Page 183 -

180 N. David et al.

Purpose of models

General goal of simulating social complexity

Basic methodological conceptions
(types of validity:
through prediction, retrodiction, structural similarity)

Validation techniques
(diverse)

Relationship to modelling strategies
(subjunctive models, context-specific models)

Fig. 9.2 Validation implies considering the purpose of the model

There are also different methodological motivations behind the use of a model,
such as those conceived to predict or explain and those merely conceived to describe.
Regardless of what method is used, the reproduction of characteristics of the object
domain is important, but this can be assessed through rather different approaches
during the model development process. If it is prediction you are seeking, validation
consists of confronting simulated behaviour with the future behaviour of the target
system (however, attempting to establish numerical prediction is not a normal goal
in simulation). If it is explanation, validation consists of building plausible mech-
anisms that are able to reproduce simulated behaviour similar to real behaviour. If
the goal is the more general aim of descriptiveness, explanation may probably be a
goal as well, and a creative integration of ways for assessing the structure and results
of the model, from quantitative to qualitative and participatory approaches, will be
applied.
In conclusion, one should bear in mind that there is no one special method for
validating a model. However, it is important to assess whether the simulation is
subjected to good practices during its conception, whether it fits the intended use of
the model builder and whether it is able to reproduce characteristics of the object
domain. Assessing whether the goals of the modellers are well stated and the models
themselves are well described in order to be understood and sanctioned by other
model builders are sine qua non conditions for good simulation modelling.
In the remainder of this section, we revise the purpose of validating simulations
along three dimensions: (1) the general goal of validation in social complexity;
(2) basic methodological conceptions of validity types; and (3) typical validation
techniques used in social simulation.