When universities write policies about generative AI, what are they actually worried about? That’s the question Luo (2024) set out to answer, and the findings are more revealing than the policies themselves probably intended. The study, published in Assessment & Evaluation in Higher Education, analyzed GenAI policy documents from 20 of the world’s top-ranked universities (per the QS 2024 rankings) using Bacchi’s “What’s the problem represented to be?” framework. It’s a method designed to look past what policies say and examine what they assume. And what these policies assume, overwhelmingly, is that students can’t be trusted to produce their own work.
Across all 20 institutions, Luo found the same dominant concern: students may not submit original work. Terms like “original work,” “own work,” and “authors” appeared frequently. GenAI was almost universally treated as external assistance, something separate from a student’s own intellectual contribution, placed in the same category as ghostwriting or getting someone else to complete an assignment. Several universities required students to sign “originality declaration forms,” and unauthorized use of GenAI was labeled a serious disciplinary offense. The messaging is clear: if AI touched it, it’s not really yours.
Luo argues that this pattern reveals a deeply rooted assumption, one that rarely gets questioned. The policies treat originality as synonymous with independent production. A student’s work is original if it was produced alone, without AI involvement. But that definition of originality, Luo points out, doesn’t hold up in a world where knowledge production is increasingly collaborative and mediated by technology.
The paper references Johnson-Eilola and Selber (2007, cited in Luo, 2024), who described the persistent image of the “lone genius in the attic slaving away on a piece of written work” as the implied gold standard. Almost two decades later, that image is still shaping institutional policy, and GenAI has made its limitations impossible to ignore.
There’s a related blind spot Luo identifies that I think is especially important. GenAI isn’t just ChatGPT. It’s now embedded in everyday productivity tools: Microsoft Office, Google Workspace, email clients. Luo notes, drawing on Chan (2023, cited in Luo, 2024), that students may not even be fully aware they’re using AI when it’s woven into the software they already rely on for coursework. Policies that draw a hard line between “AI-assisted work” and “original work” don’t account for this reality. The line is already blurred, and it’s going to keep blurring.
GenAI and Student Originality
The study also identifies a troubling pattern in how originality is policed. It’s almost always approached through the lens of academic misconduct and surveillance. Originality gets reduced to a number, like Turnitin’s “originality score,” or a compliance exercise, like signing a declaration form.
Very few of the policies Luo reviewed explored what actually makes student work meaningful or original in any substantive sense. Only one university out of twenty offered a more nuanced take, acknowledging that human text production involves associations and remixing from prior reading, a process not entirely unlike what AI does. And even that university ultimately concluded the comparison still warranted treating AI-generated content as ghostwriting.
Luo (2024) argues that “a critical silence in higher education policies concerns the evolving notion of originality in the digital age and a more inclusive approach to address the originality of students’ work is required.” (p. 651). That’s the core of the paper. The policies aren’t wrong to care about originality. They’re wrong about what originality means.
This connects directly to what Sarah Eaton argued in her work on postplagiarism, which I’ve covered on this blog. Eaton’s point was that hybrid human-AI writing is becoming the new normal, and that AI-generated texts are “edited, revised, reworked, and remixed” by humans in ways that make it nearly impossible to draw a clean line between human and AI contributions. Luo builds on that argument and applies it specifically to how policies construct the problem. The framing isn’t neutral. It shapes how students, teachers, and institutions understand what counts as legitimate work.
The effects Luo identifies are worth spelling out. At the discursive level, the policies create a hierarchy where fully human-produced work is treated as more valuable than human-AI collaborative work. Luo (2024) contends that:
the highlighted distinction between ‘AI-assisted work’ and ‘original work’ in policies may still construct a hierarchy – work that is thoroughly ‘human’ is understood to be more original whereas AI-collaborated work is useful but less valued. In this context, many emerging forms of human-AI collaborative work could be sidelined due to a lack of ‘originality’. Subjectively, this representation may stigmatise students who use GenAI regardless of how they use it, assuming that their work is less authentic or valuable. Considering the lived effects, students may become reserved, if not resistant, to leveraging such technology in their learning. (p. 660)
At the subjectification level, students are positioned as potentially untrustworthy and teachers as gatekeepers. Assessment redesign gets framed as a way to reduce cheating. Luo (2024) observes that “far too often the representations of students and technology convey a conception of education that orientates around ‘catching’ students rather than engaging them” (p. 660). At the lived level, students may avoid using GenAI for legitimate learning purposes out of fear of being flagged, which defeats the entire point of thoughtful AI integration.
I’ve seen this dynamic up close in other policy research I’ve covered. McDonald, Johri, Ali, and Hingle Collier (2025) analyzed GenAI guidance from 116 US R1 universities and found similar contradictions: institutions simultaneously encouraging GenAI use and treating it as a misconduct risk, with privacy guidance that was vague and inconsistent. The pattern is the same across both studies.
Policies want to look forward and backward at the same time, embracing GenAI as a tool for innovation and treating it as a threat to be contained. Those two positions can coexist, but not without much clearer thinking about what originality actually looks like in a world where AI is part of the process.
Luo’s proposed alternative is a context-dependent understanding of originality, one that considers disciplinary context, learning goals, and the specific nature of what students are offloading to AI. A student who uses GenAI to analyze a dataset and then applies evaluative judgment to refine the output is doing something meaningfully different from a student who submits raw AI-generated text. Policies need to reflect that range. Luo calls for a spectrum approach to originality, recognizing different degrees of human-AI collaboration and grounding the assessment of originality in what students actually demonstrate they can do.
The paper also recommends that universities move away from compliance-driven policy language. Luo (2024) argues that “policies can place more emphasis on the available support to students in producing original work that is meaningful to their learning” and stresses that “open communication and collaboration with students and faculty members should be stressed over compliance in developing policies to foster a culture of trust and care” (p. 662). The practical suggestions are concrete: partner with students in policy development, host consultations about what originality means in their disciplines, and emphasize support over surveillance to build a more learning-centered approach.
I think Luo is right that the concept of originality needs serious rethinking. And I’d add that the longer universities avoid that conversation, the wider the gap grows between what their policies say and what their students actually experience. The world has moved on from the lone genius in the attic. It’s time the policies caught up.
References
- Chan, C. K. Y. 2023b. “Is AI Changing the Rules of Academic Misconduct? An in-Depth Look at Students’ Perceptions of AI-Giarism.” arXiv Preprint.
- Eaton, S. E. (2023). Postplagiarism: Transdisciplinary ethics and integrity in the age of artificial intelligence and neurotechnology. International Journal for Educational Integrity, 19(23). https://doi.org/10.1007/s40979-023-00144-1
- Johnson-Eilola, J., and S. A. Selber. 2007. “Plagiarism, Originality, Assemblage.” Computers and Composition 24 (4): 375–403.
- Luo, J. (2024). A critical review of GenAI policies in higher education assessment: A call to reconsider the “originality” of students’ work. Assessment & Evaluation in Higher Education, 49(5), 651-664. https://doi.org/10.1080/02602938.2024.2309963
- McDonald, N., Johri, A., Ali, A., & Hingle Collier, A. (2025). Generative artificial intelligence in higher education: Evidence from an analysis of institutional policies and guidelines. Computers in Human Behavior: Artificial Humans, 3, 100121. https://doi.org/10.1016/j.chbah.2025.100121