Abstract
Social simulation studies are complex, because they typically combine
various sources of data and hypotheses, that are integrated by
intertwined processes, of model building, simulation experiment
execution, and analysis. Various documentation approaches exist that
support transparency and traceability of social simulation studies. The
exploitation of provenance standards allows for making the information
about what sources and activities contributed to the generation of an
entity, e.g., simulation model, queryable and computationally
accessible. Therefore, provenance patterns have been defined to capture
central activities and entities. Activities include model building,
calibration, analysis, and validation. Entities are simulation model,
simulation experiment (its specification), and research question. Here
we refine and extend this approach to address specific challenges of
social agent-based simulation studies, i.e., activities such as
collecting and analyzing primary data about human decisions, or
collecting and assessing the quality of secondary data. This allows us
to tell the whole story of these simulation studies in a comprehensive
manner. We illustrate the potential of the approach by applying it to
central activities and results of the Bayesian Agent-Based Population
Studies project and implementing it in a web-based provenance tool.