인공지능 행정이 행정절차･행정소송에 미치는 영향에 대한 시론적 고찰

In the situation where the use of artificial intelligence is expected to expand in the administrative field, how to secure the transparency and accountability of the artificial intelligence algorithm is important because of the unpredictability and autonomy of the artificial intelligence algorithm. To this end, administrative law responses should be considered from the standpoint of rule of law and due process, apart from ethical responses to the use of AI or special laws for AI regulation. In the United States, various discussions have been conducted regarding AI administration in recent years, focusing on the selection of execution targets by AI and the decision related to social security benefits. With a strong emphasis on the advantages of artificial intelligence administration, and actively adopting it, the black box nature of artificial intelligence is also acceptable in light of the standards of transparency or declaration of reasons in the administrative procedure law and the court's deference in administrative litigation. It is said that the application of administrative procedures(e.g. hearing) is not required, and rules can be established through artificial intelligence. On the other hand, in terms of emphasizing the response to the black box nature of automation administration and artificial intelligence, the information about the AI ​​algorithm should be disclosed as much as possible in relation to transparency and administrative procedures should be applied regarding AI system problems. In addition, in order to strengthen the court's control over AI, It is discussed that the court should review the selection of administrative execution targets that were previously excluded from judicial review, or limit the court's deference for AI administration. In order to secure the transparency and accountability of AI administration, the following matters need to be discussed in Korean administrative procedures and administrative litigation. First, it is possible to consider supplementing with a separate review procedure to enhance the objectivity and fairness of the selection of investigation targets by using artificial intelligence. Second, with regard to complaints from the Social Security Administration, the AI system for handling complaints must be designed to reflect the disposition standards and forms. In this regard, establishing the enhanced disposal standards can be an effective way to compensate for the opacity of artificial intelligence. Third, with regard to the administrative litigation law response to AI administration, to minimize the ripple effect caused by AI system problems, it allows judicial review of AI algorithms even before specific administrative actions are performed by AI. And in order to ensure transparency in AI administration, measures to strengthen the criteria for declaration of reasons should be considered. Finally, if the introduction of artificial intelligence increases administrative efficiency and decreases the cost of administrative procedures, this can be an incentive to reduce the exceptions of administrative procedure. By minimizing the reason for administrative procedure exceptions, focusing on secrecy, we will be able to move toward expanding the scope of administrative procedures in parallel with the expansion of AI administration.