Page 177 -

174 N. David et al.

9.1 Introduction

The terms verification and validation (V&V) are commonly used in science but
their meaning may be controversial in the natural and the social sciences. Putting
aside the epistemological underpinnings of the terms, in simulation the distinction
of meaning has a mere pragmatic nature inherited from computer science and
software engineering. Often, verification is used in the context of evaluating the
computational implementation of a model in terms of the researchers’ intentions. In
turn, validation typically refers to an evaluation of the credibility of the model as a
representation of the subject modelled.
In disciplines that make use of computational models, the role of V&V is
related to the need of evaluating models along the simulation development process.
Basically, the very idea of V&V is comparing models with observations and
descriptions of the problem modelled. This may include other models that have been
verified and validated to some level, or even the implementation of replications in
order to verify and validate models in more depth.
This chapter introduces a methodological perspective on V&V and describes
different strategies and techniques to validate models of social complexity. Some
aspects of what can be called either verification or validation are also discussed,
namely comparison between models and model replication, whereon verification
and validation are superimposed or indistinguishable. These are important but fre-
quently neglected methods of promoting V&V, particularly since social simulation
models can be very sensitive to implementation details (making them hard to verify),
and data from social systems can be difficult or even impossible to collect (making
the respective models hard to validate).
The use of simulation for modelling social complexity is very diverse. Often,
V&V do not refer to an explicit stage in the simulation development process, but
to the modelling process itself according to good practices and in a way that grants
credibility to using the simulation for a specific purpose. Normally, the purpose is
dependent on different strategies and dimensions, along which simulations can be
characterised, with reference to different kinds of claims intended by the modeller,
such as theoretical claims, empirical claims or simply subjunctive theoretical claims.
The term subjunctive is used when very abstract simulations are used for thinking
about scenarios in possible worlds, such as describing “what would happen if
something were the case.” There cannot be V&V without considering the purpose
of the simulation.
In the next section of the chapter, we will deal with the meaning of the terms
V&V in the context of the simulation development process. In Sect. 9.3, methods
and techniques commonly associated with validation are described. The comparison
and replication of simulation models as an essential aspect of V&V is discussed in
Sect. 9.4. The chapter closes with Sect. 9.5, where the relationship of validation with
different modelling strategies is described.