Prof. Kraft and Dr. Bartel present an XAI Approach for Company Ratings
The DVfVW Deutscher Verein für Versicherungswissenschaft holds its annual conference on March 17-18.
In Forum 3 on Insurance Business Topics, Prof. Dr. Mirko Kraft (Coburg University) and Dr. Holger Bartel (CEO RealRate Inc.) present joint work with Prof. Dr. Jochen Leidner (Coburg University and and University of Sheffield) on Explainable Artificial Intelligence (XAI).
We present a real world use case: Creating fair and explainable financial strength ratings of all German life insurers. To achieve this we developed a hybrid AI model, combining highly structured expert input with classical machine learning, enabling us to work even on small data. Most importantly, from the given model equations we can derive a causal structure represented by a neural network. Our work contributes to making Artificial Intelligence more explainable. This „explainability by design“ ensures fairness and interpretability. So there is no black box AI anymore. The study showed some quite interesting results, e. g. that for a company being big does not automatically mean to be also financially strong.
DVfVW Annual Conference 2022
„Explainable Artificial Intelligence (XAI): the Use Case of Ratings for German Life Insurers“
Authors:
Prof. Dr. Mirko Kraft, Coburg University
Dr. Holger Bartel, RealRate Inc.
Prof. Dr. Jochen L. Leidner, Coburg University and University of Sheffield
Host: Prof. Dr. Hato Schmeiser, University St. Gallen
Discussant: Niklas Häusle
The presentation and recorded video are available for members of DVfVW at https://www.dvfvw.de. More research can be found at https://realrate.ai/methodology. The complete ranking list ist to be found here: https://realrate.ai/rankings.
