T-14: Academic Writing in the Age of Generative AI

Yes, AI can write your course syllabus for you… but should you let it? This session investigates the perils and pitfalls of relying on popular Generative Artificial Intelligence (GenAI) tools to generate course syllabi, specifically by taking a qualitative look at a variety of AI-generated syllabi and relating it back to course learning objectives. Understanding how Large Language Models work and empirically evaluating their output is key to deciphering which aspects of academic labor and instructional design need to remain human.

This presentation describes our work developing and deploying a generative AI tool to support academic writing at Vanier College and Concordia University. The tool was designed to provide automated, constructive feedback on writing fundamentals such as structure, clarity, and citation format. While the technical aspects met project objectives, integrating the tool into writing centers and classrooms raised challenges. We outline our technical development, practical implementation in classroom settings, and the lessons learned from integrating the tool within established academic environments. This session offers practical insights for educators considering AI support in writing instruction.

The integration of generative AI (GenAI) tools like ChatGPT into academic writing offers both opportunities and challenges. This study investigates linguistic shifts in a dataset of 15,920 research papers (6,574 authors) before and after ChatGPT’s release. We measure changes in lexical diversity, cohesion, readability, and syntactic complexity using TAACO. Post-ChatGPT papers tend to be shorter, more succinct, and show higher lexical diversity. They also exhibit subtle changes in readability and reduced noun/argument-based overlaps, suggesting reorganized cohesive structures. Observed across physical sciences, these findings highlight both the promise and intricacy of AI-driven transformations in scholarly communication.