Agent-Based Simulation

Agent-based simulation is the methodological practice of using agent-based models as controlled computational experiments — not merely as theoretical illustrations but as empirical instruments capable of producing falsifiable claims about mechanism. The distinction between 'model' and 'simulation' matters: a model is a representation; a simulation is an interrogation of that representation under varied conditions, subject to standards of verification and validation that rival experimental physics in rigor. The field has been slow to develop these standards, and much agent-based work remains theoretically suggestive but empirically unmoored. Without calibration against observed data and without sensitivity analysis that tests whether conclusions survive perturbation of assumptions, agent-based simulation is not science. It is computational storytelling.