Reductionism, Holism and System Dynamics
Martin Rafferty
Department of Accounting and Finance
London South Bank University
Borough Road, LONDON SE1 OAA, UK
Abstract: The objective of this paper is to examine the concepts of holism and
reductionism as they relate to System Dynamics (SD) and to a lesser extent Systems
Thinking (ST) then to relate the findings of that examination to some of the
disillusionment with SD and the resultant lack of commercial take up. This paper
looks at the concepts of Reductionism and Holism as applied throughout the SD
literature in a broad range of academic areas from medicine and philosophy to physics
and looks at different definitions of holism, and reductionism. Practical applications
of Reductionist and Holistic principles are examined. From this analysis the
fundamental nature of SD as a Holistic/Reductionist methodology is deduced and
differentiated from the more generally holistic nature of ST. From this basis the
conclusion that SD is a reductionist methodology in practice, though not necessarily
described as such by practitioners, is postulated. The paper recommends that there be
further work carried out in the field of theoretical and practical holism and greater
awareness of these issues within the practitioner community. Within the conclusions
there is a pointer for some further fundamental areas of work.
Keywords: “System Dynamics”, “Systems Thinking”, Holism, Holistic,
Reductionism, Reductionist, Methodology.
Research context
This paper examines the lack of debate around the concept that holistic thinking, as
opposed to reductionist thinking, is inherent in System Dynamics (SD).
System Dynamics (SD) as its name suggests is a philosophy and practical
methodology for modelling and simulating the dynamic behaviour of systems; social,
financial, ecological and otherwise over time. It is closely related to Systems Thinking
(ST) but is often seen either as a subset of that philosophy or as an overlapping
complement to it. ST itself is a philosophy that has emerged from Feedback theory
and General Systems theory.
It is widely noted that System Dynamics adopts the ‘big picture’ viewpoint;
attempting to capture the nce of systemic behaviour rather than that of the
individual parts of the system see Anon 2 (date unknown) where this is described as
10,000 meter thinking. This “big picture’ viewpoint is also often described as the
holistic view. In practice this could mean for instance looking at the overall behaviour
of a system such as a business rather than looking at individual facets of that business.
The sometimes contrary and oft described opposite of the holistic approach is the
reductionist approach where an attempt is made to understand a ‘system’ from an
examination of its individual parts. This latter approach is closer to the norm for
existing systems analysis methodologies such as Structured Systems Analysis and
Design methodology (SSADM), and the Object Modelling Technique (OMT) and any
number of others. There are also specific techniques which adopt a mixed
holistic/reductionist stance that could be examined but that do not ‘belong’ to any
particular methodology. One example being Agent Based Modelling, Cannessa &
Riolo (2006) Kuo et. al. (2005), further investigation of these techniques and
methodologies is beyond the scope of this paper.
In adopting the systems approach there is often an implicit if not explicit assumption
that the system and its behaviour comprise more than the sum of the individual
component parts. This idea is often expressed as synergy or the synergistic viewpoint.
It can also be readily appreciated that the holistic viewpoint is looked on as a top
down approach whereas the reductionist approach is looked on as a bottom up
approach.
This paper sets out to examine the detail behind these assumptions and terms by an
examination of the literature and a sample of case studies where one or other of the
two approaches have been applied (allegedly).
Note that throughout the paper I am using the two terms ‘problem’ and
‘understanding’ in an almost interchangeable way. The reasoning for this is that an
attempt to gain understanding can be seen as a problem however in many definitions
understanding a system, concept or issue is not seen as a problem merely an attempt
to understand. Thus the two terms are largely interchangeable.
Defining the Reductionist approach
Reductionism is seen as the traditional means of approaching problem solving, at least
in the large scale, and is common across many disciplines. Jackson (2005) provides a
useful overview of some of the underpinnings of reductionism. Reductionism also
goes under many names including ‘stepwise refinement’, ‘disaggregation’ and simply
‘breaking the problem down’. It should be noted that reductionism tends to refer to
understanding rather than problem-solving but the latter seems more apt in this case.
Reductionism as a principle has one major thing going for it — it works. How do we
know it works? We know it works because it has been used in practice for many
thousands of years. Problems are broken down into constituent parts and are, possibly,
reconstituted into a single whole solution or provide a single systemic understanding.
Alternatively the individual sub-problems can acquire relatively self contained
solutions with no reference to other parts of the understanding of the problem.
Reductionism is the paradigm of understanding that has been applied by the human
race to many differing scenarios with great success.
This idea of breaking problems into their constituent parts comes naturally to people
and goes some way towards relieving the burden of bounded rationality, Simon
(1957). That is, there are some large/complex problems, issues or concepts than can
only be properly understood by first subdividing them into their constituent parts.
With this approach there is the implicit assumption that there is little need or attempt
to understand the context or ‘whole’ problem or system.
Defining the Holistic approach
As has been noted in the introduction above SD is often referred to as a holistic, or
more closely aligned with a holistic, approach to understanding the dynamic
behaviour of a system than existing methods of problem definition and solution. We
find evidence, or support, for the holistic approach from a number of sources across
many disciplines, some examples follow; from the cultural domain we have Hofstede
et al (1993) and from the medical domain Roberts et al (2002). Hofstede describes the
benefits of a holistic approach by using the metaphor of a number of blind men
studying an elephant the resultant confusion about the overall structure being an
illustration of “the need for pooling subjective patterns”. That is if the blind men were
to share their individual insights they may perhaps understand that it is an elephant
that they are dealing with and not a “snake, a stick, a disk, a column, a wall or a rope”
of course they may not arrive at this understanding. Roberts describes evidence from
secondary statistical sources of the links between mind and body and a determination
of medical success based on an appraisal of the patient’s mental attitude to illness.
From this viewpoint a method of treating physical illness by changing mental attitude
is described. On a governmental scale there is evidence that the creation of the
American Department of Homeland Security is an holistic attempt to address security
issues, Baranoff (2004).
There are of course some challenges to the concept and definition of holism. Popper
provides us with a view on holism in general when commenting on the behaviour of a
swarm of gnats “this ‘whole’ can be used to dispel the widespread ‘holistic’ belief that
a ‘whole’ is always more than the mere sum of its parts. I do not deny that it may
sometimes be so” and in a somewhat challenging assertion “the cluster of gnats is an
example of a whole that is indeed nothing more than the sum of its parts” Popper
(1979, p. 210). This is somewhat counterintuitive at first glance for the swarm can see
in more directions, determine threats with greater sensitivity and generally behave in a
fashion which suggests that the swarm is in some way acting with greater overall
ability. Popper counters this idea with the argument that the movement of the swarm
is simply the “sum of the movements of its constituent members, divided by the
number of members”. It is this author’s contention that both views are correct, the
swarm cannot be any more than the sum of its parts, in the way Popper describes, but
it is also apparent that the swarm can sense more keenly and in a more comprehensive
fashion than any individual member and in this fashion it has characteristics as a
swarm that no individual member possesses; for example the swarm can see what is
ahead and behind at the same time. In this latter example each member acts as a
sensory organ for the whole.
In addition to general arguments about holism there are detailed arguments about the
nature of holism in general and distinctions between different types of holistic
relationship, Stanford Encyclopaedia of Philosophy (2006).
The major benefit of thinking/understanding in holistic terms is that it is completely
intuitive for most systems. Hence we view an elephant as an elephant not as 4 limbs,
tail, trunk and head. The same applies in business; externally we see a business as a
business not as the individual silo’s or components of its operations. The internal view
is likely to differ.
There may be a further challenge to holism from Chaos theory which introduces
concepts of the unpredictability of systems which in a holistic sense is not acceptable.
In any case it may not be possible to produce definitive answers using holistic or ST
methods but nonetheless patterns are observable which at some level rebuts Chaos
theory.
With this holistic approach there is the implicit assumption that there is little need or
attempt to understand the individual components of the system so long as an
understanding of the overall behaviour is gleaned.
A third view of holism - reductionism
There are many other means and methods of defining holism and its uses, Rebernik
and Mulej, (2000). One further definition of holism that will be examined here could
be that it is an attempt to understand each of the parts of the system by first
understanding the whole system. In a similar fashion reductionism can be seen as an
attempt to derive a holistic understanding from an understanding of all of the
constituent parts of the system. Both of these methods of defining holism and
reductionism are something of a compromise from the two definitions presented
above. Nonetheless these latter two definitions are in common usage, see
http://en.wikipedia.org/wiki/Reductionism and http://en.wikipedia.org/wiki/Holism
for examples.
The advantage of this (pair of) view(s) is that it represents something of a continuum
and as both perspectives are somewhat interchangeable and have similar outputs, an
understanding of the overall system and the constituents thereof, it is very flexible.
The role of Holism in System Dynamics
System Dynamics can be viewed as a tool which on its own will describe the
behaviour of a system over time. The latter part of the previous sentence is of course a
truism as without the temporal dimension there is no behaviour. The simple bathtub
example is presented below as typical of a model than can be developed directly in
SD with no reference to further modelling techniques or more comprehensive
surrounding methodology. As a representation of a real world system the model can
still produce useful insights and outputs as it stands.
fate of inflow Rate of out
Water level
ie Be)
Antlow: Outflow
Figure 1: The bathtub model in Stella 7
@ 1: Water level
0.00 a0 6.60 bo 1200
Fave 1 Tine 14:00 Now 7004
Saas 2 Wate evel
Figure 2: Behaviour of the water level in the bathtub model,
given Inflow Rate = 1 and Outflow Rate = 0.5
A more typical and comprehensive approach to developing a SD model is typically
structured around the following steps; Problem recognition, Conceptual modelling,
Causal Loop Diagramming, System Dynamics models, Experimentation (sensitivity
analysis) and finally implementation, distilled from, Sterman (200, p89) and Towill
(1993, p205).
Focussing for now on the modelling aspects of this approach we can define an inverse
pyramid of holism-reductionism. This is shown in figure 3 below.
Conceptual
Model fa
5 3
3 —E——EE a
= g
s m
a Causal Loop 3
8 Diagram 3
a oO
& fe)
Ww J
5 — P
3 System 5
(o} Dynamics =
Model 2.
rr)
3
yy
Figure 3: Inverse pyramid of holism-reductionism for differing levels of modelling
Relating the SD process to holism - reductionism
There are many books and papers that describe a process such as that outlined above
for moving from problem definition to completed SD model, Sterman (2000), Towill
(1993). The vast majority of these describe a process that moves from an
understanding of the whole to an understanding of the parts. That is they take the first
of our third definitions of holism given above and apply that process to the problem or
issue at hand. It is the contention of this paper that the definition of holism used in
these cases and described above in the third definition is not the kind of approach that
most people would understand to be holistic.
Why is it important to get the definition right?
SD has been around for a long time and ST for a long time before that; even if it went
under other names. SD has not however been adopted into the mainstream of systems
analysis, understanding, decision support or problem solving tools and techniques. Its
degree of deployment does not compare in any way, for example, to that of object
oriented modelling, Goth (2002). Further there is support for the view that there is a
lack of growth in the field from within the SD community, Stevenson (2007) and this
disquiet is not new Towill (1993, p204).
It is important to get the definition right because to move the field of SD forward as a
credible tool for the enhancement of understanding we need to sell that tool to people
out there in the real world who are grappling with real world problems. If we give
people a definition of SD as a holistic systems approach to problem solving without
mentioning or properly explaining that the definition of holistic is not all it might at
first appear to be then we are not selling what we say we are.
This lack of clarity and others of a similar type has done and will lead to
disillusionment and ultimately rejection of the field of SD as a whole by the public
and perhaps practitioners as well.
Reductionism in SD
To illustrate the potential confusion between reductionism and holism figures 4a and
4b below show the two generic looping structures that characterise SD. (Both figures
were created in Stella 7)
Population
oa!
Births
Figure 4a: A reinforcing loop
Population, =
Deaths
Figure 4b: A balancing loop
In figure 4a as population increases this feeds back to births which also increase. In
figure 4b as Population increases, Deaths increase, causing a negative change in
population. The latter example assumes some other factor which is increasing
population, such as births, this has been omitted for clarity.
Now you know what each of the systems in figures 4a & b are because I have given a
description and labelled them as reinforcing or balancing loops. However the smallest
unit of structure that is visible is not the loop but any of the components of the loop
and beneath this level is a further level of detail which is the equation and within the
equation individual expressions and then individual terms. In fact to get either of these
structures to perform as I have described I need to adopt a reductionist viewpoint and
get in there with the equations and produce the underlying mathematical structure that
determines behaviour, almost regardless of how I draw the loops.
Population
Births
Figure 4c: a loop
To be contrary I could use the loop drawn in figure 4a change the underlying
equations and permit negative inflows so that as population increases births decrease
which turns the reinforcing loop back into a balancing one. I accept that these loops
are simplistic however the fact that they do exist and will work as described illustrates
the object of this exercise which is to make the confusion between a high level
(holistic) system view and a low level contradictory (reductionist) one evident. Note
that the determination of loop polarity follows the rules in Sterman (2000, p143-147).
The reductionist viewpoint is therefore necessary to convert the graphical (holistic?)
description of the system to one that matches my preconceived ideas of same.
Conclusions and policy recommendation
Being pragmatic about it this paper does not set out to produce any startling
conclusions or bring about any paradigm shifts it seeks simply to produce a little
clarification on an area of SD and ST that is sometimes taken for granted. As Sterman
points out “all models are wrong” and there is no attempt here to preach perfection or
the ‘right’ way to do things, Sterman (2000, p846). Further support comes from
Towill “there is an undoubted ‘art’ side to building a system dynamics model because
establishing the ‘truth’ requires creativity”, Towill (1993a, p206).
My conclusion is therefore that SD is not an holistic means of systems modelling
except that it is partly and the other part of it is reductionist and the two are
inseparable. A logical extension of this is that holism and reductionism are themselves
inseparable other than as subjective viewpoints.
I further conclude that ST is a purer holistic modelling technique than SD and by
definition I am differentiating the two.
Tam also concluding that the fact that SD is not a purely holistic method of systems
modelling is not a problem; selling it as if it were is.
Thope that I have illustrated in a meaningful and straightforward way the necessity of
truly understanding the fundamental nature of SD as a blended continuum of holism
and reductionism. Due to the subjective nature of both concepts neither (label)
description is likely to be satisfactory on its own to cover the entire scope of options
for development within SD.
This has been a somewhat abstract look at the debate between holism and
reductionism, though with a practical message at its heart, and several other similar
topics are available for further work in the same vein. We might for instance think of
the mantra that ‘reinforcing loops produce exponential behaviour’ as worthy of
examination as it appears to be incorrect in specific cases and therefore incorrect as a
rule. This and several other concepts at the heart of SD would be well worth a re-
examination.
The message in any further work would be the same - get it right before we sell it or it
will get returned as it has been in the past.
References
Anon 1, Holism and Nonseparability in Physics, Stanford Encyclopaedia of
Philosophy, downloaded from http://plato.stanford.edu/entries/physics-holism on 29
September 2006
Anon 2, Systems Thinking and the STELLA Software: Thinking, Communicating,
Learning and Acting More Effectively in the New Millennium, downloaded from
http://www.iseesystems.com/resources/Articles/STELLA %20IST%20-
%20Chapter%201.pdf on 27" February 2007
Baranoff E G, Risk Management: A Focus on a More Holistic Approach Three Years
After September 11, Journal of Insurance Regulation; Summer 2004, Vol. 22, Issue 4,
p71-81, lp
Canessa E E C, Riolo R LR R, An agent-based model of the impact of computer-
mediated communication on organizational culture and performance: an example of
the application of complex systems analysis tools to the study of CIS, Journal of
Information Technology (Palgrave Macmillan), December 2006, Vol. 21 Issue 4,
p272-283
Goth G, Has Object-Oriented Programming Delivered? IEEE Software,
September/October 2002, downloaded from
http://ieeexplore.ieee.org/iel5/52/22181/01032867.pdf2arnumber=1032867 on the 27"
February 2007
Hofstede G, Bond M H, Chung-leung L, Individual Perceptions of Organizational
Cultures: A Methodological Treatise on Levels of Analysis, Organization Studies,
Vol. 14 Issue 4, p483-503, 1993
Jackson M, Systems thinking, Creative Holism for managers, Wiley, 2005
Johannessen J, Olaisen J, Systemic philosophy and the philosophy of social science,
Part I: Transcendence of the naturalistic and the anti-naturalistic position in the
philosophy of social science, Kybernetes, Vol 34, Number 7/8, 2005
Johannessen J, Olaisen J, Systemic philosophy and the philosophy of social science,
Part II: the systemic position, Kybernetes, Vol 34, Number 9/10, 2005
Kuo C, Huang H, Jeng M, Jeng L, Separation model design of manufacturing
systems using the distributed agent-oriented Petri net, International Journal of
Computer Integrated Manufacturing, March-May 2005, Vol. 18 Issue 2/3, p146-157
McEntire D A, Fuller C, The need for a holistic approach: an examination from the El
nino disasters in Peru, Disaster Prevention and Management, Vol 11, Number 2, pp.
128-140, 2002
Natke H G, Cempel C, Model-based diagnosis of systems emphasising a holistic
approach, International journal of Systems Science, Vol 31, Number 11, pp. 1497-
1504, 2000
O'Loughlin A, McFadzean E, Toward a holistic theory of strategic problem solving,
Team performance management, Vol 5, Number 3, pp. 103-120, 1999
Popper K, Objective knowledge, An evolutionary approach, Oxford University Press,
1979
Rebenik M, Matjaz M, Requisite holism, isolating mechanisms and entrepreneurship,
Kybernetes, Vol 29, Number 9/10, 2000
Roberts S A, Kiselica M S, Fredrickson S A, Quality of life of persons with medical
illnesses: Counselling’s holistic approach, Journal of counselling and Development,
Volume 80, Fall 2002
Simon H, "A Behavioral Model of Rational Choice", in Models of Man, 1957
Sterman J D, Business Dynamics, Systems Thinking and modelling for a complex
world, McGraw-Hill Higher Education, 2000
Stevenson R, Query The death of System Dynamics? System Dynamics mailing list,
downloaded from http://www.systemdynamics.org/pipermail/sdmail/2007-
January/000165.html on 20th February 2007
Towill D R, System Dynamics — background, methodology and applications, part 1,
Background and Methodology, Computing and Control Engineering, October 2003
Towill D R, System Dynamics — background, methodology and applications, part 2,
Applications, Computing and Control Engineering, October 2003
