Reconstructing the Norms of Authorship in the Era of Generative AI

Recently, academic journals have been updating their policies toward either not recognizing AI as an author or requiring disclosure of AI use. However, policy wording alone is insufficient to adequately disentangle the recurring gray areas that arise in real research and educational settings. The purpose of this study is to propose a three-axis framework—accountability, contribution, and transparency—through which generative AI use can be used to reconstruct authorship norms, and to present these axes in the form of formalized rules that can be implemented in practice. Specifically, first, under the accountability axis, the responsible agent in cases of error, fabrication, plagiarism, or bias is explicitly defined as a human, and the level of verification obligation is differentiated according to task difficulty and risk. Second, under the contribution axis, tool use is separated from intellectual contribution, and minimum requirements for human contributions that satisfy authorship criteria are specified even when AI assistance is possible. Third, under the transparency axis, the location, scope, and level of reproducibility of disclosures are tiered so as to create a workable interface between policy and practice.