About This Event
This presentation introduces the Credibility Transformer, a groundbreaking non-life insurance pricing model developed by Richman, Scognamiglio, and Wüthrich (2025). It combines the traditional Bühlmann-Straub credibility theory with modern attention mechanisms from machine learning. By adapting the architecture behind ChatGPT, the model enables certain coefficients to function as a learned portfolio prior, effectively blending individual risk factors with collective experience—much like actuaries have done for decades, but now powered by neural networks.
