THIS IS THE DEV/TESTING WEBSITE IPv4: 18.222.97.243 IPv6: || Country by IP: GB
Close Search Advanced
Journals
Resources
About Us
Open Access
Journals
Resources
About Us
Open Access

Erklärbare Künstliche Intelligenz am Beispiel von Ratings deutscher Lebensversicherungsunternehmen

Erklärbare Künstliche Intelligenz am Beispiel von Ratings deutscher Lebensversicherungsunternehmen
Read Now
Download (PDF)

Year:    2023

Author:    Bartel, Holger, Kraft, Mirko, Leidner, Jochen L.

Zeitschrift für die gesamte Versicherungswissenschaft, Vol. 112 (2023), Iss. 1 : pp. 3–32

Abstract

Artificial intelligence (AI) is already used for decision-making in practice (Lossos/ Geschwill/Morelli 2021), increasingly also in the insurance sector. However, it requires trust in the various AI methods, especially in the evaluation of companies („ratings“). Trust is formed when decision makers and users can form mental models of a system and they understand its output. AI must therefore be explainable; a pure black box is insufficient even if a system is of high quality. „Explainable AI“ (eXplainable Artificial Intelligence, XAI) is concerned with the development of AI models that are comprehensible by humans (Adadi/Berrada 2018; European Commission 2020). In this paper, desirable properties of industrial AI systems are investigated – specifically with respect to explainability – and presented and visualized using the application example of ratings of German life insurance companies. In addition to XAI as one prerequisite for technical acceptance, the interaction between the business model and customer acceptance of ratings of German life insurance companies is examined. Financial key performance indicators for German life insurance companies are often said to lack transparency; this is still the case when HGB accounting is supplemented by the Solvency and Financial Condition Reports (SFCR) according to Solvency II. We argue that the examination of explainable AI methods is a useful contribution to the practice of valuation.

Journal Article Details

Publisher Name:    Global Science Press

Language:    German

DOI:    https://doi.org/10.3790/zverswiss.2023.04.Bartel.etal

Zeitschrift für die gesamte Versicherungswissenschaft, Vol. 112 (2023), Iss. 1 : pp. 3–32

Published online:    2023-02

AMS Subject Headings:    Duncker & Humblot

Copyright:    COPYRIGHT: © Global Science Press

Pages:    30

Keywords:    explainable artificial intelligence (XAI) ratings life insurance companies

Author Details

Bartel, Holger

Kraft, Mirko

Leidner, Jochen L.

  1. Adadi, A./Berrada, M. (2018): Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI), IEEE Access 6, S. 52138–52160, DOI: 10.1109/ACCESS.2018.2870052.  Google Scholar
  2. Bartel, H. (2014a): Simple Solvency – Ein Solvenzmodell für deutsche Lebensversicherer, DOI: 10.13140/2.1.2939.9041, https://www.researchgate.net/publication/267337608_Simple_Solvency_-_Ein_Solvenzmodell_fur_deutsche_Lebensversicherer [26.01.2023].  Google Scholar
  3. Bartel, H. (2014b): Simple Solvency – Ein Solvenzmodell für deutsche Lebensversicherer, Vortrag, qx Club, Regionale Gruppe der Deutschen Aktuarvereinigung e.V. (DAV) für Berlin, DOI: 10.13140/2.1.1760.2560, https://www.researchgate.net/publication/267337670_Simple_Solvency_-_Ein_Solvenzmodell_fur_deutsche_Lebensversicherer [26.01.2023].  Google Scholar
  4. Bartel, H. (2019): Kausale Analyse von Gleichungssystemen mit strukturellen neuronalen Netzen. Technischer Bericht. DOI: 10.13140/RG.2.2.16841.26729, https://www.researchgate.net/publication/335099531_Kausale_Analyse_von_Gleichungssystemen_mit_struk turellen_neuronalen_Netzen [26.01.2023].  Google Scholar
  5. Bartel, H. (2020a): Causal Analysis – With an Application to Insurance Ratings, https://www.researchgate.net/publication/339091133_Causal_Analysis_-_With_an_Application_to_Insurance_Ratings [26.01.2023].  Google Scholar
  6. Bartel, H. (2020b): Explainable Artificial Intelligence (XAI) in Ratings, https://www.researchgate.net/publication/344992217_Explainable_Artificial_Intelligence_XAI_in_Ratings [26.01.2023].  Google Scholar
  7. Bartel, H. (2020c): Causing: Causal Interpretation using Graphs, https://www.researchgate.net/publication/341878489_Causing_CAUSal_INterpretation_using_Graphs [26.01.2023].  Google Scholar
  8. Bartel, H. (2020d): RealRate Expert System Life Insurance, https://realrate.ai/download/publications/RealRate%20Expert%20System%20Life%20Insurance.pdf [26.01.2023]  Google Scholar
  9. Bartel, H. (2023): Finanzstärke-Ratings deutscher Versicherer mittels künstlicher Intelligenz. In: Zeitschrift für Versicherungswesen, 74(2), S. 42–51.  Google Scholar
  10. Bartlett, R./Morse, A./Stanton, R./Wallace, N. (2022): Consumer-lending discrimination in the FinTech Era. In: Journal of Financial Economics 143 (1), S. 30–56, DOI: 10.1016/j.jfineco.2021.05.047.  Google Scholar
  11. Breiman, L./Friedman, J. H./Olshen, R. A./Stone, C. J. (1984): Classification and regression trees, Monterey, CA, USA: Wadsworth & Brooks/Cole, DOI: 10.1201/9781315139470.  Google Scholar
  12. Burkart, N./Huber, M. F. (2021): A Survey on the Explainability of Supervised Machine Learning, Journal of Artificial Intelligence Research 70, S. 245–317, DOI: 10.1613/jair.1.12228.  Google Scholar
  13. CFO-Forum (2009): Market Consistent Embedded Value (MCEV)-Principles, http://www.cfoforum.eu/downloads/MCEV_Principles_and_Guidance_October_2009.pdf [26.01.2023].  Google Scholar
  14. Cohen, W. W. (1995): Fast Effective Rule Induction, Proc. 12th Int. Conf. Machine Learning (ICML).  Google Scholar
  15. Crumley, D. G. (2012): Credit Rating Agencies and Conflicts of Interest, Inauguraldissertation (J.D. thesis), University of Texas at Austin, Austin, TX, USA.  Google Scholar
  16. Dierkes, S./Sümpelmann, J. (2019): Digitalisierte Peer-Group-Bestimmung und Beta-Anpassung. In: Ballwieser, W./Hachmeister, D. (Hg.) (2019): Digitalisierung und Unternehmensbewertung, S. 173–192. Stuttgart: Schäffer-Poeschel.  Google Scholar
  17. Dziugaite, G. K./Ben-David, S./Roy, D. M. (2020): Enforcing interpretability and its statistical impacts: trade-offs between accuracy and interpretability, https://arxiv.org/pdf/2010.13764.pdf [26.01.2023].  Google Scholar
  18. Europäische Kommission (2020): Weißbuch zur Künstlichen Intelligenz – Ein europäisches Konzept für Exzellenz und Vertrauen, Brüssel, https://ec.europa.eu/info/sites/info/files/commission-white-paper-artificial-intelligence-feb2020_de.pdf [26.01.2023].  Google Scholar
  19. Europäische Kommission (2021): Regulating credit rating agencies, Brüssel, https://ec.eu ropa.eu/info/business-economy-euro/banking-and-finance/financial-supervision-and-risk-management/managing-risks-banks-and-financial-institutions/regulating-credit-rating-agencies_en [26.01.2023].  Google Scholar
  20. Hochrangige Expertengruppe für Künstliche Intelligenz (HEG-KI) (2018): Ethik-Leitlinien für eine vertrauenswürdige KI, Brüssel, https://op.europa.eu/s/oVfc [26.01.2023].  Google Scholar
  21. GitHub (2021a): A Real World Example: Education and Wages for Young Workers, https://github.com/realrate/Causing/blob/develop/docs/education.md [26.01.2023].  Google Scholar
  22. GitHub (2021b): Causing: CAUSal INterpretation using Graphs, https://github.com/realrate/Causing.  Google Scholar
  23. Gründl, H./Kraft, M. (Hg.) (2019): Solvency II – Eine Einführung. Grundlagen der neuen Versicherungsaufsicht. 3. Aufl. Karlsruhe: VVW.  Google Scholar
  24. Heskes, T./Sijben, E./Bucur, I. G./Claassen, T. (2020): Causal Shapley Values: Exploiting Causal Knowledge to Explain Individual Predictions of Complex Models, NeurIPS 2020, https://arxiv.org/abs/2011.01625 [26.01.2023].  Google Scholar
  25. Holland, C. P./Kavuri, A. (2021): Artificial intelligence and digital transformation of insurance markets. In: The Capco Institute – Journal of Financial Transformation H. 54 (11/2021), S. 104–115, https://www.capco.com/-/media/CapcoMedia/Capco-2/PDFs/Capco-Journal-54AI-and-Digital-Transformation-of-Insurance-Markets.ashx [26.01.2023].  Google Scholar
  26. James, G./Witten, D./Hastie, T./Tibshirani, R. (2017): An Introduction to Statistical Learning: with Applications in R, New York, NY, USA: Springer.  Google Scholar
  27. Kingma, D. P./Ba, J. (2014): Adam: A Method for Stochastic Optimization, ICLR 2015 conference paper, https://arxiv.org/abs/1412.6980 [26.01.2023].  Google Scholar
  28. Kokina, J./Davenport, T. H. (2017): The emergence of artificial intelligence: How automation is changing auditing. Journal of Emerging Technologies in Accounting 14(1), S. 115–122.  Google Scholar
  29. Kurmann, S. (2023): KI in der Versicherungsbranche: Wenn Science-Fiction auf Realität trifft, https://www.handelszeitung.ch/insurance/kunstliche-intelligenz-fur-versicherungen/ki-in-der-versicherungsbranche-wenn-science-fiction-auf-realitat-trifft-564992 [01.02.2023].  Google Scholar
  30. Leidner, J. L. (in Vorbereitung): A Survey of Ethical Problems of Artificial Intelligence.  Google Scholar
  31. Lossos, C./Geschwill, S./Morelli, F. (2021): Offenheit durch XAI bei ML-unterstützten Entscheidungen: Ein Baustein zur Optimierung von Entscheidungen im Unternehmen? HMD Praxis der Wirtschaftsinformatik 58, S. 303–320.  Google Scholar
  32. Lundberg, S. (2020): Vortrag „The Science behind SHAP“, https://www.youtube.com/watch?v=-taOhqkiuIo [26.01.2023].  Google Scholar
  33. Lundberg, S. M./Lee, S.-I. (2017): A unified approach to interpreting model predictions, Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS). Red Hook, NY, USA, S. 4768–4777.  Google Scholar
  34. Oletzky, T./Reinhardt, A. (2022): Herausforderungen der Regulierung von und der Aufsicht über den Einsatz Künstlicher Intelligenz in der Versicherungswirtschaft. In: Zeitschrift für die gesamte Versicherungswissenschaft 111 (4), S. 495–513. DOI: 10.1007/s12297-022-00541-4.  Google Scholar
  35. Owens, E./Sheehan, B./Mullins, M./Cunneen, M./Ressel, J./Castignani, G. (2022): Explainable Artificial Intelligence (XAI) in Insurance. In: Risks 10 (230), S. 1–50, DOI: 10.3390/risks10120230.  Google Scholar
  36. Quinlan, J. R. (1986): Induction of Decision Trees, Machine Learning 1(1): S. 81–106.  Google Scholar
  37. Rall, L. B. (1981): Automatic Differentiation: Techniques and Applications. Lecture Notes in Computer Science. 120. Springer.  Google Scholar
  38. Ribeiro, M. T./Singh, S./Guestrin, C. (2016): „Why Should I Trust You?“: Explaining the Predictions of Any Classifier, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), San Francisco, California, USA, S. 1135–1144, DOI: 10.1145/2939672.2939778.  Google Scholar
  39. Rumelhart, D. E./Hinton, G. E./Williams, R. J. (1986): Learning representations by back-propagating errors, Nature 323 (6088): S. 533–536, DOI 10.1038/323533a0.  Google Scholar
  40. Samek, W./Müller, K.-R. (2019): Towards Explainable Artificial Intelligence. In: Samek, W./Montavon, G./Vedaldi, A. (Hg.): Explainable AI. Interpreting, explaining and visualizing deep learning (Lecture Notes in Computer Science 11700 /Lecture Notes in Artificial Intelligence), S. 5–22, DOI: 10.1007/978-3-030-28954-6.  Google Scholar
  41. Sellhorn, T. (2020): Machine Learning und empirische Rechnungslegungsforschung: Einige Erkenntnisse und offene Fragen, Schmalenbachs Z. betriebswirtsch. Forsch. 72, S. 49–69, DOI: 10.1007/s41471-020-00086-1.  Google Scholar
  42. Shapley, L. S. (1951): Notes on the n-Person Game – II: The Value of an n-Person Game, Technical Report RM-670, Santa Monica, CA, USA: RAND Corporation.  Google Scholar
  43. Simpson, E. H. (1951). The Interpretation of Interaction in Contingency Tables, Journal of the Royal Statistical Society, Series B. 13: S. 238–241.  Google Scholar
  44. Stuwe, A./Weiß, M./Philipper, J. (2012): Ratingagenturen: Sind sie notwendig, überflüssig, notwendiges Übel oder schädlich?, Bonn: Friedrich-Ebert-Stiftung, https://library.fes.de/pdf-files/managerkreis/09647.pdf [26.01.2023].  Google Scholar
  45. Van Hulle, K. (2019): Solvency requirements for EU insurers. Solvency II is good for you. Cambridge: Intersentia.  Google Scholar
  46. Wermter, S./Sun, R. (Hg.) (2000): Hybrid Neural Systems. Springer-Verlag, Heidelberg.  Google Scholar

Section Title Page Action Price
Holger Bartel / Mirko Kraft / Jochen L. Leidner: Erklärbare Künstliche Intelligenz am Beispiel von Ratings deutscher Lebensversicherungsunternehmen 3
Zusammenfassung 3
Abstract 4
1. Einleitung 4
1.1 Motivation 4
1.2 Überblick 6
2. Erklärbare Künstliche Intelligenz 6
2.1 Ansätze der Erklärbarkeit 6
2.2 XAI und Ratings 10
3. Use Case: Unternehmensratings 13
3.1 Ein hybrides Ratingmodell mit Expertenwissen 13
3.2 Der erklärbare kausale Graph 18
3.3 Schätzung, Modellierungszyklus und Software 24
4. Transparenz und Geschäftsmodell 26
5. Zusammenfassung 28
Literatur 29