Page 214 - DCAP601_SIMULATION_AND_MODELING

Simulation and Modelling

Notes And a simulation model which generates a macro structure which resembles real-world macro
structures from simulated micro structures which resemble micro structures observable in the
real world might be accepted as a provisional explanation of real-world macro structures.
In a second step we might apply simulation to proceed to a second stage of qualitative prediction,
where we are not interested in the general behaviour of a certain class of target systems, but in
the future behavior of a particular instance of this class of target systems — say, the future market
shares of a number of competing products in a market, trying to answer the question whether
most trademarks will survive with reasonable market shares or whether most of them will
survive only in small niches whereas one product will gain an overwhelming share of the
whole market; this would still be a qualitative answer: we might not be interested in which
trademark will be the winner, and we might not be interested in how many per cent of the
market it will win (this would be only the third use of simulation, namely to predict quantitatively
and numerically, as in microanalytical simulation and, perhaps, also in the simulation of climatic
changes where we would not be content with the outcome that mean temperatures will rise but
wanted to know when, where and how fast this process would have effects on nature and
society). Or, to return to the example of the lake, its eutrophication and the countermeasures
taken by its neighbours, we would
1. First apply simulation to the very general question whether an artificial society “living”
around an artificial lake which functions much like an empirical lake could ever learn to
avoid eutrophication (something like a tragedy-of-the-commons simulation),

2. Then apply simulation to an empirical setting (describing and modelling an existing lake
and its surroundings) to find out whether in this specific setting the existing lake can be
rescued, and
3. Eventually to apply simulation to the question which political measures have to be taken
to make the lake neighbours organise their economy in away that the best possible use is
made of the lake — and obviously this would be a discursive model in which stakeholders
should be involved to negotiate and find out what “best possible use” actually means for
them.

Notes Note that to involve stakeholders in the development of a simulation model like
this it will be necessary to validate the model— otherwise stakeholders would not believe
it was worthwhile to work with the simulation model.

12.2.3 Types of Validity

With Zeigler we should distinguish between three types of validity:
1. Replicative Validity: The model matches data already acquired from the real system
(retrodiction).
2. Predictive Validity: The model matches data before data are acquired from the real system.
3. Structural Validity: The model “not only reproduces the observed real system behaviour,
but truly reflects the way in which the real system operates to produce this behaviour.”
(Zeigler 1985: 5).

Zeigler here addresses three different stages of model validation (and development). Social
science simulation does not seem to have followed this path in all cases:
Since often data are very poor in the social sciences, early models, too, tried to be structurally
valid and did not bother much about replicative or predictive validity.

208 LOVELY PROFESSIONAL UNIVERSITY