How can we make models more robust, and more rapidly adaptable, to increase the world's resilience? In the face of the dramatic changes provoked by the current pandemic, many machine learning-based systems fail to adapt quickly. The re/insurance industry, with many of its insurance products based on AI models, is particularly impacted. Dynamically changing environments dramatically affect the risk landscape of their portfolios.

This track aims to bring together practitioners from academia, industry (not exclusively re/insurance) and startups, to present and discuss the broad topic of modelling in non-stationary environments, where dynamic modelling is required to cope with an intrinsically changing environment that also reacts to the actions taken by the model. Monitoring and addressing model performance degradation in complex enterprise-level machine learning systems will also be a focus of the track, with an emphasis toward AI safety approaches to guarantee models' fairness, interpretability and robustness.

Sub-topics might include:

• Agent-based modelling
• Hybrid model-based and statistical modelling
• Reinforcement Learning
• AI Safety
• Causal inference