Beyond Control and Enthusiasm: A Ten-Phase Model for Pedagogical AI Integration in Academic Writing

Abstract: The rapid adoption of generative AI among students and faculty has confronted universities with a pedagogical challenge that neither prohibitionist control strategies nor uncritical enthusiasm adequately address. The present paper puts forward the argument that a deliberate, process-oriented framework is required; one that renders the use of AI visible, discussable and educationally meaningful. The present study draws on the cognitive process tradition of Flower and Hayes (1981) and Bereiter and Scardamalia's (1987) distinction between knowledge-telling and knowledge-transforming writing. In this paper, a foundational conceptual distinction is introduced between AI-supported authorship and AI-displaced authorship. On this basis, a ten-phase model of academic writing is proposed. This model functions as a transparency scaffold rather than a prescriptive sequence, thereby making the recursive subprocesses of academic writing explicit. This enables students to reflect on their use of AI within each phase. Seven competence dimensions—including process awareness, AI-related information literacy, evaluative judgment, didactic prompt design, transparency and documentation, ethical and legal awareness, and cooperative AI use—translate the model's logic into teachable and assessable criteria.