Page 178 -
P. 178

9 Verifying and Validating Simulations                          175

            9.2 The Simulation Development Process


            Several chains of intermediate models are developed before obtaining a satisfactory
            verified and validated model. What does it mean to verify and validate a model
            in social simulation? Is there a fundamental difference between verifying and
            validating models? The purpose of this section is to define the role of V&V within
            the scope of the simulation development process.
              The most common definitions of V&V are imported from computer science,
                                                       1
            as well as from technical and numerical simulation, having intended distinct—
            although epistemologically overlapping—meanings. The reason for distinguishing
            between the terms derives from the practice of determining the suitability of certain
            models for representing two distinct subjects of inquiry. This is represented in
            Fig. 9.1, in which V&V are related to a simplified model development process.
            Two conceptual models mediate between two subjects of inquiry. The latter are (1)
            the target theory or phenomenon and (2) the executable computational model. The
            conceptual model on the right, designated here as the pre-computational model,
            is basically a representation in the minds and writing of the researchers, which
            presumably represents the target. This model must be implemented as an executable
            computational model, by going through a number of intermediate models such
            as formal specification or textual programs written in high-level programming
            languages.
              The analysis of the executable model gives rise to one or more conceptual models
            on the left, here designated as post-computational models. They are constructed
            based on the output of the computational model, often with the aid of statistical


                                           Target
                                          theory or
                                          phenomenon
                      Validation                              Validation
                          Publication / Theory      Conceptualisation and
                        dissemination / Application  model construction

             Post-computational                                    Pre-computational
                models                                                models
                                                      Implementation
                          Conceptualisation and     (physical construction of
                           model construction
                                                    computational models)
                      Verification                             Verification
                                          Executable
                                         computational
                                           model
            Fig. 9.1 Verification and validation related to the model development process (David 2009)


            1
            Numerical simulation refers to simulation for finding solutions to mathematical models, normally
            for cases in which mathematics does not provide analytical solutions. Technical simulation stands
            for simulation with numerical models in computational sciences and engineering.
   173   174   175   176   177   178   179   180   181   182   183