The essay that shifted in style
Chloe, 16, submits a 1,500-word essay for her SEC Ethics coursework on Kant's categorical imperative. Her teacher, Mr Camilleri, notices a shift in style between the first two sections and the last, and a reference to a book the school library does not hold. A free AI-detection tool returns 82% 'likely generated'. The tool's documentation says false positives are common for students writing in a second language, and Chloe's first language is not English. The school's assessment policy, last revised three years ago, says that work must be 'the candidate's own' but says nothing about AI assistance; another teacher has told students that using chatbots to improve wording is acceptable. Chloe's grade depends on this essay, and she plans to study law. Mr Camilleri asks for ethical advice.
1. Is Chloe's conduct dishonest if she used AI to rewrite sentences but supplied all the argument herself? Argue using virtue ethics and one duty-based consideration.
Read the worked answer
Virtue ethics. The relevant virtues are honesty, integrity and diligence. Honesty concerns what she represents to others, not merely what she did. If the essay is substantially her own argument and she has checked and understood every sentence, using a tool to improve expression is closer to using a dictionary or a proofreader than to cheating. But the assessment exists to test her ability to construct and express an ethical argument; if she presents polished prose as evidence of unaided writing, she is allowing a false impression of her competence to stand. Integrity also asks whether she would be comfortable explaining exactly what she did - a useful test of whether conduct is honest.
Duty-based consideration. There is a duty not to deceive the examiner about the authorship of the work being assessed, and a duty of fairness to other candidates assessed under the same conditions. Kant's universalisability test is informative: 'students may use AI to polish wording where the school has not forbidden it' can be universalised without contradiction, but 'students may present assisted writing as entirely their own' cannot, because assessment depends on the truth of that claim.
Judgement. Her conduct is not dishonest merely because a tool was involved. It becomes dishonest to the extent that she concealed help that a reasonable examiner would want to know about, or submitted writing she could not reproduce or defend. The ambiguity of the school's own rule, and the other teacher's permission, count significantly in her favour: she did not breach a clear instruction.
2. What should Mr Camilleri do, given the policy is silent and the detector unreliable?
Read the worked answer
Do not decide on 82%. The tool's own documentation concedes frequent false positives for second-language writers, so treating it as proof would risk punishing Chloe for the way she writes. Procedural fairness requires evidence proportionate to the seriousness of the allegation.
Gather better evidence. Ask Chloe to explain her argument and her sources in conversation, including the book the library does not hold - she may have borrowed it, bought it, or read an extract online. Check the document's version history, earlier drafts, and any classwork in the same topic. Offer her the chance to respond before any conclusion. Interviewing her is not an accusation; unexplained failure to explain her own argument would be far stronger evidence than any detector score.
Apply the rules that existed, not new ones. At the time of submission there was no AI policy. The policy says 'the candidate's own work', which is genuinely ambiguous. It would be unfair to invent a prohibition retrospectively and fail her under it.
Proportionate response if AI use is confirmed. A serious conversation about disclosure, and possibly a requirement to resubmit or to write a supervised version, rather than the collapse of her whole qualification. If no evidence emerges, the matter ends and she should not be left under suspicion.
Institutional duty. Mr Camilleri should take the case to the head of department as a trigger to write a clear, prospective AI policy, with examples of permitted and prohibited use, and to train staff. The ambiguity is the school's failing, not hers.
3. Consider fairness to students who did not use AI.
Read the worked answer
Two distinct fairness questions. First, fairness in competition: if some students use AI to produce better prose and are not penalised, those who follow the rules are disadvantaged in grades that affect progression to university. Second, fairness in disclosure: if rules are unstated, students who asked and were told 'chatbots are fine for wording' are treated the same as those who used AI to generate whole sections, which is unfair to both.
Why consistency matters more than severity. A school that punishes Chloe severely while leaving the ambiguity, and the other teacher's permission, uncorrected is applying a standard retrospectively and selectively. That is arbitrary. The fair approach is a clear rule announced in advance, applied to everyone equally, with a distinction between assistance with expression and generation of content or argument. Until that rule exists, the fairest thing is to judge Chloe by the standards that actually applied when she wrote.
The school's obligation to the many, not just the one. Fairness also requires checking whether other submissions in the same cohort raise the same issue. Investigating only the essay that looked suspicious punishes the conspicuous rather than the culpable, and it may disadvantage students whose second-language writing style triggers detection tools.
Fairness to Chloe. She is also entitled to fair treatment: no retroactive rule, no decision on an unreliable percentage, and a chance to explain. A fair process protects everyone in the class, including those who did nothing wrong, because it makes the standard knowable.