Keynote Speaker: Credible Simulation Models - Provenance beyond Reproducibility

Abstract

When expressing concerns about the credibility of simulation studies, simulation data have been traditionally in the focus. However, what about another and, some might argue, even more central product of simulation studies, i.e., the simulation model itself? How can the credibility of a simulation model be assessed? Therefore, information about the process of generating a simulation model is needed. This provenance relates entities (or artifacts) and activities involved in the generating process. Based on simulation studies we will illuminate how the provenance of a simulation model relates the refinement, extension, composition, calibration and validation of simulation models to the diverse sources used in these processes. To exploit this information, unambiguously means for specifying entities play a central role. For example, a formal domain-specific language for modeling facilitates assessing and reusing simulation models. Similarly, a declarative domain-specific language for specifying simulation experiments, helps utilizing simulation experiments done with earlier models for future models. Thus, provenance, information about the past, does not only allow to understand the present, but also to design the future, in opening up new avenues for generating and analyzing simulation models.