Folks,
Heres a small further contribution to the qualitative versus quantitative
debate.
Some of you may remember C P Snow (later Lord Snow). For those who do not,
he came from humble origins, read physics at Cambridge and (despite that
handicap) went on to become very great and very good in the British
establishment. He is famous for a lecture he gave in about 1959 in which he
bemoaned the emergence of two cultures - roughly scientific and liberal
arts - each of which expressed important aspects of the human condition but
neither of which understood, nor cared to understand, the other.
During and after a very distinguished career in the civil service he became
a novelist with a series called Strangers and Brothers in which he traces
the life of a semi-autobiographical character. These are, unfortunately, no
longer much read but they have always fascinated me so I read them from time
to time. Yesterday I was re-reading the eponymous novel of the series in
which one of the characters makes the following remarkable statement:
"What a wonderful invention a map is. Geography would be incomprehensible
without maps. Theyve reduced a jumble of facts into something you can read
at a glance. Now I suspect that economics is fundamentally no more difficult
than geography. Except that its about things in motion. If only someone
could invent a dynamic map."
Bear in mind that this was published in 1940!
It seems to me that this is what the qualitative phase of SD does for
problems with both soft and hard variables, and that makes SD into an
amazing
discipline. I know of no other operations research method which even comes
close.
However, a geographical map is drawn using clear and accepted conventions.
We seem to have evolved three variants: Stock/flow diagrams, CLDs, and
Influence diagrams. I wouldnt die in a ditch over which is used except
that, for me , SF diagrams are cumbersome to draw and CLDs typically involve
a smallish number of variables (there is no reason why that has to be so but
its typical of practice) and dont distinguish between physical effects
and those of information flow and management action. An ID does make that
distinction where it is useful to do so and can be drawn with larger numbers
of variables - 50 or 60 is not uncommon (there are numerous examples in my
books and papers and in the writings of others, notably Eric Wolstenholme).
I am, however, VERY clear in my mind on two points.
The first is that mapping is an intellectual exercise which has to be done
with pencil and paper. I have, to my horror, seen someone trying to draw an
ID directly into a graphics package. The role of software is to record
thought not to support it and that person predictably got a bad result
because he was wasting his brain time in clicking on icons and dragging them
about, not on understanding a problem. (Im not going to be PC and write she
when the person in question was male).
The second is that a map puts onto ONE piece of paper a model of the
geography drawn to a useful scale. A good IDs main strength (and this is
also true of a good CLD or SF) is a portrayal of complex problem. That
accomplishment is of inestimable practical value so my question in the
debate is to ask for demonstration of the added value of quantification in
cases where, as I have tried to explain, there may be serious uncertainties
in data and causal mechanisms.
This second, one piece of paper, argument makes we worry about cases I have
seen in which people try to understand and explain a model by taping
together 10 or 15 sheets of SF diagram as printed out by a graphics package.
Yes, I KNOW that some of the packages have a tunnel-down feature to produce
aggregated SF diagrams but that, of course, is not the same as the full
diagram and conceals much.
Well, there it is, but C P Snow still gets my vote as a wise old bird.
Weve also had some chat about scales and Jay suggested how one might define
a scale for integrity. That intrigued me and I found two slightly different
dictionary definitions of the term. One mentions moral principles, the other
honesty, and both have the idea of upright behaviour. Of course, these are
all rather imprecise and I wonder if two or more people would have the same
understanding of integrity, even if they shared a common cultural
background. If they had different backgrounds they might have very divergent
ideas of what adhering to moral principles might mean. Suppose, however,
that 5 people knew that Smith (whom they all know) was a 3 on Jays scale
and Brown (also well known) was an 8 and that they agreed, roughly, with
that. They are now asked to position Jones on that scale. I suspect that the
best they would do is to agree on where Jones lies relative to Smith and
Brown but with no more than better/worse/in between precision. Suppose they
agree on in between, but I seriously doubt that they would all agree that
Jones is a 5 and there would be some degree of dispersion between, perhaps,
4 and 6. Thus, while one can certainly make up a scale for anything you
choose, the meaning of that scale (what is Integrity?), its calibration (do
we all agree on Smith and Browns positions) and its practical use (where is
Jones to be located) are all FRAUGHT with grave uncertainty. Does anyone
know what the psychometric literature says about this sort of thing, as Jay
and I seem to be doing an awful lot of guesswork and assertion.
Someone mentioned doctors asking about pain on a scale from 1 to 10 (in
fact, it would have to be zero for a complete absence of pain). I asked a
few doctors I know all of whom said they would never waste time on such a
silly (their words) question. The reason is plain. Zero means something but
what is 10? The same as the pain from a broken leg? Ive never had one so I
dont know. The pain of childbirth? Im unlikely to know that! In any case,
childbirth varies enormously from case to case, or so I understand.
In short, these qualitative scales seem to me to require to be used with
very great care. Software will accept anything; the question is whether it
means anything when youve programmed it and whether the results take you
very far, or even take you in the wrong direction.
Regards,
Geoff
Professor R G Coyle,
Consultant in System Dynamics and Strategic Modelling,
Telephone +44 (0) 1793 782817, Fax ... 783188
email
geoff.coyle@btinternet.com