FREIDA: A Framework for developing quantitative agent based models based on qualitative expert knowledge: an example of organised crime

Developing ABMs of organized crime networks supports law enforcement strategies but is often limited by scarce quantitative data. This challenge extends to other psychosocial contexts like mental health and social systems. While qualitative data from reports and interviews is more accessible, current ABM methodologies struggle to integrate both data types effectively. To address this, we propose FREIDA, a mixed-methods framework that combines qualitative and quantitative data to develop, train, and validate ABMs in data-sparse contexts. FREIDA's four-phase process includes data acquisition, conceptual modeling, computational implementation, and model assessment. Using Thematic Content Analysis (TCA), Expected System Behaviors (ESBs) are translated into Training Statements (TS) for calibration and Validation Statements (VS) for assessment. Iterative sensitivity analysis and uncertainty quantification refine the model's accuracy. We apply FREIDA to a case study of the Netherlands cocaine network, producing the Criminal Cocaine Replacement Model (CCRM) to simulate kingpin removal dynamics. FREIDA enables robust ABM development with limited data, aiding law enforcement decisions and resource allocation.