Erklärbare Künstliche Intelligenz am Beispiel von Ratings deutscher Lebensversicherungsunternehmen
JOURNAL ARTICLE
Cite JOURNAL ARTICLE
Style
Format
Erklärbare Künstliche Intelligenz am Beispiel von Ratings deutscher Lebensversicherungsunternehmen
Bartel, Holger | Kraft, Mirko | Leidner, Jochen L.
Zeitschrift für die gesamte Versicherungswissenschaft, Vol. 112 (2023), Iss. 1 : pp. 3–32
Additional Information
Article Details
Author Details
Dr. Holger Bartel, RealRate GmbH
Prof. Dr. Mirko Kraft, Hochschule Coburg, Fakultät Wirtschaftswissenschaften, Friedrich-Streib-Straße 2, 96450 Coburg, Deutschland
Prof. Dr. Jochen L. Leidner, Hochschule Coburg, Fakultät Wirtschaftswissenschaften und University of Sheffield, Department of Computer Science
References
-
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 -
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 -
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 -
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 -
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 -
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 -
Bartel, H. (2020c): Causing: Causal Interpretation using Graphs, https://www.researchgate.net/publication/341878489_Causing_CAUSal_INterpretation_using_Graphs [26.01.2023].
Google Scholar -
Bartel, H. (2020d): RealRate Expert System Life Insurance, https://realrate.ai/download/publications/RealRate%20Expert%20System%20Life%20Insurance.pdf [26.01.2023]
Google Scholar -
Bartel, H. (2023): Finanzstärke-Ratings deutscher Versicherer mittels künstlicher Intelligenz. In: Zeitschrift für Versicherungswesen, 74(2), S. 42–51.
Google Scholar -
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 -
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 -
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 -
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 -
Cohen, W. W. (1995): Fast Effective Rule Induction, Proc. 12th Int. Conf. Machine Learning (ICML).
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
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 -
GitHub (2021b): Causing: CAUSal INterpretation using Graphs, https://github.com/realrate/Causing.
Google Scholar -
Gründl, H./Kraft, M. (Hg.) (2019): Solvency II – Eine Einführung. Grundlagen der neuen Versicherungsaufsicht. 3. Aufl. Karlsruhe: VVW.
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
Leidner, J. L. (in Vorbereitung): A Survey of Ethical Problems of Artificial Intelligence.
Google Scholar -
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 -
Lundberg, S. (2020): Vortrag „The Science behind SHAP“, https://www.youtube.com/watch?v=-taOhqkiuIo [26.01.2023].
Google Scholar -
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 -
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 -
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 -
Quinlan, J. R. (1986): Induction of Decision Trees, Machine Learning 1(1): S. 81–106.
Google Scholar -
Rall, L. B. (1981): Automatic Differentiation: Techniques and Applications. Lecture Notes in Computer Science. 120. Springer.
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
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 -
Van Hulle, K. (2019): Solvency requirements for EU insurers. Solvency II is good for you. Cambridge: Intersentia.
Google Scholar -
Wermter, S./Sun, R. (Hg.) (2000): Hybrid Neural Systems. Springer-Verlag, Heidelberg.
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
Bartel, H. (2020c): Causing: Causal Interpretation using Graphs, https://www.researchgate.net/publication/341878489_Causing_CAUSal_INterpretation_using_Graphs [26.01.2023].
Google Scholar -
Bartel, H. (2020d): RealRate Expert System Life Insurance, https://realrate.ai/download/publications/RealRate%20Expert%20System%20Life%20Insurance.pdf [26.01.2023]
Google Scholar -
Bartel, H. (2023): Finanzstärke-Ratings deutscher Versicherer mittels künstlicher Intelligenz. In: Zeitschrift für Versicherungswesen, 74(2), S. 42–51.
Google Scholar -
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 -
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 -
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 -
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 -
Cohen, W. W. (1995): Fast Effective Rule Induction, Proc. 12th Int. Conf. Machine Learning (ICML).
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
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 -
GitHub (2021b): Causing: CAUSal INterpretation using Graphs, https://github.com/realrate/Causing.
Google Scholar -
Gründl, H./Kraft, M. (Hg.) (2019): Solvency II – Eine Einführung. Grundlagen der neuen Versicherungsaufsicht. 3. Aufl. Karlsruhe: VVW.
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
Leidner, J. L. (in Vorbereitung): A Survey of Ethical Problems of Artificial Intelligence.
Google Scholar -
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 -
Lundberg, S. (2020): Vortrag „The Science behind SHAP“, https://www.youtube.com/watch?v=-taOhqkiuIo [26.01.2023].
Google Scholar -
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 -
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 -
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 -
Quinlan, J. R. (1986): Induction of Decision Trees, Machine Learning 1(1): S. 81–106.
Google Scholar -
Rall, L. B. (1981): Automatic Differentiation: Techniques and Applications. Lecture Notes in Computer Science. 120. Springer.
Google Scholar -
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 -
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 -
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 -
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 -
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 -
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 -
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 -
Van Hulle, K. (2019): Solvency requirements for EU insurers. Solvency II is good for you. Cambridge: Intersentia.
Google Scholar -
Wermter, S./Sun, R. (Hg.) (2000): Hybrid Neural Systems. Springer-Verlag, Heidelberg.
Google Scholar
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.