Abstract

Papers that study real world phenomena with the help of agent-based models often offer direct policy advice. The measures recommended refer to measures taken by real policy makers, not to interventions in a model. Yet the recommendations are based only on the results that simulated interventions had in simulated worlds. So how can a policy maker justify a (real) policy choice on the basis of such simulation results? I discuss four decision procedures and their appropriateness for this question. In particular, I argue that the applicability of specific rational decision procedures to any given ABM depends on (i) the result of the AGM, (ii) the way the AGM represented the output, and (ii) the degree to which the AGM is empirically validated.