Chatbots can now produce a passable essay in seconds. In response, we can ban the tools, revive closed-room exams, use AI detectors or ask students to prove that every sentence is their work. Each may have a place, but none solves the basic problem. A polished final essay now tells us less than it used to about how the work was done.
So, the academic writing course that I teach seems especially vulnerable to generative artificial intelligence. In this course, students produce structured, grammatical, formal prose; but from submitted text alone, I can no longer discern how the work was produced, whether the claims have been checked or how much judgement went into shaping the argument. This uncertainty existed before AI; students could already enlist a private tutor or to ask someone else to write a take-home essay.
- Your students use GenAI. Let’s redesign learning so thinking still happens
- The imperative for productive struggle
- The best way to teach students to think critically about AI? Make them argue with it
My response has been to broaden what I assess. My approach does not depend on authenticating every sentence. The student’s account of how they worked becomes the starting point for feedback. I no longer treat the finished text as sufficient evidence of learning, so I ask students to make more of the writing process visible and assess their judgement within it. Drafting remains part of the work; a chatbot may supply an early version, and the student continues writing from there, turning it into an argument they understand and can defend.
Students can use AI for certain tasks, and it is sometimes required. When students use it, they must question, revise and take responsibility for output from large language models (LLMs).
For shorter tasks, students submit a marked-up text. Text copied verbatim from AI is in bold. AI output that has been meaningfully revised is in italics. Text written entirely by the student is left unmarked. Students also add brief comments explaining what they changed and why.
This method allows me to see at sentence-level how students describe their writing process. It provides a tangible point for discussion that doesn’t necessarily require authentication. Did students blindly copy the output? Did they spot the weaknesses, refine unclear statements, edit the organisation and ultimately take responsibility for the argument?
Longer assignments can become unwieldy when I try to annotate everything inline. Instead, students submit an AI appendix with an explanation of the prompts, raw text from the LLM and a note on how they verified and edited. The rule is simple. Students may use AI or avoid it without penalty, but concealing its use is penalised.
I have also adjusted the rubric. I assess the quality of argument, organisation, lucidity and referencing. The latter is manageable since the tasks are sequential. Students begin with summarising a paper selected from a list, using that as the framing device for a literature review. Then, they discuss the introduction and developing bibliography with the class in a seminar, allowing me to challenge or recommend further reading. Finally, I can verify and cross-check new references and inappropriate citation of common texts in the final assignment.
Part of the grade now reflects the judgement that students exercise when working with generated material. Can the student flag fake references, misplaced confidence or incoherent framing? Can they fix those problems? Can they explain their decisions?
On assignments requiring the use of AI, students produce outputs, edit them, report on their process and reflect on the experience in seminars. Here, students bring their drafts and LLM outputs and work through a paragraph from inception to final form, explaining where AI entered the process and what has changed. The discussion starts with argument and sources, then flows into editing the style and wording. In these dialogues, I can gauge whether students truly “own” the work.
This reveals differences in how students use AI. For example, some students might copy and paste verbatim and minimally edit content. Others start with AI-generated content but edit, shape, restructure and rewrite to convey different ideas. Some barely engage with the technology. Any of these approaches can support learning, provided students take responsibility for the final text and show how they checked and developed it.
A student can lie and present AI-generated text as their own, but no educational design can eliminate dishonesty, just as anti-plagiarism policies failed to eradicate plagiarism, even if process-based assessment alters the incentives. My course is small enough that I see students weekly in seminars of five to seven people and asking them to explain their reasoning is routine feedback. I can become familiar with their writing habits over time, which makes it harder to sustain a false story. The fact that there is no penalty for AI use means that the cost of dishonesty is higher than the reward, so it is simpler to be truthful.
Of course, this level of engagement isn’t feasible in all courses. Other faculty such as those teaching finance, public health, engineering, political theory and marketing may not be able to allocate time for this kind of dialogue with every student. However, I imagine that a single student-led conversation per term, where students choose a paragraph and explain how it developed and where they brought in AI, would suffice. Or, for larger assignments, the narrative could go into a short appendix.
These changes affect the meaning of the work. A draft is now one step in the writing process, and AI can contribute to an early draft, with students continuing to write as the argument develops. By the end, they should be able to justify what they’ve written and feel comfortable defending it.
GenAI has accelerated certain types of writing significantly and introduced new challenges to assessors. Silence and prohibition don’t reveal anything about the underlying process. A more productive response is to create opportunities to see and assess how students develop their writing when AI is involved.
Igor B. Martins is a researcher in the department of economic history in the School of Economics and Management at Lund University, Sweden.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
comment