GenAI Did Not Break Academia, It Revealed It: A Proposal for Structural Reform and Responsible AI Integration

Team name: Lobby D

Use of AI tools :

• LLMs (Claude and Gemini)

• For research, finding suitable documentation, clean redaction of the contribution

• They helped us get useful research papers suitable for the topic we're addressing, which was important for documentation, inspiration, and argumentation.
  They also helped us refine and formalize our prose. The ideas, arguments, and research were entirely our own; AI helped us express them with the clarity a formal academic paper requires.

Initial contribution: GenAI Did Not Break Academia, It Revealed It: A Proposal for Structural Reform and Responsible AI Integration

Final contribution:
Reflection on the process (important)
Three categories of feedback shaped the transition from our first essay to this final paper.

• The first concerned the technical limitations of AI watermarking. Several comments pointed out that watermarking techniques remain unreliable and easily circumvented, and that building policy proposals around them was premature. We took this seriously, researched the topic further, and ultimately made the decision to drop the entire regulatory and technical framework from our contribution. This was a significant cut, but it made the final paper more honest and more focused.

• The second concerned the lack of research-based grounding in our initial essay. Reviewers asked for stronger evidence and academic support for our claims. In response, we undertook a thorough documentation phase before writing anything new, building a related work section that situates our thinking within existing litterature on critical pedagogy, assessment reform, and ML-driven feedback systems.
  We also conducted an original student survey distributed across our networks and social media, to ground our theoretical hypotheses and bring first-hand empirical data into the argument. We present this survey as a first initiative and call for a larger-scale study spanning multiple institutions and countries that could produce the representative data needed to inform real policy reform.

• The third concerned the vagueness and feasibility of our propositions. Early feedback challenged us to move beyond diagnosis and present something concrete and implementable. Our final paper responds to this directly: the oral assessment proposition now includes a detailed logistical model grounded in existing lab (TD) session infrastructure, with specific calculations for group sizes, time slots, and staffing; and the ML feedback proposition is anchored in a documented body of current research on NLP and institutional feedback analytics.

Together, these three rounds of challenge pushed us from a philosophical essay into a structured academic paper with empirical grounding, targeted propositions, and an honest acknowledgment of what remains to be tested.

Our contribution gives two focused, actionable propositions: (1) a reform of academic assessment toward oral and live evaluation formats that cannot be delegated to AI, and (2) the integration of machine learning into institutional feedback infrastructure to give schools the tools to identify and respond to student disengagement in real time.

Please find our full contribution in the following link: https://drive.google.com/file/d/1KFkBnQ2ajwjf37d3G6QrXqdVbODIgsVR/view

Video Pitch