📚 Is Using AI to Write Homework Cheating? New Guidance on Exam Misconduct | 用AI写作业算作弊吗：考试违规认定的新指引
Since generative AI tools became widely available, students, teachers and awarding bodies have been arguing about one deceptively simple question: if you use an AI chatbot to help with your homework, are you cheating? The short answer is that the tool itself is almost never the problem. What matters is whether the work you submit honestly represents your own knowledge, understanding and effort. This article explains how the new misconduct guidance works in practice, where the red lines sit, and how to use AI without risking your qualification.
自从生成式 AI 工具普及以来，学生、老师和考试局一直在争论一个看似简单的问题：用 AI 聊天机器人帮忙做作业，算不算作弊？简短的回答是：工具本身几乎从来不是问题所在。真正重要的是，你提交的作业是否诚实地反映了你本人的知识、理解与努力。本文将解释新的违规认定指引在实操中如何运作、红线在哪里，以及怎样使用 AI 才不会危及你的成绩与资格。
1. Why This Question Matters Now | 为什么这个问题现在如此重要
AI is no longer a novelty. It sits inside search engines, word processors, translation apps and revision platforms. Every academic year, more students use it casually, often without thinking of it as a "tool" at all. Meanwhile, schools have had to write policies faster than the technology changes, and awarding bodies have had to decide what counts as malpractice when the "unauthorised material" in question is a chatbot on a phone.
AI 已经不再是什么新鲜事物。它嵌入了搜索引擎、文字处理软件、翻译应用和复习平台。每一学年，都有更多学生随手使用它，甚至根本不把它当成一种"工具"。与此同时，学校不得不以快于技术变化的速度撰写政策，考试局也不得不裁定：当所谓的"未授权材料"是手机里的一个聊天机器人时，什么才算违规。
The result is a patchwork: some departments ban AI outright, some encourage it, and some say nothing at all. That inconsistency is dangerous for students, because ignorance of a rule is not usually accepted as a defence. Understanding the underlying principles, rather than memorising one school's rules, is the safest strategy.
于是就形成了一种拼凑式的局面：有的院系全面禁止 AI，有的鼓励使用，还有的干脆不作说明。这种不一致对学生很危险，因为"不知道规定"通常不能作为免责理由。掌握背后的判定原则，比死记某一所学校的条文更安全。
2. What Formal Rules Actually Call Cheating | 正式规则中"作弊"的界定
Formal misconduct regulations rarely mention specific technologies. Instead, they define categories of behaviour. The most common categories are plagiarism, collusion, impersonation, fabrication, and the introduction of unauthorised material into an assessment. Notice that none of these categories names a device or an app.
正式的违规条例很少点名具体技术，而是界定行为类别。最常见的类别包括：抄袭、串通合作、代考（冒名顶替）、数据或引文造假，以及在考核中携带未授权材料。请注意，这些类别里没有一个提到某种设备或应用。
What all of these categories share is misrepresentation. Cheating is, at its core, presenting something as yours when it is not, in a context where the examiner is relying on it being yours. That is why a calculator is allowed in some papers and banned in others: the question is never "is this technology?", but "does using it stop the assessment from measuring what it claims to measure?"
这些类别的共同点是"虚假呈现"。作弊的核心，是把不属于你的东西当作你自己的成果提交，而在这个场景中，考官正是依赖"这是你的成果"来评判。这也解释了为什么计算器在某些试卷中允许、在另一些试卷中禁止：问题从来不是"这算不算技术"，而是"使用它是否让这项考核无法测量它声称要测量的东西"。
So the useful definition for AI is this: misconduct occurs when AI use is unauthorised and the submitted work misrepresents your own ability. Either element alone is usually not enough — but in most homework and coursework settings, both are present at once.
因此，对 AI 而言有用的定义是：当 AI 的使用未经授权，并且提交的作品虚假呈现了你本人的能力时，违规就发生了。单独任一要素通常都还不够，但在大多数作业和课程作业情境中，两者往往同时出现。
3. The Three-Use Test: Tool, Tutor or Substitute | 三种用途测试：工具、导师还是替代者
A practical way to judge your own behaviour is to ask which of three roles the AI played. If it acted as a tool, you did the thinking and it did the mechanical labour. If it acted as a tutor, it explained, questioned and gave feedback on work that was already yours. If it acted as a substitute, it produced the thinking and you produced nothing but a submission.
判断自己行为是否合规，一个实用的方法是问：AI 扮演了以下三种角色中的哪一种。如果它是工具，思考是你做的，它只承担机械性劳动；如果它是导师，它对你已有的成果进行讲解、追问和反馈；如果它是替代者，思考是它做的，你只贡献了一次提交动作。
- Tool examples | 工具类例子: spell-checking, grammar correction, formatting a bibliography, converting units, brainstorming a list of possible essay angles before you choose one. | 拼写检查、语法修正、排版参考文献、单位换算、在你做选择之前头脑风暴可能的论文角度。
- Tutor examples | 导师类例子: asking AI to explain the difference between ionic and covalent bonding, to generate five practice questions on integration by parts, to interrogate the logic of your own argument, to check whether a sentence you wrote is ambiguous. | 让 AI 解释离子键与共价键的区别、生成五道分部积分练习题、追问你自己论证的逻辑、检查你写的某句话是否有歧义。
- Substitute examples | 替代者类例子: pasting an essay question and submitting the output, asking AI to "rewrite this in your own words" and calling the result yours, generating experimental data, producing a portfolio piece. | 粘贴题目并直接提交输出、让 AI"用自己的话改写"后声称是原创、生成实验数据、产出作品集作品。
The decisive question is: could you defend this work in a five-minute conversation with your teacher, with no notes and no screen? If the honest answer is no, you did not write it — you received it.
决定性的问题是：不给笔记、不给屏幕，你能否在五分钟的对话中为自己的作品辩护？如果诚实的答案是不能，那就不是"你写的"，而是"你收到的"。
4. What the New Guidance Actually Says | 新指引究竟说了什么
Guidance issued by UK awarding organisations, including the Joint Council for Qualifications' advice on AI use in GCSE and A-level assessments, converges on a few principles. First, submitting AI-generated work as your own is malpractice. Second, where a task permits AI use, that use must normally be acknowledged. Third, AI cannot be listed as an author or cited as the source of original analysis, in the same way you cannot cite a friend as a source.
英国各考试局发布的指引——包括联合资格委员会（JCQ）关于 GCSE 与 A-level 评估中 AI 使用的建议——在若干原则上是趋同的。第一，把 AI 生成的作品当作自己的成果提交属于违规。第二，当某项任务允许使用 AI 时，通常必须对该使用予以说明。第三，AI 不能被列为作者，也不能被引作原创分析的来源，就像你不能把自己的朋友引作文献一样。
International guidance points the same way. The International Baccalaureate has taken the position that AI tools may be used as a resource in the same way as any other source, provided their use is acknowledged and the work remains the student's own. Ofqual's public statements follow a similar logic: the technology is not automatically banned, but misuse that undermines the validity of an assessment can be treated as malpractice.
国际层面的指引方向一致。国际文凭组织（IB）的立场是：AI 工具可以像任何其他来源一样被当作资源使用，前提是使用情况被说明、作品仍属学生本人。英国资格考试监管机构 Ofqual 的公开表态遵循类似逻辑：技术不会被自动禁止，但破坏考核有效性的滥用行为可能被认定为违规。
Three practical consequences follow. Teachers are expected to design tasks that are harder to fake. Centres are expected to have an AI policy that students have actually been told about. And students are expected to be able to explain, when asked, exactly what they did and what the AI did.
由此可推出三个现实后果：教师被期望设计更难"造假"的任务；学校被期望拥有一套确实已告知学生的 AI 政策；学生被期望在被问到时，能准确说明"我做了什么、AI 做了什么"。
5. Homework vs Coursework vs Exams | 作业、课程作业与考试情境的差异
The same act can be acceptable in one context and misconduct in another. Homework is usually formative: its purpose is to help you learn, and the sanctions for misuse are typically educational rather than formal. Coursework and non-exam assessment (NEA) sit in the middle: they count towards your grade, they are produced over time, and they require authentication. Written exams under invigilated conditions sit at the strict end: any unauthorised device or material is a breach by definition.
同样的行为在一种情境下可以接受，在另一种情境下就是违规。作业通常是形成性评价：目的是帮助你学习，违规后果一般是教育性的而非正式的。课程作业与非考试评估（NEA）处于中间地带：它计入成绩、耗时较长，并且需要真实性认证。在监考条件下的笔试则处于最严格的一端：任何未授权设备或材料，本身即构成违规。
| Context | 情境 | Typical rule | 通常规则 | Risk if misused | 滥用的风险 |
|---|---|---|
| Class practice tasks | 课堂练习 | Often permitted with teacher discretion | 通常由教师裁量是否允许 | Low; learning loss only | 低，仅损失学习效果 |
| Graded homework | 计分作业 | Usually must be your own; disclosure expected | 通常必须独立完成，应主动说明 | Medium; mark voided, meeting with teacher | 中，成绩作废、约谈 |
| Coursework / NEA | 课程作业 / NEA | Formal declaration of authenticity required | 需签署真实性声明 | High; malpractice investigation | 高，可能进入违规调查 |
| Invigilated exam | 监考考试 | No unauthorised devices or materials | 不得携带未授权设备或材料 | Severe; loss of grade or disqualification | 严重，成绩取消或取消资格 |
The trap for many students is a "grey zone" habit carried across contexts: using AI freely for practice, then unconsciously relying on it in a piece of coursework that is formally declared as your own. The habit does not become innocent because it is habitual.
许多学生踩的坑，是把"灰色地带"的习惯从一个情境带到另一个情境：平时自由使用 AI，然后在需要正式声明为独立完成的课程作业中不自觉地依赖它。习惯本身因为习以为常，并不会因此变得正当。
6. The Academic Integrity Spectrum | 学术诚信光谱
Rather than asking whether AI is "allowed", place your specific action on a spectrum. Green actions are generally safe, amber actions require disclosure and teacher approval, and red actions are misconduct regardless of intention.
与其问 AI 是否"被允许"，不如把你具体的行为放到一条光谱上。绿色行为通常安全，黄色行为需要说明并获得教师批准，红色行为无论主观意图如何都构成违规。
| Zone | 区域 | Behaviour | 行为 | Action required | 需要做什么 |
|---|---|---|
| Green | 绿色 | Explaining concepts, generating practice questions, checking grammar in your own draft | 讲解概念、生成练习题、检查自己草稿的语法 | None, but good practice to mention | 无需，但告知是好习惯 |
| Amber | 黄色 | Brainstorming ideas, structuring an argument, translating single terms, receiving feedback on your own paragraph | 头脑风暴、搭建论证结构、翻译单个术语、就你自己的段落获取反馈 | Disclose; confirm with the teacher first | 需说明；事先与教师确认 |
| Red | 红色 | Submitting AI-written text, paraphrasing output to hide it, fabricating data, using AI in an exam | 提交 AI 撰写的文本、改写输出以掩盖、伪造数据、在考试中使用 AI | Do not do it; it is malpractice | 不要做，这构成违规 |
Note that paraphrasing an AI draft does not move it from red to amber. Laundering text through a paraphraser, a translator or a second AI model is treats as deliberate concealment, and deliberate concealment generally increases the severity of the sanction rather than reducing it.
要注意：把 AI 草稿改写一遍并不会把它从红色变成黄色。把文本经改写器、翻译器或另一个 AI 模型"洗"一遍，会被视为刻意隐瞒，而刻意隐瞒通常会加重处罚，而不是减轻。
7. Red Lines You Should Never Cross | 绝不能越过的红线
Some behaviours are treated as misconduct in essentially every policy. Submitting AI-generated prose, code, artwork or data as your own original work is the clearest. So is using AI during an invigilated assessment, or bringing an unauthorised device into the exam room even if you never switch it on.
有些行为在几乎所有政策中都被视为违规。把 AI 生成的文字、代码、艺术作品或数据当作自己的原创提交，是最明确的一条。在监考评估中使用 AI，或把未授权设备带入考场——即使你从未开机——同样如此。
Other red lines are less obvious but equally serious: fabricating sources or citations that AI invented, using AI to complete a language assessment when the assessment is testing your own language production, using AI to answer questions in a live oral or viva, and colluding with others to share AI-generated answers as if they were individually produced.
另一些红线不那么显眼，但同样严重：引用 AI 编造出来的文献或出处；在考查你本人语言产出能力的语言评估中用 AI 完成；在实时口试或答辩中用 AI 答题；与他人串通，把 AI 生成的答案当作各自独立完成的结果分享。
Misconduct ⇌ Unauthorised Use × Misrepresentation of Authorship
Read that as a logical condition rather than a mathematical one: risk becomes misconduct when both factors are present. Remove either one — get authorisation and disclose it, or make the work genuinely yours — and the risk largely disappears.
把这个式子理解为逻辑条件而非数学运算：当两个因素同时存在时，风险就变成违规。去掉其中任何一个——要么获得授权并如实说明，要么让作品真正属于你——风险基本就消失了。
8. Declaration and Transparency: The Safest Strategy | 声明与透明：最稳妥的策略
If your school permits AI use for a task, disclosure is what separates legitimate assistance from misrepresentation. A good declaration is specific enough that a reader could reproduce your process. It normally names the tool and version, the date, the prompts used, the output received, and exactly what you did with it.
如果学校允许在某项任务中使用 AI，那么"如实说明"就是合法辅助与虚假呈现之间的分界线。一份好的声明应该具体到读者能够复现你的过程：通常需要写明工具名称与版本、日期、使用的提示语、得到的输出，以及你究竟如何处理它。
- Tool and version | 工具与版本: name the specific model, not just "an AI". | 写明具体模型，而不是笼统的"一个 AI"。
- Purpose | 用途: one sentence on what you needed it for. | 一句话说明你需要它做什么。
- Prompts | 提示语: include the key prompts in an appendix if the piece is long. | 长文可在附录中列出关键提示语。
- Verification | 核验: state how you checked facts, quotations and calculations independently. | 说明你如何独立核验事实、引文与计算。
- Your contribution | 你的贡献: describe what you wrote, changed, rejected or added. | 描述你撰写、修改、否决或增补了什么。
A useful sentence to keep in mind: "I used AI to do X; everything else in this piece, including the argument and the final wording, is mine." If you cannot write that sentence honestly about a piece of assessed work, the problem is not the sentence.
可以记住一句话："我用 AI 做了 X；除此之外，包括论证与最终措辞在内，本文其余部分均为我本人所作。"如果你无法诚实地就某份计分作品写下这句话，那么问题并不在这句话上。
9. AI Detectors: Promise and Limits | AI 检测工具：作用与局限
Teachers increasingly use AI-detection software, but these tools are probabilistic, not forensic. They estimate how likely a text is to have been machine-generated, and they are known to produce false positives — particularly for fluent non-native English writers, for formulaic academic phrasing, and for heavily edited drafts.
教师越来越常使用 AI 检测软件，但这类工具是概率性的，而不是鉴定性的。它们估算一段文本由机器生成的可能性，并且已知会产生误报——尤其是对写作流畅的非母语者、对套话式的学术表达，以及对经过大量修改的草稿。
A detection score alone is rarely enough to establish malpractice. What carries weight in a real investigation is a body of evidence: sudden shifts in style between drafts, version history that shows an entire essay appearing in one paste, citations that do not exist, technical vocabulary you cannot explain aloud, or your own answers in class being far weaker than your submitted work.
仅凭一个检测分数，很少足以认定违规。真正在调查中有分量的是证据组合：不同稿次之间风格的突变、显示整篇文章一次性粘贴出现的版本历史、并不存在的参考文献、你无法口头解释的专业术语，或者你的课堂表现在提交作品面前明显偏低。
The practical implication is uncomfortable but important: the strongest defence is not a cleverer paraphraser. It is being able to talk about your work, explain your choices, and reproduce your reasoning without assistance.
由此得出一个不太舒服但很重要的推论：最强的辩护不是更聪明的改写工具，而是你能够谈论自己的作品、解释自己的选择，并在没有辅助的情况下复现自己的推理。
10. Penalties and Outcomes | 处罚与后果
Sanctions exist on a scale that reflects intent, the value of the assessment and whether the behaviour was repeated. In low-stakes internal work, the outcome is often a conversation, a chance to redo the task, or a mark of zero for that piece. In formally assessed coursework, a centre may report suspected malpractice to the awarding body.
处罚呈阶梯式，与主观意图、评估的权重以及是否重复发生相关。在校内低风险任务中，结果通常是一次谈话、一次重做的机会，或该份作业记零分。在正式计分的课程作业中，学校可能会向考试局上报疑似违规。
| Level | 层级 | Typical consequence | 常见后果 | Typical trigger | 常见触发情形 |
|---|---|---|
| 1 | 一级 | Warning, resubmission, loss of homework mark | 警告、重做、作业分数作废 | First minor incident, admitted honestly | 首次轻微情形且主动承认 |
| 2 | 二级 | Zero for the component; formal record | 该模块记零分；正式记录在案 | AI text submitted in graded work | 在计分作业中提交 AI 文本 |
| 3 | 三级 | Loss of the qualification; reported to the board | 该科资格作废；上报考试局 | Coursework malpractice, concealment | 课程作业违规、刻意隐瞒 |
| 4 | 四级 | Disqualification from a series of exams; university consequences | 取消某一考季的考试资格；影响大学申请 | AI use in an invigilated exam, repeat offence | 监考考试中使用 AI、重复违规 |
University consequences deserve their own warning. Many admissions processes now ask about academic misconduct, and an offer can be withdrawn on the basis of a school report or a board decision discovered after the fact. The damage is rarely confined to one piece of homework.
大学层面的后果值得单独提醒。如今许多招生流程会询问学术违规记录，而一旦事后通过学校报告或考试局裁定被发现，预录取可能被撤销。损害很少只局限在某一份作业上。
11. Appeals and Evidence: Protecting Yourself | 申诉与证据：如何保护自己
If you are accused of AI misuse, the first principle is honesty. An admitted, minor, isolated incident is almost always treated more leniently than a denial that later collapses. The second principle is evidence: keep your process visible.
如果你被指控滥用 AI，第一原则是诚实。一次已承认、轻微且孤立的事件，几乎总比一个后来崩塌的否认被处理得更宽。第二原则是证据：让你的过程保持可见。
- Keep draft versions, notes, plans and reading notes with dates. | 保留带日期的草稿版本、笔记、提纲与阅读笔记。
- Keep your research trail: sources, quotations, page numbers, screenshots of relevant pages. | 保留研究轨迹：文献、引文、页码、相关页面截图。
- If AI was used with permission, keep the declaration and the teacher's approval message. | 若经允许使用 AI，保留声明与教师的批准信息。
- Ask for the specific rule and the specific evidence you are being assessed against. | 要求对方说明具体依据的规则与具体证据。
- Use the formal route: centre-level appeal first, then the awarding body's appeals procedure. | 走正式程序：先校内申诉，再考试局申诉程序。
During an interview or viva, answer questions in your own words and explain your reasoning step by step. If you genuinely do not understand a term in your own submitted work, say so and explain how it got there. Examiners are usually less interested in punishing confusion than in discovering who did the thinking.
在面谈或答辩中，用自己的话回答问题，并一步步解释你的推理。如果提交作品里有一个术语你确实不理解，坦白说明并解释它是怎么出现的。考官通常更在意的不是惩罚困惑，而是找出"到底是谁在思考"。
12. Practical Checklists and AI-Resilient Habits | 实用清单与抗 AI 依赖的习惯
For students, the aim is not to avoid AI forever, but to remain the author of your own learning. That means building habits in which AI accelerates understanding rather than replacing it. Write your first draft without AI. Ask AI to quiz you rather than to write for you. Explain a concept aloud before you ask for an explanation. Keep an AI-free timed practice session every week.
对学生的目标不是永远回避 AI，而是始终做自己学习成果的作者。这意味着要养成让 AI 加速理解、而非替代理解的习惯：第一稿不用 AI 完成；让 AI 考你而不是替你写；在索要解释之前先自己口头讲一遍某个概念；每周保留一次不用 AI 的限时练习。
For teachers and tutors, the parallel aim is task design that makes authorship visible: in-class writing, oral defence of submitted work, annotated bibliographies, staged deadlines with drafts, and explicit AI permissions stated per task rather than per year. A clear "AI allowed / AI allowed with declaration / AI not allowed" label on each assignment prevents most disputes before they start.
Published by TutorHao | Exam Prep Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图