Law enforcement occurs in a complex environment that contains a variety of actors that interact with one another. These interactions create emerging collective behavior over time. For example, inspectors will try to influence non-compliant actors to become compliant, while inspectees may comply or thwart inspections. Inspection agencies such as the Inspectie Leefomgeving en Transport (ILT) evaluate inspectees’ adherence to regulation and aim to boost compliance within an industry. However, they have limited resources and a wide range of potential societal challenges to address. They face an action dilemma, having to decide on what actions to take without full knowledge of whether their actions lead to higher compliance or improved social outcomes. Previous studies of the inspection environment rely on behavioral theories to investigate the underlying motivations of inspectees’ behavior. However, these theories presume inspectees’ motivations and characterize them homogeneously, often assuming they have perfect rationality. This leads to an inaccurate depiction of inspectees, reducing them to one-dimensional actors when in reality, their behavior is motivated by multiple factors and can be idiosyncratic. Data science techniques provide the opportunity to understand behavioral phenomena with data, leveraging datasets to identify statistical patterns in behavior to help inspectorates make decisions within a degree of certainty. A particular modeling technique that focuses on representing empirical data without assuming behavioral motivations is phenomenological modeling. Coupled with agent-based modeling (ABM), phenomenological modeling allows the researcher to simulate possible outcomes of observed behavior before the underlying motivations are understood. This provides insight into the macro-level behavior produced by micro-level interactions within the complex inspection environment...