Tucker, Jason , Belfrage, M., Lorig, F., Frantz, C. & P. Davidsson | 2025
Conference paper, Annual Modeling and Simulation Conference (ANNSIM'25)
The application of Agent-Based Modeling and Simulation (ABMS) has few established guidelines and often suffers from insufficient model documentation. We assess the prevalence of best practices associated with different types of model documentation in light of the European Union’s AI Act (AI Act). Our analysis reveals that best practices are often implemented together but ultimately reinforce the pre-existing view that ABMS frequently lacks adequate model documentation. This deficiency hinders evaluability, making it difficult to conduct quality assurance prior to application and meaningful evaluation post application. We propose a framework that highlights the importance of different types of model documentation and the attributes they enable, which are valuable to both modelers and policy actors, albeit for different reasons. The AI Act provides a valuable opportunity to improve model documentation. By proactively developing and establishing guidelines, we can stay ahead of emerging legal requirements.
