{"course":{"slug":"ai-for-teachers","title":"AI for Teachers","url":"https://addaly.com/learn/ai-for-teachers","level":"beginner","track":"work"},"lesson":{"slug":"handling-a-case-fairly","title":"Handling a case fairly","url":"https://addaly.com/learn/ai-for-teachers/handling-a-case-fairly","position":39,"of":57,"minutes":9,"answer":"When a case reaches a decision, the standard is what a reasonable colleague would conclude from the evidence you can show (the process record, the conversation, the student's own explanation) never a detector score.","keyIdea":"When a case reaches a decision, the standard is what a reasonable colleague would conclude from the evidence you can show (the process record, the conversation, the student's own explanation) never a detector score, and the default outcome is redoing the work, with the second-language writer given explicit protection from the accusation that reads unsurprising prose as machine prose."},"body":{"format":"markdown","text":"## The question is not \"was it AI\"\n\nMost of the anxiety in a suspected case comes from trying to answer the wrong question. Whether a specific paragraph came from a model, a cousin, an older sibling or a website is usually unknowable and usually irrelevant. The policy's question is whether the student did the work the task existed to make them do. That question has evidence. The other one has guesses.\n\nReframe every case that way and most of them become manageable.\n\n## What counts as evidence\n\nIn rough order of weight:\n\n1. **The student's own account.** From the two-minute conversation the detection lesson described. Not whether they confessed: whether they can talk about the work as someone who did it. Can they explain why they chose this example, what they cut, what they found hard.\n2. **The process record.** Version history, the in-class paragraph, the plan, the drafting folder. Does the finished piece connect to a visible beginning.\n3. **Comparison with their other work.** Not \"this is too good for her\": students improve, and that judgement is the one most likely to be wrong and most likely to fall on the student whose ability you underestimated. But: a piece that argues in a way the student has never argued, on a topic she showed no grasp of in class last week, is a reason to look at the other evidence more carefully.\n4. **Specific features of the text**: a citation that does not exist, a reference to a resource the class never used, an American spelling throughout in a British-spelling class. Weak on their own, useful with the rest.\n\nNot evidence: a detector score. The detection lesson explained why, and the class policy promised not to use one. If a colleague or a head asks for one, the answer is that it would add a number with a known bias against second-language writers to a decision that has real evidence available.\n\n## The standard\n\nThe standard is what a reasonable colleague, shown the evidence, would conclude. Not certainty (you will rarely have it) and not suspicion. If two teachers looking at the conversation notes and the process record would both say \"she did not write this\", act. If one would and one would hesitate, the evidence is not there, and the outcome is different.\n\nWrite the evidence down before deciding. A page: what the task was, what level it was set at, what the process record shows, what was said in the conversation, what the student says. A decision made after writing that page is better than one made before it, and it is the page the appeal will be judged on.\n\n## The outcomes\n\nThe class policy set the ladder, and the first step on it is the outcome for nearly every first case: **redo the work, under supervision, and talk about what the task was for.** It fits the harm, it is not humiliating, and it produces the learning the original task was meant to.\n\nWhen the evidence is short of the standard, the outcome is still available in a softer form: \"I could not follow your argument when we talked. Write me the second half again, here, and let's see.\" Not an accusation, not a penalty. A student who wrote the original will produce the second half easily. One who did not has redone the exercise without a finding against them, which is the right result when you were not sure.\n\nEscalation (the school's formal procedure, a note on a record) is for the second and third time, or for the high-stakes piece where the school's rules require it. Not for the first case, and not on a teacher's own authority.\n\n## The second-language safeguard\n\nPut this in writing in the department: a student writing in a second or third language is never referred on the basis of the text alone. Their prose is the prose that every automated and half-automated judgement misreads, and the harm of a wrong accusation to a student already working in a language not their own is severe and lasting. For those students the process record and the conversation carry the whole weight, and the conversation is held in a way that gives them room: in their strongest language if a colleague can help, with time to think, without an audience.\n\n## The meeting\n\nIf a parent is involved, the meeting is about the evidence page, the policy the student agreed to, and the outcome, in that order. Not about AI. Not about the character of the child. \"The task was Level 0. The process record shows the essay arrived at once. When we talked, she could not explain the argument. Under the policy she agreed to, she redoes it with me. No mark penalty. Here is the policy.\" Most parents, shown that, are on the teacher's side, because the process is fair and the outcome is mild.\n\n## What not to do\n\nDo not raise it in front of the class. Do not open with \"did you use AI\". Do not use words like cheating or plagiarism in the first conversation. Do not compare the student with a sibling. Do not act on a detector. Do not decide before writing the page. Each of these is a way to be wrong loudly, and a loud wrong accusation costs a child more than a missed case costs anyone.\n\n## The honest limit\n\nSome students will get away with it. That has always been true and it is more true now. The alternative (a regime that catches more by suspecting everyone) is worse, and the detection lesson showed who it catches. A fair process that occasionally misses is the right trade, and a teacher who has designed the high-stakes tasks so that the work has to happen in the room has already moved most of what matters out of reach of the question."},"check":{"question":"After a conversation, a student cannot explain the argument in her essay, the version history shows a single paste, and she says a cousin helped. The teacher has no other evidence. What is the fair outcome?","options":[{"text":"Record it formally as AI cheating, since the single paste and the failed conversation taken together are conclusive evidence.","correct":false,"tempting":"Two signals are consistent with several explanations, including the one she gave; \"conclusive\" is stronger than the evidence."},{"text":"Take no action at all, since without proof that a model was used there is nothing in the policy to act on.","correct":false,"tempting":"The policy's concern is whether she did the work, not which tool she did not use; the evidence that she did not do it is substantial."},{"text":"Have her redo the piece under supervision, record the evidence, and leave the tool question out.","correct":true},{"text":"Run a detector on the essay to settle whether the cousin or a machine wrote it, and then decide the outcome on that basis.","correct":false,"tempting":"A detector score cannot distinguish a cousin from a model and would add a number with a known bias to a decision that does not need it."}]},"prev":{"slug":"process-evidence-that-holds-up","title":"Process evidence that holds up","url":"https://addaly.com/learn/ai-for-teachers/process-evidence-that-holds-up"},"next":{"slug":"rethinking-homework","title":"Rethinking homework","url":"https://addaly.com/learn/ai-for-teachers/rethinking-homework"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Handling a case fairly', in AI for Teachers. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-teachers/handling-a-case-fairly (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Handling a case fairly","courseTitle":"AI for Teachers","module":{"index":5,"slug":"integrity-and-ai-written-work","title":"Integrity: AI-written work and what to do about it"},"answer":null,"keyIdea":"When a case reaches a decision, the standard is what a reasonable colleague would conclude from the evidence you can show (the process record, the conversation, the student’s own explanation) never a detector score, and the default outcome is redoing the work, with the second-language writer given explicit protection from the accusation that reads unsurprising prose as machine prose.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-question-is-not-was-it-ai","runs":[{"text":"The question is not “was it AI”"}]},{"t":"p","runs":[{"text":"Most of the anxiety in a suspected case comes from trying to answer the wrong question. Whether a specific paragraph came from a model, a cousin, an older sibling or a website is usually unknowable and usually irrelevant. The policy’s question is whether the student did the work the task existed to make them do. That question has evidence. The other one has guesses."}]},{"t":"p","runs":[{"text":"Reframe every case that way and most of them become manageable."}]},{"t":"h2","id":"what-counts-as-evidence","runs":[{"text":"What counts as evidence"}]},{"t":"p","runs":[{"text":"In rough order of weight:"}]},{"t":"ol","start":1,"items":[[{"text":"The student’s own account.","b":true},{"text":" From the two-minute conversation the detection lesson described. Not whether they confessed: whether they can talk about the work as someone who did it. Can they explain why they chose this example, what they cut, what they found hard."}],[{"text":"The process record.","b":true},{"text":" Version history, the in-class paragraph, the plan, the drafting folder. Does the finished piece connect to a visible beginning."}],[{"text":"Comparison with their other work.","b":true},{"text":" Not “this is too good for her”: students improve, and that judgement is the one most likely to be wrong and most likely to fall on the student whose ability you underestimated. But: a piece that argues in a way the student has never argued, on a topic she showed no grasp of in class last week, is a reason to look at the other evidence more carefully."}],[{"text":"Specific features of the text","b":true},{"text":": a citation that does not exist, a reference to a resource the class never used, an American spelling throughout in a British-spelling class. Weak on their own, useful with the rest."}]]},{"t":"p","runs":[{"text":"Not evidence: a detector score. The detection lesson explained why, and the class policy promised not to use one. If a colleague or a head asks for one, the answer is that it would add a number with a known bias against second-language writers to a decision that has real evidence available."}]},{"t":"h2","id":"the-standard","runs":[{"text":"The standard"}]},{"t":"p","runs":[{"text":"The standard is what a reasonable colleague, shown the evidence, would conclude. Not certainty (you will rarely have it) and not suspicion. If two teachers looking at the conversation notes and the process record would both say “she did not write this”, act. If one would and one would hesitate, the evidence is not there, and the outcome is different."}]},{"t":"p","runs":[{"text":"Write the evidence down before deciding. A page: what the task was, what level it was set at, what the process record shows, what was said in the conversation, what the student says. A decision made after writing that page is better than one made before it, and it is the page the appeal will be judged on."}]},{"t":"h2","id":"the-outcomes","runs":[{"text":"The outcomes"}]},{"t":"p","runs":[{"text":"The class policy set the ladder, and the first step on it is the outcome for nearly every first case: "},{"text":"redo the work, under supervision, and talk about what the task was for.","b":true},{"text":" It fits the harm, it is not humiliating, and it produces the learning the original task was meant to."}]},{"t":"p","runs":[{"text":"When the evidence is short of the standard, the outcome is still available in a softer form: “I could not follow your argument when we talked. Write me the second half again, here, and let’s see.” Not an accusation, not a penalty. A student who wrote the original will produce the second half easily. One who did not has redone the exercise without a finding against them, which is the right result when you were not sure."}]},{"t":"p","runs":[{"text":"Escalation (the school’s formal procedure, a note on a record) is for the second and third time, or for the high-stakes piece where the school’s rules require it. Not for the first case, and not on a teacher’s own authority."}]},{"t":"h2","id":"the-second-language-safeguard","runs":[{"text":"The second-language safeguard"}]},{"t":"p","runs":[{"text":"Put this in writing in the department: a student writing in a second or third language is never referred on the basis of the text alone. Their prose is the prose that every automated and half-automated judgement misreads, and the harm of a wrong accusation to a student already working in a language not their own is severe and lasting. For those students the process record and the conversation carry the whole weight, and the conversation is held in a way that gives them room: in their strongest language if a colleague can help, with time to think, without an audience."}]},{"t":"h2","id":"the-meeting","runs":[{"text":"The meeting"}]},{"t":"p","runs":[{"text":"If a parent is involved, the meeting is about the evidence page, the policy the student agreed to, and the outcome, in that order. Not about AI. Not about the character of the child. “The task was Level 0. The process record shows the essay arrived at once. When we talked, she could not explain the argument. Under the policy she agreed to, she redoes it with me. No mark penalty. Here is the policy.” Most parents, shown that, are on the teacher’s side, because the process is fair and the outcome is mild."}]},{"t":"h2","id":"what-not-to-do","runs":[{"text":"What not to do"}]},{"t":"p","runs":[{"text":"Do not raise it in front of the class. Do not open with “did you use AI”. Do not use words like cheating or plagiarism in the first conversation. Do not compare the student with a sibling. Do not act on a detector. Do not decide before writing the page. Each of these is a way to be wrong loudly, and a loud wrong accusation costs a child more than a missed case costs anyone."}]},{"t":"h2","id":"the-honest-limit","runs":[{"text":"The honest limit"}],"role":"limit"},{"t":"p","runs":[{"text":"Some students will get away with it. That has always been true and it is more true now. The alternative (a regime that catches more by suspecting everyone) is worse, and the detection lesson showed who it catches. A fair process that occasionally misses is the right trade, and a teacher who has designed the high-stakes tasks so that the work has to happen in the room has already moved most of what matters out of reach of the question."}]}],"check":{"question":[{"text":"After a conversation, a student cannot explain the argument in her essay, the version history shows a single paste, and she says a cousin helped. The teacher has no other evidence. What is the fair outcome?"}],"options":[{"text":[{"text":"Record it formally as AI cheating, since the single paste and the failed conversation taken together are conclusive evidence."}],"correct":false,"tempting":[{"text":"Two signals are consistent with several explanations, including the one she gave; “conclusive” is stronger than the evidence."}]},{"text":[{"text":"Take no action at all, since without proof that a model was used there is nothing in the policy to act on."}],"correct":false,"tempting":[{"text":"The policy’s concern is whether she did the work, not which tool she did not use; the evidence that she did not do it is substantial."}]},{"text":[{"text":"Have her redo the piece under supervision, record the evidence, and leave the tool question out."}],"correct":true,"tempting":null},{"text":[{"text":"Run a detector on the essay to settle whether the cousin or a machine wrote it, and then decide the outcome on that basis."}],"correct":false,"tempting":[{"text":"A detector score cannot distinguish a cousin from a model and would add a number with a known bias to a decision that does not need it."}]}],"order":[0,1,3,2],"why":null},"prev":{"slug":"process-evidence-that-holds-up","title":"Process evidence that holds up","minutes":9},"next":{"slug":"rethinking-homework","title":"Rethinking homework","minutes":9}}}