| Abstract | Recent years have seen a proliferation of agent-based models (ABMs), but with the exceptionof a few "classic" models, most of these models have never been replicated. We argue that
replication has even greater benefits when applied to computational models than when applied to
physical experiments. Replication affects model verification, in that it aids in determining if the
implemented model reflects the conceptual model. It affects model validation, since a
replication of a conceptual model may change the output from an implemented model and thus
alter the correspondence between the model and the real world. Replication also affects
validation by forcing the model developer and replicator to re-examine assumptions made in the
original model. In addition replication fosters shared understanding of the details of modeling
decisions within the research community. To facilitate the practice of replication, we argue for
the creation of standards for both how to replicate models and how to evaluate the replication. In
this paper, we present a case study of our attempt to replicate an ABM developed by Axelrod and
Hammond. We detail our effort to replicate that model and the challenges that arose in
recreating the model and in determining if the replication was successful.
|
