自動化行政決定之容許性——從行政程序法及個資保護法制的觀察面向

Permissibility of Automated Administrative Decisions: Artificial Intelligence, Administrative Procedure and Personal Data Protection

法學作為規範之學，其任務本就在於對於行為之容許性提出規範性要求。相較於私部門已廣泛運用人工智慧，就公行政中何處運用，以及行政法學如何評價此一人工智慧的運用，似有待進一步釐清。本文擬先就人工智慧的規整面向予以分析，就人工智慧的運用略作分析後，始能立基此概觀進一步討論何以目前各國立法仍多處於倫理準則之故，並簡短論及歐盟的立法模式。其次，就人工智慧在行政各領域運用，並就行政決定自動化在事實面向上的機會與風險加以分析，其次，則是針對法律層面提出何以引進人工智慧的理由、界限與形塑。再者，針對全自動化行政決定所可能引發的質疑與其可能性，本文擬以德國聯邦行政程序法第35a條為例，介紹其立法模式與基本內涵，以供我國後續可能的立法參考；並且進一步思考，在人工智慧運用下，行政決定應負的說理義務如何可能。最後，基於人工智慧之運用多係植基於巨量資料之處理，其中涉有個人資料者，即有個人資料保護法制之適用。而歐盟資料保護法制亦就人工智慧之運用有初步規範，本文亦就其中重要之規範內涵，略加介紹，以完整此一問題觀照面向。

As a normative science, the task of law is to put forward normative requirements for the permissibility of behavior. Compared with the widespread use of artificial intelligence in the private sector, further clarifications of where it is used in public administration and how administrative law evaluates this use of artificial intelligence are needed. This article will first analyze the regulatory aspects of artificial intelligence. After briefly analyzing the application of artificial intelligence, we further explore on the basis of this overview why the current legislation in various countries is still based on ethical standards, and briefly discuss the legislative model of the European Union. Secondly, it analyzes the application of artificial intelligence in various administrative fields and examines the opportunities and risks of the automation of administrative decisions from a factual perspective. This article also puts forward the reasons, boundaries and shaping of the introduction of artificial intelligence from a legal perspective. Furthermore, in view of the doubts and possibilities that may arise from fully automated administrative decisions, this article intends to exemplify Article 35a of the German Federal Administrative Procedure Act and introduce its legislative model and basic connotation, so as to serve as a reference for possible subsequent legislation in our country. Meanwhile, it considers the probability of exerting the reasoning obligation for administrative decisions with the use of artificial intelligence. Finally, as the application of artificial intelligence is mostly based on the processing of huge amounts of data and hence if personal data is involved, personal data protection laws will apply. Given that the EU data protection law has preliminary regulations on the use of artificial intelligence, this article also briefly introduces the important regulatory connotations to complete the perspective of this issue.