AI-Assisted Academic Writing as an Inquisitorial Appellate Court

Abstract

The rapid proliferation of generative AI tools such as ChatGPT, Gemini, and Claude has ignited a fierce debate in academia over the legitimacy of AI-assisted writing. Critics equate its use with cheating, citing risks of hallucinations, plagiarism, and erosion of scholarly standards. Proponents counter that AI can enhance productivity and free researchers to focus on substantive intellectual contributions. This paper argues that AI-assisted academic writing, when practiced responsibly and under proper constraints, is not only defensible but, under the right institutional constraints, can enhance the efficiency of scholarly production without compromising its quality. I first develop ten arguments in defense of AI-assisted writing, addressing objections ranging from its moral status and deskilling concerns to questions of attribution, disclosure, and cognitive biases. I then propose a framework drawing on an analogy to inquisitorial appellate courts, in which the academic author functions as an appellate judge who receives an initial output and must independently investigate, verify, and refine it before endorsing it as their own. Just as appellate systems rely on institutional safeguards-appointment criteria, rules of procedure, multi-tier review, and misconduct sanctions-to ensure legitimacy, the academic system can develop analogous mechanisms to govern responsible AI use. Drawing on the law and economics literature on judicial incentives, I develop a formal framework for aligning academic incentives with responsible AI deployment. The paper synthesizes these insights into a set of proposed best practices and concludes by discussing the limitations of the appellate court analogy.

Keywords: Generative AI, Academic Writing, Appellate Courts, Law and Economics, Institutional Design

JEL Classification: D02, D91, K40, O33, I23