📚 Year 12 CCEA Computer Science: Essay Writing Framework and Sample Essay | Year 12 CCEA 计算机：论文写作框架与范文
In the CCEA AS Computer Science examination, extended writing questions carry substantial marks. These essays test your ability to apply knowledge, analyse societal impacts, and construct coherent arguments. Without a clear framework, even well-prepared students may lose marks due to poor structure or lack of evaluation. This article provides a step-by-step guide to building high-scoring essays, complete with a sample essay on the ethical dimensions of artificial intelligence.
在 CCEA AS 计算机科学考试中，扩展写作题占有较高分值。这些论文考查你运用知识、分析社会影响以及构建连贯论证的能力。没有清晰的框架，即便准备充分的学生也可能因结构混乱或缺乏评估而失分。本文提供高分论文分步指南，并附上一篇关于人工智能伦理维度的范文。
1. Understanding the CCEA Essay Requirements | 理解 CCEA 论文要求
CCEA's AS Computer Science Unit 1 (Computer Systems) often includes an extended response question in Section B. Candidates are asked to discuss a scenario involving legal, ethical, moral, or cultural issues, drawing on their understanding of computing concepts. The examiners look for accurate knowledge, application to the given context, and a sustained, balanced judgement. Typically the question is worth between 9 and 12 marks, so it deserves roughly 15 to 20 minutes of the paper.
CCEA AS 计算机科学第一单元（计算机系统）通常在 B 部分设置一道扩展写作题。考生需要围绕一个涉及法律、伦理、道德或文化问题的情境展开讨论，并运用自己的计算机概念知识。考官期待看到准确的知识、对给定情境的运用，以及持续且平衡的判断。这类题通常为 9 至 12 分，因此应在试卷中为它留出约 15 至 20 分钟。
A distinctive feature of CCEA extended responses is the assessment of Quality of Written Communication (QWC). Spelling, punctuation, grammar and the precise use of specialist terminology are all credited. This means that vague phrases such as 'computers are bad for people' score far less than precise statements such as 'algorithmic decision-making can entrench existing socioeconomic bias because training data reflects historical patterns of discrimination'.
CCEA 扩展写作题的一个显著特点是评估书面表达质量（QWC）。拼写、标点、语法以及专业术语的准确使用都会计分。这意味着像 "计算机对人有害" 这样模糊的表述，得分远低于 "算法决策会固化既有的社会经济偏见，因为训练数据反映了历史上的歧视模式" 这样精确的陈述。
2. Where Extended Writing Appears in the Specification | 扩展写作在考纲中的位置
Extended writing is not confined to one topic. It can be attached to almost any area of Unit 1: the social and ethical consequences of networking, the impact of legislation on software development, the environmental cost of data centres, or the cultural effects of globalised digital services. The scenario is usually contemporary, so examples drawn from news stories about social media, facial recognition, or autonomous vehicles are valuable.
扩展写作并不局限于某一个主题。它几乎可以挂靠在第一单元的任一领域：网络的社会与伦理后果、立法对软件开发的影响、数据中心的环境成本，或全球化数字服务的文化影响。题目情境通常具有时代性，因此引用社交媒体、人脸识别或自动驾驶汽车等新闻案例非常有价值。
Because the mark scheme rewards application, you must anchor every abstract point to the specific scenario in the question. If the scenario mentions a hospital trust, your discussion of data protection should refer to patient records, not to a generic 'organisation'. Generic answers are capped at the lower mark bands even when the computing knowledge is sound.
由于评分标准奖励 "运用"，你必须把每一个抽象论点都锚定在题目给出的具体情境上。如果情境提到某家医院信托机构，你关于数据保护的讨论就应指向患者病历，而不是泛泛的 "某个组织"。即便计算机知识正确，泛泛而谈的答案也只能停留在较低分数段。
3. Mark Scheme Anatomy: AO1, AO2 and AO3 | 评分标准剖析：AO1、AO2 与 AO3
CCEA's assessment objectives map neatly onto the three things an essay must do. AO1 rewards correct knowledge; AO2 rewards applying that knowledge to the scenario; AO3 rewards analysing consequences and evaluating competing positions to reach a reasoned conclusion. A high-scoring essay interleaves all three rather than separating them into a knowledge paragraph and an opinion paragraph.
CCEA 的评估目标恰好对应一篇论文必须做到的三件事。AO1 奖励正确的知识；AO2 奖励把知识运用于情境；AO3 奖励分析后果并评估相互竞争的观点，从而得出有理据的结论。高分论文会把三者交织在一起，而不是把它们拆成一个知识段和一个观点段。
| Assessment Objective | What it looks like in an essay | 中文说明 |
| AO1 Knowledge | Accurate definitions and concepts, e.g. 'GDPR requires a lawful basis for processing personal data'. | 准确的定义与概念，例如 "GDPR 要求处理个人数据必须有合法依据"。 |
| AO2 Application | Using that concept on the scenario, e.g. 'the trust must document its lawful basis for holding patient records'. | 把概念用于情境，例如 "该信托机构必须记录其保存患者病历的合法依据"。 |
| AO3 Analysis and Evaluation | Weighing consequences for stakeholders and reaching a justified judgement. | 权衡对各利益相关方的影响，并得出有据可依的判断。 |
4. The Five-Part Paragraph Framework | 五段式段落框架
The most reliable way to hit all three assessment objectives is a repeating paragraph structure. Each body paragraph should contain a Point, an Explanation, Evidence from the scenario, an Impact on stakeholders, and a Judgement. This structure forces you to move from describing to analysing within every paragraph, which is exactly what examiners reward.
要同时命中三项评估目标，最可靠的方法是采用可重复的段落结构。每个主体段都应包含：论点（Point）、解释（Explanation）、来自情境的证据（Evidence）、对利益相关方的影响（Impact），以及判断（Judgement）。这一结构迫使你在每一段内部从描述走向分析，而这正是考官所奖励的。
Body paragraph = Point + Explain + Evidence + Impact + Judgement
At the whole-essay level, the equivalent framework is: introduction (define the issue and state your line of argument), three or four body paragraphs (each using the five-part structure), and a conclusion that answers the question directly. Two paragraphs is rarely enough for 12 marks; five body paragraphs usually means the last one is rushed and unevaluated.
在整篇论文层面，对应的框架是：引言（界定问题并表明你的论证立场）、三到四个主体段（每段使用五段式结构），以及一个直接回应题目的结论。两段对 12 分的题目几乎不够；而五个主体段通常意味着最后一段写得仓促且缺乏评估。
5. Planning Under Timed Conditions | 限时条件下的规划
Allow three to four minutes for planning, even when the clock feels tight. A short plan prevents the two most expensive mistakes: repetition and missing evaluation. Write the scenario's key facts in the centre of your page, then branch out into legal, ethical, social, economic and cultural consequences, and finally select the three or four strongest branches.
即使时间紧张，也要留出三到四分钟规划。简短的规划能避免两个代价最高的错误：重复和漏掉评估。把情境的关键事实写在页面中央，然后向外分出法律、伦理、社会、经济与文化五个后果分支，最后从中选出最强的三到四个分支。
- Legal / 法律: legislation, regulation, compliance, liability — 立法、监管、合规、责任
- Ethical and moral / 伦理与道德: fairness, privacy, autonomy, honesty — 公平、隐私、自主、诚信
- Social and cultural / 社会与文化: access, inequality, employment, identity — 可及性、不平等、就业、身份认同
- Economic and environmental / 经济与环境: cost, efficiency, energy use, e-waste — 成本、效率、能耗、电子垃圾
6. Writing the Introduction and Conclusion | 引言与结论的写法
The introduction should do two things only: define the key term in the question and state the position you will defend. Do not retell the scenario at length, and do not list what you are 'going to talk about'. A single sentence such as 'Artificial intelligence can improve diagnostic accuracy, but its opacity creates accountability gaps that current regulation does not close' is worth more than three descriptive sentences.
引言只需做两件事：界定题目中的关键术语，并表明你将捍卫的立场。不要长篇复述情境，也不要罗列你 "打算谈论什么"。像 "人工智能可以提高诊断准确率，但其不透明性造成了现行监管无法填补的责任缺口" 这样一句话，胜过三句描述性的话。
The conclusion must give a direct answer with a reason. Use hedged but committed language: 'On balance, the benefits outweigh the risks for routine administrative tasks, but not for decisions affecting individual liberty, because the consequences of error are irreversible.' Avoid new evidence here, and avoid the empty phrase 'in conclusion, there are many advantages and disadvantages'.
结论必须给出直接答案并说明理由。使用有分寸但明确的措辞："总体而言，在常规行政任务中收益大于风险，但在影响个人自由的决策中并非如此，因为错误的后果不可逆。" 这里不要引入新证据，也要避免 "总之，有很多优点和缺点" 这种空洞表述。
7. Using Technical Vocabulary Precisely | 准确使用专业术语
QWC marks are earned by using subject terminology correctly, not by using a lot of it. A term used wrongly costs more than it gains. Check that you can distinguish machine learning (systems that improve from data) from artificial intelligence (the broader field), and that you know that a biased algorithm is usually a symptom of biased training data rather than of malicious code.
QWC 分数来自正确使用学科术语，而不是术语堆砌。用错一个术语的失分大于得分。请确认你能区分机器学习（从数据中改进的系统）与人工智能（更广泛的领域），并明白有偏见的算法通常是训练数据有偏的表现，而非恶意代码造成的。
| Term | Accurate use in a sentence | 中文对照 |
| Anonymisation | 'Anonymisation removes identifiers so that data subjects can no longer be identified.' | 匿名化删除标识符，使数据主体无法再被识别。 |
| Accountability | 'If no human can explain the decision, accountability cannot be assigned.' | 如果无人能解释该决策，责任就无法归属。 |
| Digital divide | 'Unequal broadband access widens the digital divide between urban and rural users.' | 宽带接入不均扩大了城乡用户之间的数字鸿沟。 |
| Automation | 'Automation displaces routine roles faster than retraining programmes can absorb them.' | 自动化取代常规岗位的速度快于再培训项目的吸纳能力。 |
8. Evaluation Language and Judgement | 评估语言与判断
Evaluation is the difference between a description and an argument. Use comparative connectives such as 'whereas', 'in contrast', and 'this matters more than ... because ...'. Use degree modifiers such as 'to a large extent', 'in most cases', and 'arguably'. Crucially, every evaluative claim needs a because-clause; a judgement without a reason scores nothing.
评估是描述与论证之间的分水岭。使用比较性连接词，如 "whereas"（然而）、"in contrast"（相比之下）、"this matters more than ... because ..."（这比……更重要，因为……）。使用程度修饰语，如 "to a large extent"（在很大程度上）、"in most cases"（在多数情况下）、"arguably"（可以说）。关键在于，每一条评估性论断都需要一个 because 从句；没有理由的判断不得分。
Another useful move is the counter-argument and rebuttal: state the strongest opposing view fairly, then explain why your position still holds. This demonstrates the balance that CCEA rewards and is far more convincing than dismissing the other side in a few words.
另一个有用的技巧是 "反论与反驳"：先公正地陈述最强的对立观点，再解释为何你的立场依然成立。这体现了 CCEA 所奖励的平衡性，也比只用几句话否定对方更有说服力。
9. Common Pitfalls That Cost Marks | 常见失分点
The most frequent weakness is drift into description: three paragraphs explaining how something works and only one sentence of judgement. The second most frequent is ignoring the scenario, producing an essay that could be attached to any question. Both are avoidable with the five-part paragraph structure described above.
最常见的弱点是从分析滑向描述：用三段解释某事物如何运作，却只有一句判断。第二常见的错误是忽视情境，写出一篇可以套用到任何题目上的文章。这两个问题都可以用上文介绍的五段式段落结构来避免。
- No judgement at the end of a paragraph or the essay — 段落或全文结尾缺少判断
- Unsupported assertions such as 'this is unethical' with no reason — 只说 "这不道德" 而不给理由
- Terminology errors, e.g. confusing encryption with hashing — 术语错误，如把加密与哈希混为一谈
- Poor time management leaving the conclusion unwritten — 时间管理不当，结论未写
10. Sample Essay: The Ethical Dimensions of Artificial Intelligence | 范文：人工智能的伦理维度
Question: 'Discuss the ethical issues raised by the increasing use of artificial intelligence in decision-making.' (12 marks)
题目："讨论人工智能在决策中日益广泛的使用所引发的伦理问题。"（12 分）
Introduction. Artificial intelligence refers to computer systems that perform tasks normally requiring human intelligence, such as pattern recognition and prediction. Where such systems make or inform decisions about people, ethical questions arise about fairness, accountability and privacy. This essay argues that the most serious ethical risk is not that machines will be cruel, but that they will be opaque, and that opacity in decisions affecting individuals' rights is unacceptable.
引言。人工智能指执行通常需要人类智能之任务（如模式识别与预测）的计算机系统。当此类系统对人类作出决策或为决策提供依据时，就会产生关于公平、责任与隐私的伦理问题。本文认为，最严重的伦理风险不是机器会变得残忍，而是机器不透明，而在影响个人权利的决策中，不透明是不可接受的。
Fairness and bias. Machine learning models learn patterns from historical data. Where that data encodes past discrimination, the model reproduces it at scale. A recruitment tool trained on a decade of hiring records may systematically downgrade female applicants without any programmer intending it. The ethical problem is not only the outcome but the illusion of objectivity: a numerical score appears neutral and is therefore harder to challenge than a human decision. This matters more than efficiency gains because it can entrench inequality across an entire organisation's processes.
公平与偏见。机器学习模型从历史数据中学习模式。当数据中编码了过去的歧视时，模型会大规模复制这种歧视。一个在十年招聘记录上训练的招聘工具，可能在程序员并未刻意为之的情况下系统性地压低女性申请人的评分。伦理问题不仅在于结果，还在于 "客观性" 的错觉：数字评分看起来中立，因此比人类决策更难被质疑。这比效率提升更重要，因为它可能在组织的全部流程中固化不平等。
Accountability and transparency. Many high-performing models are effectively black boxes: even their developers cannot fully explain a specific output. Where a decision deprives someone of a job, a loan or parole, the affected person has a legitimate interest in knowing the reason. GDPR-style rights to explanation exist precisely because unexplained adverse decisions remove the ability to appeal. The ethical position here is that explainability is not a technical luxury but a precondition for justice.
责任与透明。许多高性能模型实际上是黑箱：连开发者也无法完全解释某个具体输出。当一项决策剥夺了某人的工作、贷款或假释机会时，被影响者有权知道原因。类似 GDPR 所赋予的 "获得解释的权利" 之所以存在，正是因为无法解释的不利决策剥夺了申诉的可能。此处的伦理立场是：可解释性并非技术上的奢侈品，而是正义的前提条件。
Privacy and autonomy. AI systems improve by consuming data, which creates pressure to collect more of it. Continuous monitoring in workplaces and public spaces can shift behaviour even when no sanction follows, producing a chilling effect on lawful activity. There is also a moral question about consent: individuals rarely know which models process their data or what inferences are drawn.
隐私与自主。人工智能系统通过消耗数据来改进，这产生了收集更多数据的压力。工作场所与公共空间的持续监控即便没有伴随惩罚，也会改变行为，对合法活动产生寒蝉效应。此外还有关于同意的道德问题：个人往往不知道自己数据被哪些模型处理、又被推断出什么。
Conclusion. On balance, the ethical problems of AI in decision-making are serious but manageable, provided that three conditions are met: training data is audited for bias, decisions affecting rights remain explainable and appealable, and data collection is proportionate to a stated purpose. The decisive factor is not the sophistication of the technology but the willingness of institutions to accept scrutiny. Where that scrutiny is absent, the risks clearly outweigh the benefits.
结论。总体而言，人工智能在决策中的伦理问题虽严重但可管理，前提是满足三个条件：训练数据经过偏见审计；影响权利的决策保持可解释且可申诉；数据收集与既定目的相称。决定性因素不是技术的先进程度，而是机构是否愿意接受审查。在缺乏审查的情况下，风险显然大于收益。
11. Annotated Commentary on the Sample Essay | 范文点评
The introduction earns AO1 by defining artificial intelligence accurately and earns AO3 by stating a contestable thesis. Notice that it does not waste words on the scenario or on a shopping list of topics. Examiners can identify the candidate's line of argument within two sentences, which makes the rest of the essay easy to credit.
引言通过准确定义人工智能拿到 AO1，通过提出可争议的中心论点拿到 AO3。注意它没有在情境复述或话题清单上浪费笔墨。考官能在两句话内识别出考生的论证立场，这使后文更容易得分。
Each body paragraph follows the same five-part pattern and contains at least one comparative evaluative sentence with a because-clause, for example 'this matters more than efficiency gains because it can entrench inequality'. The conclusion reaches a qualified judgement with three conditions, which is stronger than a one-sided verdict because it acknowledges complexity while still answering the question.
每个主体段都遵循相同的五段式模式，并至少包含一句带 because 从句的比较性评估，例如 "这比效率提升更重要，因为它可能固化不平等"。结论给出了带有三个条件的有限判断，这比一边倒的裁决更有力，因为它在承认复杂性的同时仍回应了题目。
What would lose marks here? Omitting the accountability paragraph and describing instead how neural networks are trained; giving no judgement at the end of paragraphs; or claiming that AI is 'simply unethical' without distinguishing between contexts such as recruitment and medical diagnosis.
哪些做法会失分？省略责任那一段，转而描述神经网络如何训练；段落结尾不作判断；或断言人工智能 "就是不道德的"，而不区分招聘与医疗诊断等不同情境。
12. Revision Plan and Exam-Day Checklist | 复习计划与考场清单
Build a bank of six or seven flexible scenarios with prepared arguments: facial recognition in public spaces, autonomous vehicles, algorithmic recruitment, medical diagnosis, predictive policing, social media recommendation, and remote examination proctoring. For each, note two benefits, three risks, one relevant law or regulation, and one stakeholder group.
建立一个由六到七个可灵活套用的情境组成的素材库，并预先准备论点：公共空间人脸识别、自动驾驶汽车、算法招聘、医疗诊断、预测性警务、社交媒体推荐、远程考试监考。为每个情境记录两点好处、三点风险、一条相关法律法规，以及一个利益相关方群体。
- Plan for 3-4 minutes; never start writing immediately — 规划 3 至 4 分钟，切勿立即动笔
- Signpost your argument in the introduction — 在引言中明示你的论证方向
- Use the five-part paragraph structure consistently — 一致使用五段式段落结构
- Include at least one legal reference and one stakeholder impact — 至少包含一条法律引用与一个利益相关方影响
- Reserve 2-3 minutes to write a decisive conclusion — 留出 2 至 3 分钟写出明确的结论
- Proof-read for terminology and spelling, since QWC is credited — 检查术语与拼写，因为 QWC 计入分值
Practise under timed conditions at least twice before the exam, and mark your own work against the three assessment objectives. If every paragraph contains a judgement, you are writing at the top of the mark range; if judgements appear only in the final sentence of the essay, rewrite and try again.
考前至少进行两次限时练习，并对照三项评估目标自评。如果每一段都包含判断，你就处在最高分数段；如果判断只出现在全文最后一句，那就重写再来一次。
Published by TutorHao | Year 12 Computer Science 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