Qualitative v Quantitative Modelling: the Evolving Balance 1
1998 INTERNATIONAL SY STEM DY NAMICS CONFERENCE
"Qualitative v. Quantitative Modelling: The Evolving Balance" .
Eric Wolstenholme,
Professor of Business Learning, Leeds Business School and Director COGNITUS
Key words: System Dynamics, Modelling, Qualitative analysis, Quantitative Analysis.
Abstract
This paper addresses the issue of what are the wise uses of qualitative mapping and what are
the conditions that require formal quantitative modelling within System Dynamics.
The background to the evolution of qualitative and quantitative system dynamics will be
explored. This analysis will recognises that although the history of feedback thought
repeatedly contains the assertion that formal, quantitative models are essential for
understanding the dynamics of complex systems, the need for quantification is relative and
depends on the purpose of analysis, which in tum is related to the methods used and the
audience addressed.
The central theme of the paper will be to examine the strengths and weaknesses of qualitative
and quantitative system dynamics and to relate these to their respective tool sets. The paper
will also focus on evidence from the author's extensive recent use of qualitative and
quantitative system dynamics in education, training, research and consultancy studies of the
way in which qualitative and quantitative system dynamics can be linked together to
consolidate management learning, both in projects and in organisations.
The paper concludes that both qualitative and quantitative system dynamics are important and
related to the purpose of analysis. It is suggested that within studies the true power of system
dynamics to address problem solving lies in a judicious blend and intertwining of both
qualitative and quantitative ideas, aimed at addressing as broad an audience as possible whilst
remaining sufficiently rigorous to be useful. Within organisations it is suggested that there is
a need to cement together the use of qualitative system dynamics in management
development and quantitative system dynamics modelling for strategic and operational
learning in teams.
Introduction
System Dynamics as Quantitative Computer Simulation
In its original concept, the subject of system dynamics was purely a computer simulation
method. The core idea being to bring the emerging power of computer simulation to the
analysis of complex socioeconomic issues '’. Stock-flow diagrams were used to capture the
structure of systems and to keep a diagrammatic representation of the equations listing of the
resultant system dynamics simulation model. Essentially, model creation centred around a
preliminary diagram from which the initial simulation equations were written. However, the
equations soon dominated the model development process and the diagram, as in most
Qualitative v Quantitative Modelling: the Evolving Balance 2
systems analysis of the day, was often a retrospective attempt to monitor and communicate
the development process.
Enter Causal Loop Diagrams
Given that system dynamics models were basically control feedback models, an early
innovation was to introduce the idea of causal loop diagrams °, later referred to in the U.K. as
influence diagrams ‘°. Causal loop diagrams provided a high level means of conceptualising
models in terms of their feedback loop structure. These diagrams were then converted into
stock flow diagrams for the purpose of simulation modelling.
Qualitative System Dynamics based on Causal Loops
It was later suggested by a number of authors °”*°, that causal loop diagrams could be used in
a free standing mode without computer simulation to assist issue structuring and problem
solving.
One strand of this work resulted in the development of cognitive mapping "° and its associated
software. The other resulted in the development of qualitative system dynamics, firstly as a
prerequisite for quantitative system dynamics ", then as a free standing concept referred to as
systems thinking ” aimed at providing some level of insight into managerial issues by
inferring, rather than calculating, the behaviour of the system represented. Note that the use of
the term systems thinking in this way should not be confused with its use in the UK to
describe the systems approach to managerial problem solving represented by the various
strands of soft OR ®.
This assertion that causal loop diagrams alone could add value to issue structuring and
behaviour assessment was based on the fact that even in this mode such diagrams were
sufficiently rigorous to provide a significant increase in assistance to thinking compared with
other emergent diagrammatic tools. For example, the system problem solving approaches
embedded in the methods of soft operational research and, in particular, the rich pictures of
the soft systems methodology 14 .
Qualitative System Dynamics based on Stocks and Flows
At the same time the development of system dynamics simulation software tured towards
the use of visual interactive components *"*"’, Such software allowed the development of
models electronically by creating stock flow diagrams directly on the computer screen as
icons which could be opened to insert data to create simulation models, without recourse to
separate equation formulation. This method only allowed models to be changes via the
diagram and hence ensured that diagrams were always a one-to one representation of the
equation listing. These preliminary stock-flow diagrams were termed system maps to
distinguish them from system dynamics models and such software was designed to operate
either in mapping or modelling mode. Some software ” now allows the mapping to be carried
out in either stock flow of causal loop diagrams, although no way has yet been established to
directly convert a causal loop representation directly to a simulation model.
Qualitative v Quantitative Modelling: the Evolving Balance 3
System Archetypes
One of the cornerstones of qualitative system dynamics based on causal loop diagrams was
the use of generic structures. Of particular significance was the use of generic structures in
both diagramming formats. The stock-flow version of this was in the form of multiple stock
flows or main chain structures and the causal loop version of this was in the form of system
archetypes . These were patterns of loops related to patterns of behaviours (for example,
limits to growth and fixes that fail) which emerged from extensive cumulated simulation
modelling experience. These archetypal structures have since been further categorised '*. They
have also been developed to include model boundaries and condensed into truly generic two
loop combinations *°.
The current issue, addressed in this paper, centres on to what extent can qualitative mapping
and quantitative modelling contribute to understanding the dynamics of complex systems?
Spectra of Quantification
The Context of Quantification
Before answering this question it is of interest to take a systems perspective by standing back
and reflecting on the context of the issue. Two spectra are considered of relevance to this
context. The first is the spectrum of the audience addressed by these methods and the second
is the spectrum of the methods of problem solving available to shed light on complex issues.
Both of these spectra range from the extremes of purely qualitative to purely quantitative.
The Audience Spectrum
Firstly, the audience. Many people need insights into complex systems who are towards the
qualitative end of the audience spectrum. These include non-visual thinkers ”, non-
quantitative thinkers or those who would never have an interest in the learning overhead
associated with learning a modelling approach to problem solving. The non-quantitative
audience is thought to be much larger than the quantitative and to include many busy and
successful senior managers. There is strong evidence that people toward the quantitative end
of the spectrum are more introverted and, whilst more capable of using the methods, would
not have the same breadth of interest to became decision makers in influential positions.
Although this situation might suggest that the quantitative people (analysts) should help the
non-quantitative, it is well established that insights cannot be transferred easily *”. The
implication here is that methods must be found to enable everyone to be involved in the
modelling process and to lea these insights for themselves. In fact, one of the biggest
problems in growing the field of system dynamics is that of finding people with the right
blend of technical and managerial skills.
The Spectrum of Problem Solving Methodologies
Secondly, the methods of problem solving. These range from the extremes of total intuition
and speculation at the qualitative end of the spectrum to rigorous algorithms at the
quantitative end of the spectrum. Any way of assisting intuitive thinking involves some
degree of formal and logical issue structuring and both systems thinking and system dynamics
Qualitative v Quantitative Modelling: the Evolving Balance 4
can be thought of as tools for this purpose. An important point here is that system dynamics
simulation is itself not particularly formal and neither is qualitative system dynamics
particularly informal, relative to the extremes of this spectrum. Systems thinking and system
dynamics might be thought about as map-supported and model-supported speculation
respectively. No map or model is ever a complete analysis and there is always still a need for
further speculation beyond the insights reached.
The important conclusion of this section of the paper is that any claims for the value of
quantitative system dynamics simulation analysis over systems thinking is relative to where
they lie on the spectrum of problem solving methods and the needs of the spectrum of
audiences requiring assistance with their thinking. Further, that in applying any problem
solving methods there is a need to create a balance between the need to remain sufficient
quantitative to be applicable and sufficiently flexible to be relevant, in terms of both audience
and method.
The Strengths and Weaknesses of Qualitative System Dynamics
Systems thinking undoubtably provides a significant level of assistance to thinking.
‘Causal loop’ systems thinking enhances linear and laundry list * thinking by introducing
circular causality and provides a medium by which people can externalise mental models and
assumption and enrich these by sharing them. Further, it facilitates inference of modes of
behaviour by assisting mental simulation of maps. By identifying policy links in maps, it
allows focussed speculation of how to intervene to redesigned systems. By using archetypal
structures, particularly total generic two loop structures, involving policies, boundaries and
delays, it enables potential unintended consequences to be anticipated and hence increases the
chances of plans being achieved. The methods bring much needed tools to the strategic areas
of management and allow a wide range of managers to access the power of feedback thinking.
‘Stock-flow’ systems thinking requires additional skills and learning than its causal loop
counterpart, but adds the dimension of process representation into the analysis. It facilitates
similar analysis to causal loops but, by depicting processes as flows rather than influences, it
is not easy for the inexperienced to interpret feedback loop structures easily.
The dangers claimed for qualitative system dynamics is that it is easy to apply inappropriate
insights to problems and that, although the methods overcome the quantification problems of
simulation, they still require the strengths of feedback loops to be assessed (by at least
considering in the order of magnitude of variables) and still require at element of visual
thinking ability. One of the major criticisms is that the ability to become expert at causal loop
mapping requires people to have undertaken much quantitative modelling and that this issue
is never explained by its proponents.
The Strengths and Weaknesses of Quantitative System Dynamics
Quantitative system dynamics on the other hand adds the dimension of data to mapping
structures and hence allows computer simulation of systems to ascertain their behaviours over
time. Inferring behaviour of even the simplest of feedback structures is very difficult even for
the trained since the human ability to consider any more than a very small number of
variables at once is limited. Quantitative system dynamics brings together the best
Qualitative v Quantitative Modelling: the Evolving Balance 5
combination of both people and their creative thinking ability and computers and their data
manipulation ability.
Computer simulation modelling adds significant value to qualitative mapping by enabling
deeper and more rigorous analysis. It allows a combination of soft and hard elements which is
unique and powerful.
The problem with computer simulation is that data is often not available and hence models
are still speculative. Further, they are often idealised representations of the world constrained
by the restrictive nature of stock-flow diagrams. One of the major problems is that of how to
produce simple, balanced and elegant models at an appropriate level of aggregation in time
and space to be useful. There is always a tendency to produce models which are too detailed
and complex and to insufficiently validate them against the mental models of their creators.
Much skill is required in achieving good modelling practice both in terms of assembling the
appropriate teams and the modelling process. The ideal is to create true ‘modelling for
learning’ * where the process on taking part is as important an outcome as any predictive
answers. A gain the blend of skills for expert modelling competence is difficult to find.
Mixing Qualitative and Quantitative System Dynamics within projects
Since both qualitative and quantitative analysis are important to the success of any system
dynamics study it is important to sequence the process of application of the method to capture
both approaches. On such sequence being extensively and successfully used in practice by the
author is referred to as ‘intertwined project learning’ and is shown in Figure 1. In practice this
process is iterative at each stage, as implied by the circular arrows. Further each stage may be
self contained and the process can be terminated at any point if required.
The first step is issue definition and exploration and the second, knowledge capture. Causal
loop diagrams are used in both of these stages for the prime purpose of management
engagement and knowledge extraction. These stages often involve the use of archetypal
structures to stimulate the overall modelling process. The third stage is to develop the output
from stages 1 and 2 into ‘first pass’ qualitative maps related to a consensus view of the issues
of managerial concern and to feed back individual maps to managers in a group forum. These
maps are then used to define the scope of a study in terms of its degree of resolution in time
and space and its boundaries. They are also used as a basis for group speculation and
hypotheses development of likely system behaviour, alternative policy generation and
redesign of organisational boundaries and responsibilities.
The forth stage is to begin the process of serious data collection and quantification and the
development of simulation models to substantiate the hypotheses created and test out
organisational redesign ideas and plans.
At this point those people involved in the modelling process have usually developed
numerous insights from the modelling experience and are keen to in implement their findings.
However, the most serious factor effecting the success of implementation centres on
conveying the insights to colleagues and, as previously intimated, it is very difficult to
Qualitative v Quantitative Modelling: the Evolving Balance 6
transfer a learning to others. Stages five, six and seven of the modelling process therefore
focus on the important area of disseminating the results and insights.
Stage five is involves returning to a qualitative mode of operation and is aimed at
summarising the model in terms of an easy to understand archetypal structure. System
dynamics models can easily become large and complex and the important causal loop
structure generating the essential model behaviour can be lost in the detail. It is important
therefore to maintain a causal loop map which summarises the structure of the model at any
time. An excellent way of doing this and at the same time communicating the essence of the
model is to collapse the model down to an archetypal representation.
Stage six is not aimed at summarising insights, but at creating a way of allowing other people
to relive the experiential learning process that was undertaken by the modelling team. This is
achieved by developing the model into an easy to use microworld by which people can use
for constructive play, either individually of in a workshop setting, to investigate
organisational redesign. Stage seven provides an enhancement of stage six by embedding the
microworld into multimedia learning environments (LEs), which combines the model(s) with
cognitive help routines which can be revealed at a controlled rate to facilitate the learning
process.
Combining Qualitative and Quantitative System Dynamics within Organisations
The number of organisations now using qualitative system dynamics (in its guise as systems
thinking) and quantitative system dynamics is growing rapidly. However, the tendency is still
for these two (absolutely interdependent and complementary) approaches to be used in
isolation.
Systems thinking is finding a home in management development programmes, where it is
bringing clarity and rigour to a particularly abstract field. At the same time, system dynamics
is being used to provide senior and middle managers with models for strategy and operational
policy development at many different levels of an organisation. A number of large companies
have over twenty different system dynamics modelling projects in existence at one time
The danger is that neither systems thinking or system dynamics, if used in isolation, will
achieve their full potential. The systems thinking contribution to learning might still be too
abstract and unrelated to the specific operational realities of organisations to make a lasting
and meaningful co contribution to management development. And the specific leaning and
insights created from system dynamics modelling may be lost, without a formal process by
which to capture and relate those insights back into general and top management
development.
A method for consolidating leaming from system thinking and system dynamics in large
companies is to formalise a process which builds upon the interdependence of Systems
Thinking for general management learning to a wide spectrum of managers and System
Dynamics modelling for strategic and operational learning in small project teams. One such
process develop by the author is referred to as Accelerated Business Learning (ABL) and is
described in Figure 2.
Qualitative v Quantitative Modelling: the Evolving Balance us
The purpose of this process is to embed qualitative and quantitative system dynamics into the
learning culture of the organisation.
ABL involves capturing the output and insights from system dynamics models at any level
and sector of an organisation through the creation of reflection workshops. Here team leaders
and members from different modelling projects can share model process and content
experiences.
Further, these experiences can be captured and used as the basis for qualitative system
dynamics inputs into management development programmes. In particular, the insights can be
developed as system archetypes and the models into learning environments. The fact that
these archetypes and LEs are constructed from real in-company situations rather than those
imported from external situations provides a much more powerful impact on senior managers.
The introduction and communication of in-company modelling projects and their insights
within general company training provides, in turn, a platform for managers to sponsor more
detailed training in system dynamics modelling for their staff, and hence new modelling
initiatives.
The outcome is a reinforcing process of learning in system dynamics.
Conclusions
This paper has sought to shed light on to how to best combine and balance the use of
qualitative and quantitative system dynamics to bring the methods to a wider management
audience, whilst focussing on enhancing the rigour of management thinking. It is concluded
that both approaches have important contributions to make to management thinking and that
there is a need for methods of applying them which judiciously blend and intertwine their best
characteristics. A ‘within project process’ and a ‘within organisation process’ to achieve this
have been introduced.
References
1. Forrester. J.W. (1994)(1) The Collected Papers,, Productivity Press,
Cambridge, Mass.
2. Forrester. J.W. (1994)(2) Industrial Dynamics, Productivity Press, Cambridge,
Mass.
3. Randers, J. (1980), Guidelines for Model Conceptualisation, in Randers, J.
(ed), Elements of the System Dynamics Method, Productivity Press,
Cambridge, Mass., pp 117-39.
4. Coyle R.G. (1990) Management System Dynamics, Wiley, Chichester.
5. Coyle R.G. (1990) System Dynamics Modelling, Chapman and Hall, London.
6. Wolstenholme E. F., (1982) System dynamics in perspective, JORS, 33,
547-556.
Qualitative v Quantitative Modelling: the Evolving Balance 8
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Wolstenholme, E. F. and Coyle, R.G., (1983), The Development of System
Dynamics as a Methodology for System Description and Qualitative Analysis,
Journal of the Operational Research Society, 34(7): (569-581).
Wolstenholme E. F., (1985) A methodology for qualitative system
dynamics, Proc. International System Dynamics Society Conference,
Denver, U.S.A..
C. Eden, S. Jones and D. Sims (1979) Thinking in Organisations,
Macmillan,London.
C. Eden, (1989) Using Cognitive Mapping for Strategic Options
Development and Analysis (SODA), in Rational Analysis for a
Problematic World (ed) J. Rosenhead.
Wolstenholme E. F. (1990) System Enquiry - a System Dynamics
Approach, Wiley, Chichester.
Senge P. (1990) The Fifth Discipline, Doubleday
Richardson, G.P., Wolstenholme, E.F. and Morecroft, J.D.W. (1994) (Eds),
Systems thinkers, Systems Thinking, The System Dynamics Review, Vol. 10,
Nos. 2-3.
P. B. Checkland (1981) Systems Thinking, Systems Practice, Wiley,
Chichester.
Davidson, P.I, (1992) A System Dynamics Approach to Computer Based
Management Learning Environments: Implementations in Powersim, In
Modelling for Learning (1992), J. W. D.Morecroft and]. D. Sterman Eds.),
Productivity Press, 1994
Richmond, B., (1997),‘ithink Version 5 User Manual’, High Performance
Systems, Hanover New Hampshire, USA, 1997.
R. L. Eberlein and D. W. Peterson (1992) Understanding models with
Vensim. In Modelling for Learning (1992), (J. W. D.Morecroft andJ. D.
Sterman Eds.), Productivity Press, 1994
Lane, D. C. and Smart, C. (1996), Reinterpreting ‘generic structure’: evolution,
application and limitations of a concept, The System Dynamics Review, Vol.
12 No. 12, pp 87-120.
Wolstenholme E, F, and Corben, D. (1985), Towards a Core Set of
Archetypal Structures in System Dynamics,Proc. International System
Dynamics Society Conference, Cancun, Mexico.
Qualitative v Quantitative Modelling: the Evolving Balance 9
20. O'Connor. J and Seymour, J, (1994), Training with Neuro Linguistic
Programming, Harper Collins, London.
21. DeGeus A. P. (1988) Planning as Leaming, Harvard Business Review,
Volume 66, Number 2, March-April, 1988.
22. De Geus A. P. (1997)The Living Company, Nicholas Brearly Publishing,
London.
23. Lane, D. C, Modelling as Learning: A Consultancy Methodology for Enhancing
Learning in Management Teams. in Morecroft J. W .D. and Sterman J.D. (Eds)
(1994) Modelling for Learning, Organisations, Productivity Press, Portland,
U.S.A..
Figure 1, Blending Qualitative & Quantitative System
Dynamics within projects
A Hierarchy of Tools and Methods
[NL NO NL NON LDN
Issue Knowledge Qualitative Quantitative System Flight Learning
Definition Capture Models Models Archetypes Simulators Environments
NY NY N YR NY’ RLY
<< — = Engaging & Relating =——_ eee Se SE = = = —pP
@ —— — — Learning errr oP
<< — — —Disseminating -—-—-—- >
Figure 2, Blending Qualitative & Quantitative System
Dynamics within projects
The COGNITUS ABL Process ‘
Senior
ABL Leadership «———_____ Management
Development a ABL Briefings
Workshops = >.
Organisation Knowledge Base
I
I :
New Team <——— | Model Libraries Team
Training Flight Simulators & Learning Environments f ‘ Reflection
Workshops = , Workshops
System Dynamics Modelling Projects
CC BY-NC-SA 4.0