Designing Explainable AI: The Case of Dashboard Design for Fraud Detection in Public Transport Ticketing Systems

Burger, Mara; Näscher, Hans-Henning; Kipping, Gregor; Gau, Michael; vom Brocke, Jan

Abstract

Fraud detection in digital ticketing systems presents a significant challenge for public transport operators, as its implementation requires considerable financial and operational investment. In Germany’s largest ticketing system, approximately 7% of transactions involve fraudulent or unpaid tickets, causing substantial monetary losses. Moreover, existing artificial intelligence (AI)-based fraud detection solutions lack transparency and trust due to their black-box nature. Applying a design science research (DSR) approach and collaborating with a leading German public transportation operator, this study extends existing design knowledge by an instantiation and evaluation of an explainable AI (XAI)-based fraud detection dashboard, which was trained on 1.7 million transactions collected over two years. The evaluated system demonstrates high accuracy and precision on test data. Expert evaluations reveal that the system increases trust and transparency while maintaining necessary human oversight. Our findings advance the understanding of XAI in real-world settings and illustrate how design principles can be instantiated and evaluated in practice.