REPLY Models, Problems and Systems (SD7036)

[SDMAIL John Morecroft]

Posted by  "John Morecroft" <jmorecroft at london.edu>

Jack (Homer)
I was struck by your observation that a well defined problem situation
can actually be broader than the 'perceived' system. So I ended-up
spending part of a sunny May Sunday thinking about model
conceptualisation.......

I often say to my MBA students at LBS that SD professionals prefer to
model a problem situation, rather than 'the system' in order to be
discriminating about which factors to exclude from the model.  Modelling
is the art of 'leaving things out'.  A disciplined and reliable way to
converge on essential structure is important.  But I accept your
corollary that, if one's starting viewpoint is narrow and event-oriented
then yes, you first need to expand the boundary of factors deemed
relevant.  Sterman's multi-loop representation of factors affecting
traffic congestion (BD Chapter 5) is a great example.  The question is
when to stop.  I explain this winnowing-down of messy situations in
terms of the iterative process of modelling: 'I Spy dynamics in the
complexity of everyday life and can discover (following SD guidelines)
underlying feedback structure'. See my SMBD book Chapter 5 on 'Cyclical
Dynamics and the Modelling Process'.

To begin modelling with problem articulation and a dynamic hypothesis is
both a dictum and distilled wisdom of the field, nicely captured in
figures 3-1 and 3-2 of Sterman's BD chapter 3.  However I must admit
that sometimes my own experience of modelling has been different -
perhaps because modelling is fundamentally a creative process.  In
particular I remember a modelling project with BBC World Service, a
well-respected international radio broadcaster specialising in news and
current affairs.  The stated purpose of the project was to 'develop a
model to explore 10-year strategy scenarios and to support the
organisation's bid to government for future funding'. For this purpose,
World Service was conceived as a dynamical system in which public funds
from Government are deployed in various ways to build broadcasting
assets that collectively attract listeners.  There was no explicit
dynamic hypothesis, but nevertheless the project team was able to create
a reasonably compact and focussed model built around eight stock
accumulations. The structure took shape without a particularly clear
up-front or ex-ante view of feedback loops.  Did this project depart
from dictum or simply re-interpret the dictum to fit the situation?
(More information about the project and the model can be found in my
1999 article 'Visualising and Rehearsing Strategy', Business Strategy
Review, 10(3) 17-32.)

Was this project experience an anomaly?  I doubt it.  I think it's
simply a manifestation of a need for 'crafting freedom' during model
conceptualisation (but without losing sight of one's discipline
guidelines).  Let me finish with a short story about the art, craft and
science of model conceptualisation.  The story is an edited excerpt from
my 2000 article 'Creativity and Convergence in Scenario Modelling',
97-115 in a Festschrift honouring Erich Zahn, (editors Habenicht and
Waeschser), Schaeffer-Poeschel-Verlag, Spring 2000.
 
************************************************************************
***
Conclusion
Model conceptualisation requires an unusual blend of creative and
convergent thinking.  And that is why conceptualisation is such a
rewarding activity.  There is no recipe for building a good system
dynamics model of a firm, an industry or a society. There are just
guidelines and principles of feedback systems - the theory, craft and
distilled wisdom of the discipline (as found in the literature and SD
textbooks). In my experience the process is one of discovery with sudden
breakthroughs: using expert knowledge, conversations, the collective
mental database of the modelling team and principles of modelling to
make sense of dynamic complexity.  Sooner or later a pattern emerges, a
framework for the model, to be refined and clarified through mapping and
dialogue.  But progress toward this framework can be uneven - sometimes
it is very fast, sometimes it is slow. 
	
I am reminded of the time many years ago when I was on the faculty at
MIT's Sloan School.  I used to run a semester-long course for MBAs and
PhDs on "Applications of Industrial Dynamics".  The course required
students to conceptualise and build a model addressing a practical
business problem.  These students had already been introduced to the
basic principles of system dynamics and so were ready and eager to
undertake a self-contained project of their own choosing.  The course
included some lectures as well as seminar-style discussion of
models-in-progress.  I thought it was important to include a lecture on
model conceptualisation, so I invited Jay Forrester to talk about his
experiences.  I thought he would talk about the steps he followed in
this creative stage of modelling.  He did, but not in the way I had
expected.  He picked two quite different examples from his repertoire of
projects:  the corporate growth model he built for the fledgling Digital
Equipment Corporation in the early 1960s, and the World Dynamics model
from the early 1970s.

His lecture had a dramatic effect on my students (and me too).  He began
by holding-up a single sheet of paper.  It contained two large ovals -
one labelled "company" and the other labelled "market".  Between these
two ovals were lots of connecting arrows, some going from the company to
the market and some going the other way.  He then went on to explain
that this single page with its ovals and arrows was the result of two
years model conceptualisation while he was sitting on the board of the
newly formed Digital Equipment Company.   He stressed that his efforts
were by no means full-time.  But more importantly he made the point that
he needed this elapsed time to think how best to approach the problem of
representing Digital as a growth firm.  Along the way he had rejected
many possibilities to avoid detail unnecessary for exploring the
dynamics of growth over a ten-year period (for example he chose to
ignore individual product lines, many of which would come and go over
ten years).  Thereafter only eight weeks were required to create the
entire growth model of 200 equations.

After explaining much more about the structure of the Digital Equipment
model he then moved to his second example - World Dynamics.  At this
point he held-up the two pages of the World Dynamics book that show the
complete stock and flow structure of the model and the information
network (the same picture I mentioned at the start of this article).  He
then said he had created a rough sketch of this picture in an 8 hour
transatlantic flight from Paris to Boston following a meeting of the
Club of Rome!  (You can now see the original for yourself.  It appears
as figure 7 in David Lane's 2007 article, 'The power of the bond between
cause and effect: Jay Wright Forrester and the field of system
dynamics', SDR 23 95-118, the issue celebrating 50 years of system
dynamics).  

The contrast between the two examples was dramatic.  Two years for a
sector map of Digital Equipment Corporation, and eight hours for a
close-to-complete structural map of World Dynamics. The lesson is that
good model conceptualisation cannot be tightly scheduled.  Creativity
and convergence need time.  It is much better to wait for a satisfactory
modelling framework to come to mind than to rush into premature equation
formulation and simulation.  Sometimes conceptualisation happens very
quickly, as for World Dynamics, and sometimes more gradually as for
Global Oil.  Occasionally, as with the Digital model, the right approach
is a long time in gestation.  The reward is a framework for thinking
about dynamic complexity with the power to capture the knowledge, time
and attention of policymakers, and even other academics. And that's
where the fun of modelling lies too.             

_________________________________

John Morecroft | Senior Fellow
Management Science and Operations
London Business School
Posted by  "John Morecroft" <jmorecroft at london.edu>
posting date  Mon, 12 May 2008 13:16:45 +0100