MIMICO project: From Surrogate Modeling to Explainable AI in Agent-Based Simulations

Tóm tắt báo cáo

Complex agent-based models (ABMs) are essential for understanding emergent phenomena, yet their high computational cost and "black-box" nature often hinder their utility. As part of the MIMICO project, this talk explores the integration of Surrogate Modeling and Explainable AI (XAI) to address these challenges. We will discuss how meta-models can be used to approximate heavy simulations, providing a foundation for interpretability. By applying XAI techniques to these surrogates, we can move beyond simple observation to identify the key parameters and interactions that drive model behavior. This approach ensures that complex simulations are not only faster but also more transparent and trustworthy for decision-making.