Artificial intelligence is shaking higher education, but it is not the cause, according to Jeff Sparrow, researcher‑teacher at the University of Melbourne. In a essay published on 22 October 2026 by The Conversation, he estimates that the system was already weakened, even before these tools arrived.
- L’IA ne cause pas la crise de l’enseignement supérieur, mais révèle un système déjà affaibli par une logique managériale.
- L’évaluation doit être repensée, car les productions que la machine sait générer ne mesurent plus la compétence réelle.
- L’engagement des étudiants dépend de la qualité des expériences d’apprentissage conçues par les enseignants, non de la simple mise à disposition d’outils.
- Les universités doivent coordonner gouvernance et pédagogie, au lieu d’acheter des licences et de signer des partenariats sans changer les pratiques.
A revelation rather than a cause
The observation by Jeff Sparrow is unflinching. According to him, AI devastates higher education, but decades of managerial logic have already turned degrees into products, students into clients, and research into metrics. In this context, using ChatGPT to write an assignment is not a deviance, but a rational response to an institution that has itself stopped valuing learning. The author also recalls university officials who urged students not to delegate their thinking, in op‑eds themselves largely written by an AI. “The symbol is cruel, it is also revealing.”
The author of a reflection that extends this diagnosis, drawing on his recent research conducted with students, teachers and management teams, reaches a similar conclusion. AI acts, in his view, as a solvent that reveals the fragilities of a model based on content transmission, standardized production assessment and massification of pathways. When a machine produces an acceptable essay in a few seconds, it is the pedagogical value of the exercise itself that is called into question, far beyond mere cheating.
Four inseparable transformations
For this reflection, AI integration must be thought of holistically and systemically, through four transformations that evolve together. The first is technological and concerns the environment in which learners operate, made up of conversational assistants, adaptive systems and immersive spaces. The second is pedagogical: it involves moving from a transmission logic to a logic of designing learning experiences. The third touches educational practices, and the role of the teacher, who becomes an architect of learning situations, supported by new professions such as learning‑experience designer. The fourth, deemed the most decisive, concerns assessment: as long as we grade productions that the machine knows how to generate, we will assess “the ability to craft a prompt rather than the real competence”, i.e., the formulation of queries addressed to an AI assistant rather than the effective mastery of knowledge.
None of these dimensions can be transformed in isolation. Introducing AI without rethinking pedagogy yields passive consumption, while evolving assessment without supporting teachers exhausts those driving the change. Yet most institutions do not know where to start. This approach demands governance capable of linking strategic, tactical and operational levels, and of coordinating initiatives today scattered among digital departments, teaching services and faculties.
This reflection also denounces an erroneous approach. Universities purchase licences, sign partnerships with major publishers, draft usage charters, then are surprised that nothing changes in the classroom. The University of Sydney, which signs a multi‑year agreement with OpenAI a few days after a strike by its staff, illustrates this disconnect between stated strategy and lived reality.
Learning experience as the key to engagement
The cited works illuminate a central point: it is not AI per se that enhances student engagement, but the quality of the learning experience in which it is embedded. Where the setup is carefully designed, prepared by the teacher and integrated into a reflective approach, cognitive engagement advances. Where the tool is merely made available, it fuels a “cognitive debt”, documented by neuroscience, consisting of dependence and superficial learning.
Jeff Sparrow notes that the teachers least affected by ChatGPT are those who run small groups passionate about their discipline. Students, according to this reading, do not delegate what engages them, but what seems meaningless. The issue would therefore be less about fighting the machine than about rebuilding learning experiences rich enough for intellectual effort to regain its value.
