This on-going project aims to build an ontological model concerning recreational poly-drug use and to use an agent-based simulation, SimUse, to test and verify related public policies. We consider drug-use (and even more so poly-drug use) to be a complex adaptive system that needs to be studied via a methodology able to describe such a dynamic and complex social phenomenon. To contextualize the complexity of poly-drug use and highlight the necessity of a multidisciplinary approach to study this practice we will pay particular attention to justify our theoretical assertions and computational constructions by qualitative researches, statistical surveys and/or theoretical concepts. The description of the different levels composing the structure of drug use and the dynamics between these levels leads us to consider an ontological approach as the best way to capture and encapsulate poly-drug use. We will describe the global structure of our simulation via an UML diagram as well as the way we want to model the decision process concerning drug consumption. We finish with an illustration of tests that can be run with SimUse: this example presents results of different law enforcement policies and their consequences for decision-makers.