ASTIN: Explainable AI for Claims Reserving

The actuarial perspective will cover the practical limitations of current reserving workflows, where judgment enters traditional methods and why consistency is difficult, and what actuaries need from AI, including transparency, validation, and governance. It will also include a case study comparing traditional results with AI supported modeling and how ML results can serve as a second opinion framework.

The machine learning perspective will cover the Bayesian ML framework, including its architecture, model selection based on predictive power and cross validation, and how explainability is built into the model. It will also discuss why full distributions matter more than point estimates and share technical lessons from applying ML to insurance triangles.

Speaker: Yulia Nechay

Yulia Nechay is a fully qualified actuary with over 16 years of experience in non life insurance, covering reserving, pricing, reinsurance, and portfolio risk management. She is co-founder of PredicTri, where her work focuses on actuarial methodology, model validation, and explainability in AI based reserving frameworks. She is actively involved in international actuarial discussions on the adoption of AI and leads the professional Linkedin community Next Gen Triangle Modeling with AI/ML, dedicated to advancing modern reserving techniques.

Speaker: Ben Zickel

Ben Zickel is a physicist and algorithm developer with more than 25 years of experience in signal processing, machine learning, stochastic modeling, and high performance computing. He is co-founder of PredicTri, where he leads the development of Bayesian modeling frameworks, including probabilistic model design, predictive validation, and robust inference methods. His work focuses on translating advanced statistical and computational techniques into practical tools for actuarial decision making.

[....]