modelers and
facilitators), the format of meetings (how frequent, how long, and
what mix of discussants), and the "technology" for presenting and
recording policy debate (flip-charts, blackboards, paper, overheads,
video projectors and computers (with word-processing, diagramming and
modeling software)).
Several recent papers describe the style and direction of the field
work. Richmond (1987) and Senge (1987) describe a "Strategic Forum"
which they view as a "process" to enable a cross-functional management
team to improve the match between operating policies and stated
strategic objectives. A forum involves several work steps for a
management team: articulating current vision and strategy, developing
simple "reality check" models, developing more complex models by
closing feedback loops, conducting "what-if" policy testing and
defining action steps. Morecroft (1984) describes "strategy support
models" which are intended to "provide executives with insight into
whether the policies and programs (of a business strategy) are
properly coordinated and whether they are in fact capable of achieving
the market and financial objectives called for by the strategy". He
describes two phases of modeling, a first qualitative mapping phase to
identify "players", policies, and feedback structure, and a second
simulation modeling phase to develop equations and concepts and to
debate the outcome of simple simulated scenarios.
It is interesting to note that research and consulting on the process
of model-building with management teams is already well-established
outside the system dynamics field. Well-known work has been carried
out by Phillips (1986) and Eden (1985) and the topic is receiving
increasing attention in the area of decision support systems (Land et
al (1987). Some cross-fertilisation of research and methods would
likely be fruitful.
=293-
BEHAVIOURAL DECISIONMAKING AND GAMING
The value of behavioural decision theory to system dynamics is clear
enough: it can help modelers to ask better questions of managers, to
specify decision processes more plausibly, and to capture more of
managers' knowledge in maps and algebra. An important extension to
this bridge~building is to embody the new ideas explicitly into
symbols for mapping (say by including information filters in maps) and
into protocols for questioning policymakers.
Another significant area for research is game design. Behavioural
decision theory gives some guidance to game design by focussing the
game-builders' attention on the design of the "decision shell" in
which human subjects will role-play. Immediately one thinks of
“designing a decision shell" then game-building takes on many
interesting research dimensions (that go well beyond the purely
technical issue of outfitting a simulation model with the capability
for occasional human intervention). There is the question of how one
"replicates" the organizational, cultural and administrative filters
(of information) that condition choice and action. What information
(from the vast matrix of simulated data available) should be presented
to game-players? How should screens of information be organized?
What balance of graphic, verbal and visual displays is appropriate?
How much leakage of information between players should be allowed in
multi-player games? What is an appropriate protocol for gaming-
decisions? How should one gauge the adequacy and fidelity of the
decision shell? The research questions are numerous. At a more
technical level one might consider the merits of different programming
environments and computers for developing behavioural decision shells.
Finally, there is a chal Lenging and potentially large research topic
in the use of gaming to link experimentally the behavioural
decisionmaking of individuals and groups to the dynamics of large
organizations. In this kind of research a simulation game becomes a
laboratory for "testing" cognitive limitations of individuals and
groups in environments that "simulate" large organizations. Subjects
make choices in an experimentally controlled setting (the decision
shell) that provides operating information. The operating information
is generated by a simulation model that "surrounds" the decision
shell. Subjects are free to make any choice they consider
appropriate, given the available operating information, their
knowledge of operating goals and incentives, their "mental model" of
how the rest of the organization operates, and also given their own
cognitive limitations. The actions and reactions of the rest of the
organization (comprising several behavioural decision functions,
actions and levels) are represented by algebraic functions and
simulated during the game. Since the situation is entirely
experimental, one can replace the decision shell and human
decisionmaker/s with an algebraic decision rule and discover (through
analysis or simulation) an "optimal" decision rule. Knowing an
optimal decision rule and the results of many game trials with many
different players, one can discover if and when people use
systematically poor decisionmaking heuristics. One can also model the
players' heuristics and compare them with the optimal decision rule in
order to probe the. link between cognitive limits and observed dynamic
behaviour.
~294-
Research along these lines is being carried out by Sterman (1986). It
is a fascinating area that promises to yield better understanding of
the reasons for (economically) inefficient dynamic behaviour in
business and social systems; experimental methods for validating model
assumptions; and new insights into the, design of role-playing
simulation games.
BETTER MAPPING TECHNOLOGY
There is a large potential for research which leads to better mapping
technology and therefore to a richer flow of policymakers' knowledge
into maps and models. The most direct research path leads straight to
improvements in software. A more ambitious research path leads into
aspects of modern computer science and artificial intelligence.
Software for mapping, modeling and simulation has improved over the
past five years, as outlined earlier. However, there is room for
still more improvement. Mapping (of the kind allowed by STELLA)
should permit word-and-picture maps to be built at the level of
policies (Morecroft 1982) rather than at the present level of
algebraic converters. Such high level maps would allow better
communication with policymakers (because the maps are readable,
visually compact and easily changeable) and would guide equation
formulation without constraining conversation (because they stand in a
natural hierarchy above equation formulation). The needed software
should combine the flexibility of drawing and writing packages (say
like MacDraw and MacWrite) with the modeling capability of STELLA.
New software should also help modelers write good clear algebra that a
policymaker can (almost literally) read! A simple step is to allow
much longer labels so that equations look like sentences. Also
needed, but more difficult to provide, is guidance for equation
formulation - a computer environment for developing equations that
weeds-out poor formulations. Here is an ambitious but clear research
challenge: to capture ina software package (at least some of) the
expert modelers’ rules of formulations (for example, dimensional
consistency checks and extreme-condition tests).
Finally, new software should give modelers more simulation power and
flexibility. Given a credible model, one should be able to probe
“policy parameter space" as quickly as one can envisage and articulate
meaningful policy scenarios. The required flexibility here is not
only for rapid re-simulation, but more important, for rapid
reformating and reorganization of simulated graphs and charts.
The most ambitious research path leads into modern computer science
and artificial intelligence. The challenge is to better understand
how to elicit and reconstruct managers’ broad operating knowledge into
meaningful word-and-picture maps, algebraic "sentences", models and
simulations. It seems to me that an important prerequisite is to
discover more precisely what we mean by the phrase "managers'
knowledge". Branches of Artificial Intelligence (AI) may provide some
ideas (see for example Minsky's (1986) Society of Mind). However,
there is a need for focus. The likely criterion for achieving focus
is to select the work that is most informative on how symbols (words,
-295-
charts, pictures, etc.) can be used to provide a "framework" on which
to hang managers' knowledge.
I have outlined some promising paths for future research in system
dynamics. A lot has been accomplished over the last ten years, but
the remaining opportunities and challenges are enormous. Future
research should provide the technology, theory and group processes for
policy microworlds which will (in Richmond's words) "help
organizations design their own future".
NOTES
1. This paper is an abridged version of "System Dynamics and
Microworlds for Policymakers", forthcoming in the European
Journal of Operational Research.
2. The term "microworld" comes from Seymour Papert (1980) ina
fascinating book called Mindstorms. Papert is a mathematician and
computer scientist at MIT who has devoted his energy to exploring
how computers can help people to learn. A fundamental premise of
his work is that people learn effectively when they have
transitional objects to "play with" in order to develop their
understanding of a particular subject or issue. The writer has.
pen, paper and word processor with which to hone his skill of
composition. The very young child has building blocks to learn
about sizes, sorting and simple construction. The combination of
transitional objects, learner and learning process is what Papert
calls a microworld or “incubator of knowledge". But what
transitional objects can one provide for learning about
"intangible" topics like motion, geometry, mathematics and (for
our purpose) policymaking? Papert suggests the computer and
simulations:
"The computer is the Proteus of machines. Its essence is
its universality, its power to simulate. Because it can
take on a thousand functions, it can appeal to a thousand
tastes."
The combination of computer, simulation language, learner and
learning process is a computer-based microworld.
REFERENCES,
Eden, C., 1985.Perish the Thought. Journal of the Operational
Research Society 36(9):809-819.
Flint, B.B., 1986, A Role-Playing Simulation for Sales Planning and
Control, Unpublished Masters' Thesis, Sloan School of Management, MIT,
Cambridge, Mass.
Forrester, J.W.,1961. Industrial Dynamics. Cambridge Mass.:
MIT Press.
-296-
Forrester ,J.W., 1985. Dynamic Modeling of the Arms Race. System
Dynamics Group Working Paper _D-3684-3, Sloan School of Management, MIT
Cambridge Mass.
Hall, R.I., 1984. The Natural Logic of Management Policy Making: Its
Implications for the Survival of an Organization. Management Science
30(8).
Kreutzer, D.P., 1985. A Microcomputer Workshop Exploring the Dynamics of
Arms Races. System Dynamics Group Working Paper D-3689-1, Sloan School
of Management, MIT, Cambridge Mass.
Land, F., Gall, M., Hawgood, J., Miller, G., Mundle, F., 1987
Knowledge Based Management Support Systems.
Proceedings of International Business Schools Computer User's Group
and Information Systems Association Joint European Meeting, London
Business School.
Mass, NwJ., 1981 Diagnosing Surprise Model Behaviour: a Tool for
Evolving Behavioral and Policy Insight. Proceedings of the
1981 System Dynamics Research Conference, Rensseraerville, NY 254-272.
Also available as MIT System Dynamics Group Working Paper D-3323,
Sloan School of Management, MIT, Cambridge Mass.
Meadows, D.L., 1985. STRATEGEM 1: A Resource Planning Game
Environmental Education Report and Newsletter 14(2):9-13.
Minsky, M., 1986. The Society of Mind, New York NY: Simon and
Schuster.
Morecroft, J.D.W., 1982. A Critical Review of Diagramming
Tools for Conceptualizing Feedback System Models. Dynamica 8(1):20-29.
Morecroft, J.D.W., 1984. Strategy Support Models. Strategic Management
Journal 5(3):215-229.
Morecroft, J.D.W., 1985. Rationality in the Analysis of Behavioral
Simulation Models. Management Science 31(7):900-916.
Papert, S., 1980. Mindstorms. New York, NY, Basic Books.
Phillips, L.D., 1986. Computing to Consensus. Datamation 68:2-6.
Pugh-Roberts Associates, 1986. Project Management Modeling Systems
PMMS, Pugh-Roberts Associates, Five Lee Street, Cambridge Mass 02139.
Richmond, B.M., Vescuso, P., Peterson, S., 1987. STELLA for Business.
13 Dartmouth College Highway, Lyme, New Hampshire 03768. High
Performance Systems Publications.
Richmond, B.M., 1987. The Strategic Forum: From Vision to Operating
Policies and Back Again. High Performance Systems Publications.
13 Dartmouth College Highway, Lyme, New Hampshire 03768,
-297-
Senge, P.M., 1987. Catalyzing Systems Thinking Within Organizations.
System Dynamics Group Working Paper D-3877-2 Sloan School of
Management, MIT, Cambridge Mass,
Simon, H.A., 1976. Administrative Behavior, third edition,
New York, NY: The Free Press.
Sterman, J.D., 1984. Instructions for Running the Beer Distribution
Game, System Dynamics Group Working Paper D-3679, Sloan School of
Management, MIT, Cambridge Mass.
Sterman, J.D., 1985. A Behavioral Model of the Economic Long Wave.
Journal of Economic Behavior and Organization 6(1):17-53.
Sterman, J.D., and Meadows, D.L., 1985. STRATEGEM-2: A Microcomputer
Simulation Game of the Kondratiev Cycle. Simulation and Games
16(2) 2174-202.
Sterman, J.D., 1986. Testing Behavioral Simulation Models by Direct
Experiment. System Dynamics Group Working Paper D-3783-1, Sloan
School of Management, MIT, Cambridge Mass, forthcoming in Management
Science.
