Stever,
I know you werent asking about this, but your question
triggered some thoughts on soft modeling.
Basically, I think that the challenge in soft modeling is to
make it as UN-soft as possible.
The reason for this is that soft modeling tends to be abstract
and can end up merely being a direct translation of a dynamic
hypothesis -- a set of loops -- into a mathematical model.
This can add value, but not as much value as if modeling is
also driven by the need to represent physical (not quite the
right word, "concrete"?) structures. Modeling the physical
structure requires people to create loops that they hadnt been
focussed on before.
I know youre familiar with model X and model Z. For people
who may be listening in: X and Z are two models of a dynamic
hypothesis related to federal financial gurantees. One model
represents the hypothesis at an abstract level and the other at
a concrete level. We learned a lot from the concrete model; we
would have gained very little from the abstract model -- it
simply represented the loops we knew about. The two models
were comparable in size, but in the concrete model we
inadvertantly added loops in order to be concrete. The loops
needed to be in our model for a physical representation; they
had to be in the real world if our dynamci hypothesis could
exist -- and they completely changed our understanding.
The model of love that you refer to is good to the extent that
ideas like irritation (cabin fever) and missing someone (Eliza
Doolittle) are linked to physical things -- like how much time
you actually spend with someone, or the stock of
characteristics that you like in someone. The model is
unsuccessful to the extent that it doesnt do this. In Jays
market growth model, "bias" -- a soft variable --is very
important. But the representation of bias is a single
parameter; while the rest of of the model deals mostly with
more concrete things -- sales, backlogs, number of salesmen,
capacity.
There are lots of examples of important soft variables in
models -- including fatigue and morale in project models,
influencing people in Bass-type diffusion models, expectations
in almost all models (including market growth). These soft
variables are powerful because they are found in models that
are firmly anchored in concrete reality.
And now, your reward for reading through an answer to a
question you didnt ask: Take a look at Stermans model of
Kuns structure of scientific revolutions; also, Anjelai Sastri
is working with a model of "punctuated equilibria" for her
dissertation. (Hey! you didnt skip to the bottom, did you?).
Regards,
Jim Hines