Abstract

The use of artificial intelligence (AI) in public administration is possible and realistic at all levels of administration and in a wide variety of factual contexts. In view of the deficits associated with AI in terms of its control, legality and transparency, the principles of democracy and the rule of law in particular are used as arguments against its use in public administration. In some cases, being confronted with artificial intelligence integrated into state action is even treated as a case of human dignity. In contrast, the considerable efficiency and quality advantages that the appropriate use of AI can bring are often overlooked. These advantages include, on the one hand, time and cost savings associated with the scalability of AI systems (e. g., in carrying out standardized work steps in formalized contexts, such as checking tax data or subsidy applications, or in opening communication channels between authorities and citizens, such as chat bots), and, on the other hand and above all, their ability to work with higher precision, reliability and quality than human actors (e. g., in preparing forecast-based planning decisions). This is especially true for machine-learning systems, which are typically focused on detecting correlations and can therefore produce results of an entirely different quality. This contribution provides an appropriate legal framework for addressing the opportunities and risks of using AI in administration by interpreting the integration of intelligent computer systems into government action as a kind of “digital administrative assistance” (digitale Verwaltungshilfe). This seeks parallels with the law of privatization, from which the figure of conventional administrative assistance (Verwaltungshilfe) is familiar. Such a parallelization not only promises a more dispassionate, pragmatic attitude toward the use of AI in public administration – “digital administrative assistance” sounds reasonably unsuspicious. Above all, dogmatic routines and legal standards can be imported from the law of privatization, which can lead to a proper treatment of AI in the law of administrative digitization, at least in a modified form. This could render the technology-related discourses more compatible for mainstream administrative law scholarship. Conversely, the suggested parallelization also offers the opportunity to reflect on whether problem-solving strategies regarding the handling of AI systems can be transferred to administrative decision support made by private actors (e. g., as expert evaluators or procurement service providers).