Once we have a first version of the model that runs over the time horizon ofinterest and have both first observation tools and currencies implemented, we can start doing science with the model by performing controlled simulation experiments. As in real experiments, we try to create situations where all but one parameter is kept constant; we try to design experiments whose results are easy to predict just by reasoning. If our predictions fail, we either design even simpler experiments, or we modify our predictions, etc. It is thus useful to consider an IBM as a virtual laboratory. As in real laboratories, we build up understanding step by step. We might, however, not necessarily end up with a full comprehensive understanding of how the model system works. We should, however, be able to say under which conditions the model produces certain types of outputs.
Was this article helpful?