AI 개입 행정처분의 공법적 함의 — ｢개인정보 보호법｣이 드러낸 자기결정권과 공정력의 개념적 긴장, 그리고 양자의 조화로운 이해를 위한 공정력이론의 재구성 —

초록

The possibility of administrative dispositions involving artificial intelligence (AI) is gradually emerging. Against this background, this article seeks to analyze the structural tension formed between (1) the “right to refuse AI-involved administrative dispositions,” which functions as a safeguard for the protection of the right to self-determination, and (2) the binding effect, which constitutes a traditional legal effect of administrative acts. Furthermore, it aims to explore the possibility of coordinating this tension from the perspective of reconstructing the theory of the binding effect. Article 37-2 of the current Personal Information Protection Act recognizes the data subject’s right to refuse fully automated decisions that have a significant impact on rights or obligations, while explicitly excluding automatic dispositions under Article 20 of the General Act on Public Administration from its scope of application. This appears to be a legislative technique intended to prevent a conflict between the data subject’s right to refuse and the theory of the binding effect. However, it is difficult to conclude that such an approach alone fully resolves the structure of tension between the right-to-refuse regime and the theory of the binding effect. Consider cases in which, at the pre-dispositional stage, AI performs functions such as structuring factual circumstances, selecting issues, and assessing risk or suitability, and the results are absorbed into the final disposition without substantive re-examination. In such cases, even if the automated assessment does not formally constitute an administrative disposition, it performs a normative function by effectively predetermining the direction and content of the subsequent disposition. Within this structure, if the existing theory of the binding effect is applied without modification, there is a risk that the substantive process of judgment formation will remain concealed from external scrutiny, while the effectiveness of ex post remedies will be significantly weakened. Accordingly, this article draws on the analytical framework proposed by Hideyuki Takaki, which decomposes the legal effect of administrative acts into a regulatory mechanism and a blocking mechanism, in order to capture the structure in which regulation is directly transposed into blockage in AI-involved administrative dispositions. To address this problem, the article proposes a stepwise reconstruction of the application structure of the binding effect in AI-involved administrative dispositions, based on whether statutory provisions recognizing a right to refuse exist. Where such provisions exist, the notification of the right to refuse and the provision of an opportunity to exercise it are identified as core requirements, and any omission or disregard thereof should be evaluated as a serious defect that limits the occurrence of the binding effect. In particular, in AI-involved administrative dispositions, the traditional criterion requiring the manifestness of defects is structurally difficult to operate, and therefore the criteria for distinguishing nullity and revocability should also be reorganized around seriousness. Ultimately, the binding effect should be understood not as an effect that is applied unconditionally in an AI-driven administrative environment, but rather as a concept that operates conditionally depending on whether safeguards for the right to self-determination have been observed. This may have significance as a theoretical attempt to coordinate the legitimacy crisis brought about by AI administration while maintaining the traditional function of the binding effect.

키워드