Your student used ChatGPT in a writing assessment. What if they deserve an A+? – A student’s perspective

by Byungjoon (Eric) Yoo, University of Oxford

“I have made a living writing for the last thirty years. I know an AI writing when I see one.”

A humanities professor proudly announced this to his classroom during introduction week, a few years ago, while explaining the final term paper. Erin, an undergraduate, listened with mixed feelings: the previous semester, the same professor had praised her final term paper—heavily reliant on ChatGPT—as “one of the best pieces of writing I have read as an undergraduate advisor.” He proceeded on to recommend her to join the postgraduate program. During my interview with Erin, she further shared her complex emotions: on the one hand she realizes that she may have the upper hand in the “cat and mouse game” between the detectors and the detected, an endless masquerade of pretending to have used less AI than they originally had. On the other hand, however, she started questioning whether attempting to discern the “true author” would be a functional strategy at all, especially in this “brave new world,” molded by rapid technological development, we came to live within.

Erin’s story is one of myriad cases of confusion within higher education facing the rapid diffusion and uptake of Gen AI. As her professor demonstrated, some have responded to this turbulence with a proliferating repertoire of assessment reforms and numerous guardrails including AI-use declarations, renewed honor codes, detection software, and redesigned rubrics (Jin et al. 2025; Miao & Holmes, 2023). One common thread of intention that weaves these efforts could be summarized as: to calculate the individual’s authentic contribution to a piece of writing or any produced work. Defending the integrity of such calculation against the external threats exemplified by Gen AI is emphasized as a key educational goal.

This intent to ‘enclosure’ individual share within an academic work, is entangled with, or a manifestation of, the core of modern education’s central ideology: meritocracy. As a political proposition that scarce social resources should be distributed according to individual merit, this belief has long been translated into an educational worldview (Markovitz, 2019; Sandel, 2020). Here, cultivating each individual’s distinctive ability becomes the ultimate aim of education, and assessing that individual ability—with minimal error and without prejudice—becomes the ultimate criterion by which its legitimacy becomes substantiated.

The rather fruitless self-confidence of Erin’s professor and other similar institutional efforts attempting to isolate the individual’s authentic share in an AI-mediated outcomes, then is a symptom of this meritocratic inertia that still governs our higher education. Considering that academic writing has long functioned as the core­, if not only, educational ritual through which individual ‘merit’ was repeatedly ratified as a valid unit of account, it then would be no surprise that the most heated debate on student AI use in higher education gravitates around it. It indeed was a golden standard of academic merit, in an era where one could only write or type “word by word, sentence by sentence” by oneself, through their own bodily and cognitive labor. In other words, this ritual had a strong material basis for its claims; yet, in this new era wherein one could ‘produce writing’ without ‘writing,’ the taken-for-granted connection between one’s product and one’s merit seems to have been severely challenged.  Alas, we, including Erin and her professor, find ourselves increasingly struggling to acknowledge an academic writing as a proper reflection of one’s own merit, a pure working of one’s inner mind.

Of course, this does not signify the instant end of academic writing as a medium for assessing merit. Largely, two families of response seem to have emerged: sanitation and demarcation. Sanitation attempts to create a niche spatial-temporal condition in which students are completely separated from AI. Practices such as AI-free writing under supervision, handwritten exams, oral examinations are primary examples. The second, demarcation attempts to disentangle the complex final product by distinguishing the student’s share from the AI’s share via disclosure protocols, process documentation, or detection technologies. Both share an intention to restore the integrity of the link between writer and the text, thus preserving writing as a valid merit-assessing mechanism.

Yet, these attempts are facing limits that appear constitutive rather than transitional. The first struggles to justify itself on social grounds: students will leave the classroom into lives of continuous interaction with AI, and it is difficult to articulate, least of all to the students themselves, why they should be deliberately removed into sanitary lab-condition to be evaluated. If they would be subsequently evaluated by their AI-entangled reports upon graduation, what makes higher education so special to be exempt from these entangled times? The second confronts a more fundamental difficulty. Writing with AI is increasingly not an arithmetic mixture of human and machine contributions but something closer to a chemical compound, a product of genuine synthesis in which the constituent shares no longer exist as separable quantities. Soon enough, it may be virtually impossible to segmentize and separate my contribution from to that of AI.

Despite these hardships, efforts to identify and measure an individual’s distinctive ability seem unlikely to disappear anytime soon. Yet the material conditions transformed by Gen AI will continue to frustrate the ideal on which such efforts depend, rendering attempts to isolate and attribute individual merit continuously partial and unstable. Perhaps, then, what we are witnessing is not simply a temporary crisis of assessment, but the declining historical viability of writing as an educational ritual—the medium we have long considered as a transparent reflection of one’s own academic merit. If this speculation has some truth inside it, then our higher education systems and institutions would face severe pressure to reinvent or redefine its core academic rituals. We may search for series of new academic rituals (or revive the old ones) capable of sustaining the idea of individually attributable merit. We may also, as the global trend demonstrates, strive to reinforce and perfect existing strategies of sanitation and demarcation. For the time being, we are witnessing a race to define and implement the new lines of defense.

We must, however, while configuring such defensive strategies, not step away from critically responding to speculative inquiries this novel technology enabled for higher education (Macgilchrist, 2021). For instance, we might begin to imagine a world of post-individual merit, wherein we no longer can consider individual as the basic unit-of-analysis. Here, ‘merit’ exists only in an entangled state, as “networked merit” (Yoo & Lee, 2026), refusing to be ascribed to an individual. Unlike sanitization and demarcation, the strategies to forcefully detach human and AI, networked merit invites us to acknowledge the complicated relationship between students, professors, and AI as they are: as co-producers equally and holistically responsible for their outcome. When we accept this invitation—admitting that determining whether Erin or her AI agent is a true author is near impossible_—current rituals of individualized merit would also have to be reconfigured to correspond to this new material reality.

Either way, one lesson becomes clear: concurrent educational ‘problems’ posed by Gen AI should not be reduced to the fruitless dichotomy between prohibition and permission in academic writing., the “cat and mouse game” between paranoid faculties and self-doubting students. Instead, we should move on to asking more consequential questions. What transformations this technological change may further incur in higher education, and how could we continue to interact ‘educationally’ there within?

Acknowledgement

This article is a product of continued engagement from my presentation at Networked Learning Conference 2026, which would have not been existent without the academic guidance of Kyungmee and support from PRiDE research group.

References

Download

Author