Almost twenty years before ChatGPT made everyone panic about originality in student writing, two composition scholars were already arguing that the whole framework was broken. Johnson-Eilola and Selber (2007) didn’t use the word AI. They didn’t need to. Their argument was about something more fundamental: the way academic writing has always defined originality is too narrow, and it’s been causing problems long before any machine could write a paragraph.
Their paper, published in Computers and Composition, makes a case for what they call “assemblage,” texts built primarily and explicitly from existing texts to solve a communication problem in a new context. The idea isn’t that students should stop writing original work. It’s that assemblage should be recognized as an equally valid form of composition. And reading this in 2026, with AI tools that can generate, remix, and recombine text at scale, the argument feels less like a theoretical provocation and more like a prediction that came true.
The Lone Genius Problem
Johnson-Eilola and Selber (2007) start with a critique that anyone in composition studies would recognize: the field has theoretically moved past the Romantic notion of the lone creative genius, but classroom practice hasn’t caught up. Teachers still expect students to produce texts where borrowed material is clearly separated from, and subordinated to, the student’s own words. As the authors put it:
“When all is said and done, teachers seem to ask students the question: After you have read all the background material and assembled your evidence, what did you, just you, produce? Show us your words; let the words of others fade into the background. ” (p. 379)
That sentence was written in 2007 and it describes almost exactly how many universities responded to generative AI in 2023 and 2024. The first instinct was to ask: which words are the student’s, and which came from the machine? I covered Luo’s (2024) review of university originality policies and found them full of contradictions, vague definitions, and impractical expectations. Johnson-Eilola and Selber would probably say the contradictions aren’t surprising. The underlying concept of originality was already incoherent before AI entered the picture.
Citation as Surveillance
One of the most striking arguments in the paper is about what citation practices actually do in the classroom. Johnson-Eilola and Selber (2007) argue that citations exist mainly to help teachers separate what the student produced from what someone else produced, so the teacher can assign primary value to the “original” text. The hierarchy is clear: your words matter most, borrowed words are secondary.
The irony, as the authors point out, is that this hierarchy encourages the very behavior teachers are trying to prevent. Students know their “own words” are valued above everything else. So some of them hide their borrowings. They paraphrase aggressively, not because paraphrasing deepens understanding, but because it makes borrowed ideas look like original ones. The system rewards concealment.
Johnson-Eilola and Selber (2007) draw on scholars who had already challenged this logic. Rebecca Moore Howard (1993) reframed “patchwriting” as a developmental literacy practice, not a moral failure. Margaret Price (2002) questioned whether plagiarism is even a stable, monolithic concept. And James Porter (1986) showed how academics routinely engage in what he called “ethical plagiarism,” like sharing syllabi with plagiarism statements borrowed from other syllabi. The very professors who police student borrowing are themselves accomplished borrowers.
Eaton (2023) arrived at a similar conclusion from a different angle with her postplagiarism framework. The traditional plagiarism paradigm assumes clear boundaries between “mine” and “yours,” and both Johnson-Eilola and Selber and Eaton argue those boundaries have never been as solid as institutions pretend.
Assemblage as Legitimate Practice
The heart of the paper is the proposal to treat assemblage as a valid mode of academic composition. Johnson-Eilola and Selber (2007) define it as writing that works primarily through selection, arrangement, and recombination of existing materials. What counts isn’t whether the words are “yours” but whether the assembled text works, whether it achieves its rhetorical goals and produces the intended effects. “What counts, in this new context, is the ability of students to remix texts in ways that address specific issues, readers, and situations” (p. 380).
They offer examples from fields where assemblage is already standard practice. In web design, open source templates are freely downloaded, modified, and recombined. In music, DJs remix entire albums into new works. In architecture, Christopher Alexander’s “pattern language” provides a grammar of reusable design components. And in perhaps their most vivid example, Rob Ryang took existing footage from Stanley Kubrick’s The Shining and rearranged it into a feel-good romantic comedy trailer. Every frame was borrowed. The result was undeniably a new creative work with its own rhetorical argument.
The logic is compelling: “Re-inventing the wheel becomes an inefficiency, a misplaced waste of effort. ‘You borrowed that chunk? Great! Where did you get it from? Maybe I can use it, too’” (p. 400). If the goal of writing shifts from performance (displaying that you can produce original sentences) to action (solving a real communication problem), then finding, curating, and arranging existing material becomes a genuine skill, not a shortcut.
I covered Guetzkow, Lamont, and Mallard’s (2004) research on how peer reviewers define originality and found that “original approach,” the ability to frame a problem in a new way, was valued far above “original results.” Johnson-Eilola and Selber’s assemblage model aligns with that finding. A student who selects the right existing texts, arranges them strategically, and produces something that works in context is demonstrating original approach, even if they didn’t write any of the individual sentences.
What This Means for the AI Conversation
Johnson-Eilola and Selber (2007) wrote this paper when remix culture was associated with music sampling and web templates. In 2026, the most powerful remixing tool is generative AI itself. Students can prompt an AI to draft, combine, rewrite, and restructure text in seconds. The assemblage model says that’s fine, as long as the student is making deliberate, informed decisions about selection and arrangement, and as long as the result solves a genuine communication problem.
I find this persuasive up to a point. The authors are right that originality has been too narrowly defined and that the citation-as-surveillance model creates perverse incentives. Their proposal to value assemblage alongside original composition is a healthy correction. But the model needs some additional guardrails when AI is involved.
The difference between a student carefully curating and remixing source material and a student dumping a prompt into ChatGPT and submitting the output is enormous, even though both could technically be called “assemblage.” The authors acknowledge legal and ethical concerns but argue they should be separated from pedagogical goals. That separation was easier in 2007. In 2026, the ethical, pedagogical, and legal dimensions of AI-assisted writing are deeply entangled.
Still, the core argument holds. The authors redefine creativity for a remix culture:
In the end, as we see it, this all comes down to a reconfigured notion of “creativity,” one more in line with postmodern work. Creativity is no longer, as we said, re-inventing the wheel, which does not remove creativity but shifts it to the assemblage: Take what already exists and make something else, something that works to solve problems in new, local contexts. Creativity, in this rearticulation, involves extensive research, filtering, recombining, remixing, the making of assemblages that solve problems. Citation is no longer a way of marking subordinate elements in a text to downplay their value in student work but a way to reward students for their new skills, to situate texts not only in pre-existing but new contexts.” (p. 400)
If universities are serious about teaching students to work with AI, this is the direction the assessment conversation needs to go. Not “did you write this?” but “does this work, and can you explain why you built it the way you did?”
References
- 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
- Guetzkow, J., Lamont, M., & Mallard, G. (2004). What is originality in the humanities and the social sciences? American Sociological Review, 69(2), 190–212. http://www.jstor.org/stable/3593084
- Howard, Rebecca Moore. (1993). A plagiarism pentimento. Journal of Teaching Writing, 11, 233–246.
- Johnson-Eilola, J., & Selber, S. A. (2007). Plagiarism, originality, assemblage. Computers and Composition, 24(4), 375–403. https://doi.org/10.1016/j.compcom.2007.08.003
- 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
- Porter, James E. (1986). Intertextuality and the discourse community. Rhetoric Review, 5, 34–47.
- Price, Margaret. (2002). Beyond ‘gotcha!’: Situating plagiarism in policy and pedagogy. College Composition and
Communication, 54, 88–115.