The “Black Box” of Automated Taxation: A Comparative Study of Algorithmic Accountability and Taxpayer Rights in Germany and Taiwan

Chia-Hao Hsu | The Lisbon International and European Tax Law Seminars nº 17
Working Papers
The “Black Box” of Automated Taxation: A Comparative Study of Algorithmic Accountability and Taxpayer Rights in Germany and Taiwan

The modernization of tax administration has reached a pivotal juncture as revenue authorities increasingly replace human discretion with data-driven predictive algorithms and automated risk-scoring frameworks. While these technological shifts enhance administrative efficiency, they introduce an acute institutional friction with the constitutional principles of legal certainty, procedural due process and accountability. This paper examines the black box phenomenon, where the internal logic of algorithmic models becomes opaque, and the resulting crisis of legitimacy in automated fiscal enforcement.

Through a comparative analysis of the law-centric German model, which anchors automation in strict statutory authorization (e.g., §§ 88, 155 AO), and the agile, normative-vacuum approach in Taiwan, this study investigates how tax administrations can harness algorithmic efficacy without dismantling taxpayer safeguards. The research argues that current frameworks, including the German implementation of human-in-the-loop mechanisms, face a crisis of structural hollowization and formalistic oversight. To bridge these regulatory gaps, the paper proposes a normative framework transitioning from a black box to a glass box model. Key ideas include: Requiring revenue authorities to subject AI models to independent evaluations to detect and mitigate systemic data-driven inequalities. Establishing an explicit right for taxpayers to understand the dominant features that triggered an algorithmic risk score in intelligible, non-technical language. Instituting independent technical-legal committees to conduct ex-ante regulatory vetting of tax AI systems prior to deployment.

Finally, the study posits a tectonic shift in tax litigation paradigms, proposing a digital exclusionary rule akin to the police stop-and-frisk path. By shifting the focus of judicial review from the substantive accuracy of the tax bill to the procedural legitimacy of the algorithmic selection process, this framework aims to restore constitutional boundaries on state coercion in the age of sovereign algorithms.

 

Keywords: automated taxation, algorithmic accountability, risk management system, AI smart-audit selection model, digital constitutionalism

Partilhar

Call For Papers Call For Papers
Veja agora

Call For Papers