1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Integrating Archetypes and Generic Models into a Fr k for Model
Conceptualisation
David A Corben
Department of Management Science
The School of Management
University of Stirling
Stirling
FK9 4LA, UK
Abstract
Model conceptualisation is the most difficult system dynamics skill to acquire, practise and teach.
The advent of user friendly simulation tools; STELLA and ithink, have made the task of model
construction and use much easier. Model conceptualisation by contrast remains as difficult as ever.
Experiences in training managers in system dynamics show that it is a relatively simple task to turn
ip! i into and d i The bility to recreate a model from
an ithink map and a written description of operating policies is quickly acquired, as is the ability of
modifying that model to test out possible solution to problematic behaviour.
The problem arises when these p i imul, are asked to lise a simple two loop
model from a problem description, presented in the form of a newspaper article. This task is found
very difficult by most course members.
This paper. describes the development of a new technique to assist with the model conceptualisation
process. The method i types and their ponding generic Models into a framework
that helps modellers move from a problem description to a first pass ithink model and/or causal map.
An application of the method, to aid the conceptualisation of a model of Government funding of
housing associations is described.
System Dynamics : Methodological and Technical Issues, page 1
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Integrating Archetypes and Generic Models into a Framework for Model
Conceptualisation
Introduction
During last two years, the author has been involved in the delivery of a number of three day training
courses in systems thinking. These courses aim to develop both qualitative and quantitative modelling
skills in parallel, they are “hands-on” in nature; the emphasis is on learning by doing.
The majority of these courses were conducted for small groups of managers or mixed groups of
managers and analysts. Some of the courses were run as public workshops, which were open to
i from different isati but the majority were “ ‘in house” courses run for a specific
The back 21 d of the particip and the ion from which they came was
highly varied. This has given the author a chance to observe a representative cross-section of both
managers and organisations that have shown an interest in adopting system thinking in the UK.
Training Course Structure
This brief overview concentrates on the model building elements of the training course. These
are designed to form a progression in which parti are required to increase their own
contribution to the modelling process. In the first exercise participants are given explicit instructions,
the final exercises involve freeform modelling.
Exercise |
In their first exposure to the ‘think software managers are talked through a “hands on” exercise that
builds a simple two stock model of staff turnover. (Wolstenholme 1990). This task introduces the
main features of the ithink software and highlights the nature of the modelling process.
Exercise 2
This exercise requires the construction of a model from a given ithink map and equation listing. The
model is based on a real case study (Wolstenholme 1993) and particip: are taken through the
background to the model before they start to build the model. The aim of this exercise is to reinforce
the managers’ skills with the software and provides the opportunity to perform “what if? simulations
with a larger and more realistic model.
Exercise 3
The third modelling exercise is a major case study, based on the market growth model (Forrester
1968), that takes up most of the second day of the training course. The case study consists of two
parts. The first is a qualitative mapping exercise, participants work in small groups to create a causal
map from a problem description and they are also required to suggest a possible solution to the
problem. A facilitated debriefing session follows in which the problem and proposed solutions are
discussed.
The second part is a quantitative modelling exercise. The ithink map is provided as are the parameter
values, but participants are required to create their own equations, but a written description of
operation policies is provided to assist with this task. When the model has been built, the managers
are required to modify it to incorporate the solution that was identified in the qualitative exercise, but
no additional help or information is given.
Exercise 4
The fi inal day of the course is devoted to Freeform modelling of participants own problems, but first
some | practise in P; are given a series of modelling
exercise, that require participants to produce causal maps ‘and/or ithink maps from written
descriptions. Some of these are reasonably transparent descriptions of simple models, but others are
based on problem descriptions taken from newspapers are much more open-ended in nature.
of Partici Perfc
It has not been possible to carry out a quantitative evaluation of performance, but the author has
observed a large number of managers over the course of the last two years and the following insights
have been gained.
System Dynamics : Methodological and Technical Issues, page 12
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Use of Software
Teaching managers how to use the ithink software is relatively easy. Most managers can be taught
how to “fly” the software in a couple of hours.
Model Conceptualisation
This is the most difficult modelling skill for managers to acquire. Participants often stated that they
found conceptualising the hardest part of the workshop and this was confirmed by their actual
performance.
P i find isation easier if it is carried out within some kind of framework. A good
example of this, can be found in the second day case study (exercise 3). The task of incorporating the
qualitative solution into the base case quantitative model is usually carried out very successfully, with
little need for facilitator intervention. In contrast the less P of the
exercises and the freeform modelling, are found to be much more difficult (exercise 4).
Participants confronted with a blank sheet of paper, find it difficult to start the modelling process
going. If however they can be facilitated to build a working first pass model (this can be a very simple
model), then they usually have the confidence to incrementally develop this model, with only
occasional need for facilitator assistance.
Use of Models
Experimenting with a model comes naturally to most managers, but some need some prompting to
experiment ly and interpret simulation output.
A number of participants were ivated to ive and impl their own ions to the
model used in exrecise 3. It was noted with some interest that it was the managers who were most
likely to want to extend the model, analysts tended to regard the case study simply as an exercise to
be carried out and soon lost interest in the completed model. However the analysts were in generally
better at implementing such extensions.
This suggests that it is best if workshops are attended by mixed groups of managers and analysts. This
has been confirmed by experience, the workshops where managers and analysts worked together,
were all very successful and useful models of the own probl were developed on the
final day.
Identification of Areas of Difficulty
The performance of participants clearly shows that conceptualisation is the most difficult system
thinking skill to master. In particular it is the early stages of the conceptualisation process that causes
problems. Once participants have a simple model to work with, they are capable of conceptualising
and implementing enhancements to that model. The rest of this paper describes the development of a
modelling framework that aims to help managers move easily to a working model as early as possible
in the conceptualisation process.
Model Conceptualisation
Model conceptualisation is the most difficult stage in the modelling process, but is also the most
rewarding, for it is here that the big leap in understanding occurs. Given the complexity of a typical
managerial system it is possible to build any number of models of that system. The purpose of the
conceptualisation process is to identify the model that will provide a suitable vehicle for
understanding the behaviour that is of interest to the managers of that system.
In ising a model, the following need to be decided; the basic feedback structure, the level
of aggregation, the model boundaries and the time frame. It follows that the conceptualisation process
is essentially qualitative in nature. Turning the qualitative conceptual model into a quantitative model
is not a trivial task, much about the system will be learned along the way, but these will be
1 gains in und ding, that serve to enrich and the add detail to original insight, that
occurred at the moment of conceptualisation.
System Dynamics : Methodological and Technical Issues, page 13
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Conceptualisation
(The big leap in understanding)
Model
Development
(incremental
gains in
understanding)
Figure 1 The modelling process
Current Modelling Frameworks
A number of techniques for the building of system dynamics models have been proposed.
(Forrester 1961; Randers 1980; Richardson & Pugh 1981; Morecroft 1982;
Wolstenholme & Coyle 1983; Wolstenholme 1985, 1990, 1992; Mass 1986; Richmond et al. 1987,
1993)
These approaches typically break the modelling process down into a series of steps (one of which is
1) and provide and suggest techniques for each step. These methods
undoubtedly contain much useful advice, but the pp to model does
not provide enough support for novice modellers.
Some methods make use of modular building blocks (Wolstenholme & Coyle 1983; Richmond et
al. 1987, 1993). This provides the modeller with “something to work with”, but turning these building
blocks into a working model is still a difficult task for the inexperienced modeller.
The novice modeller needs to be provided with the facility to conceptualise a model in terms of
generic structures thereby providing a working model for experimentation, very early on in the
conceptualisation process. In the next section generic structures and their use in the modelling process
are reviewed.
Archetypes, Generic Models and Generic Structures
There is at present, nothing close to an accepted definition of a generic structure: it is unlikely that two
system dynamicists discussing the concept will be talking about the same thing.
Mark Paich (1985)
This statement is, unfortunately, as true now as when it was first written nearly ten years ago. In an
effort to clarify the situation, three different classes of generic structure can be identified.
(Paich 1985) (see Table 1) The names have been applied by the author. The use of these structures in
the conceptualisation process will now be discussed.
Table 1 Classification of Generic Structures
Generic Models Structures generic to a specific problem domain.
Pp that are between different
problem domains,
Building Blocks Sub-structures that are found as building blocks in
many different models.
System Dynamics : Methodological and Technical Issues, page 14
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Generic Models
The use of an appropriate generic model can greatly simplify or even eliminate the conceptualisation
process. The problems with this approach are; the need for a suitable generic model of the problem
domain under investigation to be available, there is a danger that an inexperienced modeller will
select an inappropriate model and finally the existence of a generic model may give the modeller a
preconceived view of the problem.
Archetypes
Archetypes can be used a way of ising more general probl but for this to work, the
modeller must have a good understanding of the full set of the archetypes and the totally qualitati
nature of the archetypes may cause difficulties. The archetype approach to isation, shares
with the generic model method, the probl of; availability, inappropriate choice of
structure and preconceived view.
Archetypes were originally seen as fundamental building blocks (Senge 1985), but now they are seen
as a set of more complex structures. (Senge 1990) There is still a degree disagreement over the
constituents of this set. (Senge 1990; Kim 1992)
Table 2 The base archetypes
Intended Action
Control Growth
B/R RIB
Opposition Fixes Limits
that to
System Fail Success
Reaction BIB RR
Competition Fighting Success
for to the
Control Successful
It has been shown that there is a reduced set of four “Base Archetypes”, consisting of two loops, that
can be used to represent managerial problems in terms of an intended action and a system reaction.
(Wolstenholme & Corben 1993) (see Table 2) This simplified way of looking at a system provides an
ideal vehicle for model conceptualisation.
Building Blocks
These structures are more usetul in the iterative development stage of the modelling, when processes
are being modelled in more detail. than in the conceptualisation stage.
A Framework for Model Conceptualisation
The method uses the base archety pes and yeneric models of those archetypes to ease the transition to
a simple working model. The identification of an intended managerial action and a system reaction
allows a base archetype and it’s corresponding generic model to be selected. The modeller therefore
only needs to identify two loops before they can begin experimenting with a working model, that can
be used as an aid to clarify their thinking by exploring the chosen structure’s dynamic behaviour.
The selected generic model is next customised to produce a first pass model that can then be
iteratively developed. The steps of the method are described in Table 3 and an overview of the
process is given in Figure 2. An example application of the method is described in the next section.
System Dynamics : Methodological and Technical Issues. page 15
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Table 3 The steps of the conceptualisation method
1 Specify the Intended Behaviour
Is the aim growth or control. Draw the intended behaviour loop.
2 Identify the System Reaction
Will the system respond with growth or control. Draw the system reaction loop.
3 Create a Base Archetype
Link the loops identified in 1 & 2 to create a base archetype.
4 Specify the Problem as a Generic Model
Take the generic ithink model corresponding to the base archetype and by
renaming the model elements, customise it to represent the problem. It can now
be verified that the chosen generic structure is capable of producing appropriate
behaviour.
5a Qualitative First Pass Model
Flesh out the loops, by adding intermediate variables and the organisational
boundaries.
5b = Quantitative First Pass Model
Add extra detail to the ‘think model to keep it consistent with the qualitative
model. Incorporate “order of magnitude” data into the model and experiment
with the model.
6 Iterative Model Development
From here onwards the model(s) can be developed in an iterative fashion.
Issues in Designing a F k for C
In designing this methodology, careful though has been given to striking an appropriate balance
between, providing sufficient structure to help a manager build a model and the danger of enforcing
an over prescriptive set of rules that will constrain a manager’s thinking. There is a real danger of this,
in the specific case of model development and in the more general development of an individual’s
system thinking skills.
In the case of model development, the availability of generic structures may encourage the forcing of
the problem to fit the available model. This is potentially a major problem. Operational Research, for
example, has been severely criticised for this very failing. (Ackoff 1979)
The structures used in the conceptualisation framework were chosen to be sufficiently generic to
avoid this problem. The selection of a base archetype is merely a starting point for investigation, not a
final definitive statement about the problem. The initial identification of a loop pair, can lead to a
model than is equivalent to one of the system archetypes or indeed no specific archetype at all. The
fact that some of the base archetype can be incrementally developed into more than one of the more
complex system pes shows the of this
The framework could have been built around the full set of system archetypes, but it was felt that this
would have been too restrictive. In using the system archetypes for conceptualisation, there is a real
danger that the selection of an archetype will be both a starting point and an ending point; the
modeller will miss out on the challenging but rewarding task of structuring their own view of the
system.
System Dynamics : Methodological and Technical Issues, page 16
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Figure 2 Overview of the conceptualisation method
In the wider context of an individual’s modelling skills, system thinking is more than just a simulation
technique, it is a distinctly different way of looking at the world. This ability can only be acquired by
thinking hard about real problems and therefore there can be no formal proscriptive method for
practising systems thinking. This fact has been recognised from the earliest days of system dynamics
and has been restated many times over the intervening years. (Forrester 1961; Saeed 1986;
Senge 1990; Richmond 1990).
The purpose of the lisati ‘k is th seen by the author as a tool to help the
novice system thinker through the difficulties of their early model conceptualisations, so that through
this experience they may develop their own system thinking skills. Experienced systems thinkers may
not need to use the framework to aid to their own thinking, but it can be used to good effect as a
facilitation technique, to aid the thinking of others, when model building with a group of managers for
example.
An Example: Ci
ptualising the Housing A ‘iation Model
Problem Description
The government is seeking to control its spending by imposing cash limits on the individual
government departments. In this particular case we are focusing on the Department of the
Environment and the grant aid that it gives to support social housing. Low cost housing is provided by
independent non-profit making bodies called housing associations.
The amount of grant aid given to Housing iations by the Dep of the Envi has
fallen from 90% five years ago to around 55% today. This reduction in grant aid has required the
housing associations to borrow more money in the open market the extra cost of which has caused
rent levels to rise. The majority of the people housed by housing associations have low incomes and
many cannot afford the higher rent levels. Therefore they claim housing benefit from the Department
of Social Security to help pay their rent. (Philips 1993; Page 1993)
‘System Dynamics : Methodological and Technical Issues, page 17
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Step 1: Intended Behaviour Step 2: System Reaction
Cuts in
Spending
s | byDept. of
Environment
Rising government spending triggers cuts Reduced spending by one government
in departmental spending, this causes a department can and in this case will
reduction in total government spending. create extra costs for other departments.
Step 3: Create a Base Archetype
Ss
Cuts in
Spending
s by Dept. of
Environment
Combining the intended behaviour loop with system reaction loop creates a base
archetype; in this case it is the “Fixes that Fail” archetype.
System Dynamics : Methodological and Technical Issues, page 18
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Step 4: Specify the Problem as a Generic Model
Probie SyeapEOR
change b syaptom
nescebn Te ¢
surcoser C) O onasqvece
own site of Fe
WA perpct Oumton
pecamiee FABRE a pymentatin Tine
(7) ss spenting
act on Das Spend
The generic model of the Fixes that Fail archetype is customised, by changing
the names of the model elements, so that it represents the specific problem.
Step 5a: Qualitative First Pass Model
Department of Social Security
Departmental 5
‘Spending
Budget
SA Grant aid
S$ To Housing
Associations
Recuremer Housing
Association
Department of Environment
Extra model elements have been added to the base archetype, delays and
organisational boundaries have been identified.
‘System Dynamics : Methodological and Technical Issues,
page 19
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Step 5b: Quantitative First Pass Model
Govema ent spending
ihe fh Spending
cut Grant
The direction of the “Fix” flow has been reversed (the fix is a reduction in grant
fraction). An outflow has been added to the government spending stock to
model annual budgeting. Two graphical functions have been created to quantify
departmental spending. This model is capable of producing a convincing
reference mode of behaviour.
Conclusions
Observation of a number of systems thinking training courses, has shown that it is a relatively
simple task to teach managers how to use the STELLA/iThink software. Model conceptualisation
has been identified as most difficult system thinking skill to acquire. In addition the benefit of
training mixed groups of analysts and managers has been highlighted.
A conceptualisation framework, that quickly takes managers from a problem to a working model
has been developed. The framework has been applied to the conceptualisation of a model of
public housing provision.
Currently the framework supports the “Fixes that Fail” and the “Limits to Success” base
archetypes. This limits the lication of the fr k to based
around these two base arch Further work will extend the method to cover
the rest of the base archetypes. This will allow the enhanced framework to be applied as a
conceptualisation aid to any kind of problem.
Acknowledgements
The work described in this paper has been supported by the ESRC. and Cognitus System Ltd.
References
Ackoff, R.L. 1979. The Future of Operational Research is Past, Journal of the Operational
Research Society, 30: 93-104.
Forrester, J. W. (1961). Industrial Dynamics. Cambridge MA: Productivity Press.
Forrester, J. W. (1968). Market Growth as Influenced by Capital Investment. Industrial
Management Review. 9(2), 83-105.
Kim, D. H. (1992). Systems Archetypes at a Glance. The Systems Thinker, 3(4), 5-6.
Mass, N. J. (1986). Methods of conceptualization (1981). System Dynamics Review, 2), 76-80.
Morecrofi, J. D. W. (1982). A Critical Review of Di: ‘ing Tools for Ci
Feedback System Models. Dynamica, 8, part 1, 20-29.
Page, D. (1993). Building for Communities: A Study of New Housing Association Estates, Joseph
Rowntree Foundation, York.
System Dynamics : Methodological and Technical Issues, page 20
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Paich, M. (1985). Generic Structures. System Dynamics Review, 1(1), 126-132.
Phillips, M. (1993). Quick Fix For The Homeless That’s Sure To Come Unstuck. The Guardian
5/8/93, pp 18.
Randers, J. (1980) Guidelines for Model Conceptualisation. In Elements of the Systems Dynamics
Method, ed. J. Randers. Cambridge, Mass.: MIT Press.
Richardson, G. & A.L. Pugh III (1981). Introduction to System Dy ics Modelling with
Dynamo. Cambridge MA: Productivity Press.
Richmond, B., P. Vescuso, & S. Peterson (1987). An Academic User's Guide to Stella. Lyme,
NH: High Performance Systems.
Richmond, B. (1990). Systems Thinking: A Critical Set of Critical Thinking Skills for the 90's
and Beyond. In D. F. Anderson, G. P. Richardson, & J. D. Sterman (Ed.), Proceedings of the
1990 International System Dynamics Conference (pp. 934). Chestnut Hill, Mass.
Richmond, B., S. Peterson & C Charyk (1993). Introduction to Systems Thinking and Ithink.
Lyme, NH: High Performance Systems. .
Saeed, K. (1986). Minds over methods. System Dynamics Review, 2(2), 150-156.
Senge, P. M. (1985). System Dynamics, Mental Models, the Development of Management
Intuition. In Proceedings of the 1985 International Conference of the System Dynamics
Society (pp. 788). Keystone, Colorado.
Senge, P. M. (1990). The Fifth Discipline: The Art and Practice of the Learning Organization.
New York: Doubleday.
“Wolstenholme, E. F., & R. G. Coyle (1983). The Devel of System Dy ics as a
Methodology for System Description and Qualitative Analysis. Journal of the Operational
Research Society (UK), 34(7), 569-581.
Wolstenholme, E. F. (1985). A Methodology for Qualitative System Dy ics. In Pr dii
of the 1985 International Conference of the Systems Dynamics Society (pp. 1049). Keystone,
Colorado.
Wolstenholme, E.F. (1990). System Enquiry - A System Dynamics Approach, Wiley, New York.
Wolstenholme, E.F. (1992) The Definition and Application of a Stepwise Approach to Model
Conceptualisation and Analysis. European Journal of Operational Research. 59(1992) 123-
136 North-Holland.
Wolstenholme, E.F. (1993). A Case Study in Community Care using Systems Thinking, Journal
of the Operational Research Society. Vol 44, No 9, pp. 925-934.
Wolstenholme, E.F. & D.A. Corben (1993). Towards a Core Set of Archetypal Structures. In
Proceedings of the 1993 International Conference of the System Dynamics Society (pp. 583-
592). Cancun, Mexico.
System Dynamics : Methodological and Technical Issues, page 21
