Page 209 - DCAP601_SIMULATION_AND_MODELING

Unit 12: Design and Evaluation of Simulation Experiments (II)

Notes
Figure 12.1: Drogoul’s and his Colleagues’ Interpretation of Gilbert’s and
Troitzsch’s Methodological Proposition on the Role of Simulation

This diagram does not even contain the word ‘simulation’, but in its centre we find ‘data’ which
are taken from observation and compared with results from simulation runs, for their similarity.
Gilbert’s and Troitzsch’s original diagram describing “the logic of simulation as a method”
(Gilbert and Troitzsch 1999: 16, 54, see also Troitzsch 1990: 2) is much the same: A model is built
by abstraction from a target system, it is translated into a computer programme which can then
be run and delivers results in the form of simulated data which can, and have to, be compared to
data gathered from the same kind of target systems in the real world from which the model was
abstracted.
Being aware that observation (as contrasted to just looking around in the world) presupposes at
least some primitive form of theory (which tells us which entities and which of its properties to
observe and which relations between them to register to find out whether there are some
regularities), we should admit that our assumptions and our observation are not independent
from each other (although Figure 5.1 insinuates this). And we should admit that in most cases
computational (and other) models do not directly start from observation data but from a theory
which in turn should build on, but often does not refer explicitly to observation data. Instead, we
often start from a verbal theory which expresses our (or other authors’) belief in how reality
works, comparing simulation results with stylised facts instead of observation data. A good
example of this strategy is Sugarscape where the question “can you explain it?” is interpreted as
“can you grow it?”, and where “a given macrostructure [is] ‘explained’ by a given
microspecification when the latter’s generative sufficiency has been established.” (Ep- stein and
Axtell 1996: 177)
At the other extreme, we might have microanalytical simulation which starts from a large
collection of observational data on persons and households and the population as a whole. The
model is initialised with empirical estimates of transition probabilities, age-specific birth and
death rates and so on. Tens of thousands of software agents are created with data from real world
people. And all this aims at predicting something like the age structure or kinship networks of
this empirical poulation in the far future (see for instance Harding 1996).

LOVELY PROFESSIONAL UNIVERSITY 203