Towards Personalized Explanations for AI Systems: Designing a Role Model for Explainable AI in Auditing

Jonas Rebstadt, Florian Remark, Philipp Fukas, Pascal Meier, Oliver Thomas

In: Wirtschaftsinformatik Proceedings (2022). Internationale Tagung Wirtschaftsinformatik (WI-2022) February 21-23 Erlangen-Nürnberg Germany Springer 2022.


Due to a continuously growing repertoire of available methods and applications, Artificial Intelligence (AI) is becoming an innovation driver for most industries. In the auditing domain, initial approaches of AI have already been discussed in scientific discourse, but practical application is still lagging behind. Caused by a highly regulated environment, the explainability of AI is of particular relevance. Using semi-structured expert interviews, we identified stakeholder specific requirements regarding explainable AI (XAI) in auditing. To address the needs of all involved stakeholders a theoretical role model for AI systems has been designed based on a systematic literature review. The role model has been instantiated and evaluated in the domain of financial statement auditing using focus groups of domain experts. The resulting model offers a foundation for the development of AI systems with personalized explanations and an optimized usage of existing XAI methods.

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz