Algorithms as Judges: Constitutional Limits to Automated Decision-Making in Public Administration

Abstract

The increasing use of automated decision-making systems in public administration raises fundamental questions of constitutional law. This paper analyses the legal framework and the constitutional limits of such systems in Germany. The focus is on the provisions of the German Basic Law, in particular Article 20(3) (the administration is bound by law), Article 19(4) (guarantee of legal protection) and the fundamental rights in Articles 1, 2 and 3. Particular attention is paid to Section 35a of the Administrative Procedure Act (VwVfG), which enables fully automated administrative acts. Based on a constitutional dogmatic analysis, taking into account case law, administrative practice and international developments, the study identifies key legal challenges - particularly with regard to transparency, discrimination risks and accountability. The results show that the existing legal framework does not provide sufficient guidelines for the use of complex, learning-based systems, especially where decision-making logic is no longer comprehensible ("black box" problem). Specific reform proposals are being developed to close these regulatory gaps, including binding documentation obligations, human control bodies and a revision of Section 35a VwVfG in line with the planned EU AI Act. The aim is to safeguard the digital transformation in the administration in accordance with the rule of law and to strengthen trust through transparency and legal certainty.