A Control-Based Model of Authorship in AI-Mediated Research

Abstract The rapid integration of artificial intelligence (AI) into research and writing workflows has generated a crisis of authorship attribution. Existing frameworks often rely on binary distinctions human versus AI generated failing to capture the nuanced, iterative interactions that characterize real world usage. This paper proposes a control based model of authorship, arguing that authorship is not determined by initiation or mechanical generation, but by sustained conceptual control, evaluative judgment, and iterative refinement. By distinguishing between delegation, retrieval, and directed synthesis, this framework clarifies when AI functions as a tool versus when it meaningfully shapes intellectual output. The model may provide a defensible basis for human authorship in AI mediated research while acknowledging the non-trivial role of generative systems.摘要 人工智能（AI）快速融入科研与写作工作流，引发了作者归属认定危机。现有框架往往依赖人类生成与AI生成的二元区分，无法捕捉现实应用场景中特有的细腻且迭代式交互特征。本文提出一种基于管控的作者权模型，主张作者权并非由初始创作或机械生成所决定，而是由持续的概念管控、评价性判断以及迭代式打磨所界定。通过区分委托、检索与定向合成，该框架阐明了AI何时作为工具发挥作用，又何时会实质性地塑造智力产出。该模型可为AI辅助研究中的人类作者权提供可辩护的依据，同时承认生成式系统所发挥的不可忽视的作用。