📚 AS CAIE Computer Science: Exam Techniques and Marking Criteria | AS CAIE 计算机：答题技巧与评分标准
Achieving a high grade in AS Computer Science (9618) requires more than just factual knowledge - you must understand how the examiners award marks and how to structure your answers accordingly. This guide breaks down the CAIE marking criteria, typical command words, and effective techniques for every question type you will face in Papers 1 and 2.
要在 AS 计算机科学 (9618) 中取得高分，不仅需要掌握知识，还必须理解考官如何给分、如何按标准组织答案。本文详细解析 CAIE 评分原则、常见指令词以及应对试卷一和试卷二各类题型的有效技巧。
1. Understanding the CAIE Mark Scheme | 理解 CAIE 评分方案
CAIE mark schemes use a clear three-letter system: B marks are independent marks for a correct statement, M marks are awarded for a correct method even if the final answer is wrong, and A marks are given for an accuracy-dependent answer that follows a correct method. A simple example is in a number conversion question: the working may earn M1, while the final correct denary value gets A1.
CAIE 的评分方案采用明确的三字母系统：B 分是为独立正确的陈述给出的分数，M 分是授予正确方法（即使最终答案错误）的分数，A 分则是在正确方法基础上给出准确答案的分数。例如在进制转换题中，计算过程可获得 M1，最终正确的十进制值得到 A1。
| Mark type | Meaning | Example |
| B | Independent correct statement, no method needed | State one role of the Program Counter |
| M | Correct method, awarded even if the final answer is wrong | Showing the division steps in a conversion |
| A | Accurate answer, usually dependent on the M mark | Final 8-bit binary value |
Understanding how B, M and A are allocated helps you prioritise showing clear workings. Never skip steps - method marks are often easier to gain than the final answer mark. The mark scheme is written before the exam is sat, so each question has a fixed, finite list of acceptable answers.
理解 B、M、A 的分配方式，能帮助你优先写出清晰的解题过程。绝不要跳步——方法分往往比最终答案分更容易拿到。评分方案在考试之前就已写好，因此每道题都有一份固定且有限的"可接受答案"清单。
2. Command Words and What They Demand | 指令词及其要求
Every question begins with a command word that tells you exactly how much detail is required. If you write a paragraph where one word would do, you waste time; if you write one word where an explanation is needed, you lose marks.
每道题都以一个指令词开头，它明确告诉你需要写多少细节。如果一个词就能答完的题你写了一整段，就是浪费时间；如果需要解释的题你只写一个词，就会丢分。
- State / Give / Identify / Define: one short phrase, no justification. English: a one-line answer. 中文：一两句短语，不需要论证。
- Describe: say what happens, point by point. 中文：逐点说明"发生了什么"。
- Explain: say why or how, using cause and effect. 中文：用因果关系说明"为什么"或"如何"。
- Compare: give both similarities and differences. 中文：既写相同点，也写不同点。
- Discuss / Evaluate / Justify: give arguments for and against, then reach a conclusion. 中文：正反两面论证，并给出结论。
- Write / Complete / Trace / Convert / Calculate: produce code, output or a value. 中文：写出代码、输出结果或数值。
A reliable rule: the number of marks equals the number of distinct points the examiner expects. A three-mark explain question needs three separate reasons, not one long sentence.
一条可靠的规律是：分值等于考官期望的不同要点数量。一道 3 分的 Explain 题需要三个各自独立的理由，而不是一个长句子。
3. Timing and Mark Allocation in Papers 1 and 2 | 试卷一与试卷二的时间与分值分配
Paper 1 (Computer Principles) lasts 1 hour 30 minutes and is worth 75 marks. Paper 2 (Problem Solving and Programming) lasts 2 hours and is also worth 75 marks. In both cases that works out at roughly 1.2 minutes per mark, so a 5-mark question deserves about 6 minutes.
试卷一（计算机原理）时长 1 小时 30 分钟，满分 75 分；试卷二（问题求解与编程）时长 2 小时，同样满分 75 分。两者换算下来大约每分 1.2 分钟，因此一道 5 分题应分配约 6 分钟。
Do not treat the paper as one continuous block. Work in passes: first sweep through and answer everything you can do confidently, then return to the harder items with the time left. Long programming questions in Paper 2 should be started no later than 60 minutes from the end.
不要把整份试卷当成一个连续的整体。采用多轮作答法：第一轮先把有把握的题全部做完，再用剩余时间回头处理难题。试卷二中分值较高的编程大题，最迟应在距考试结束 60 分钟时开始动手。
4. Paper 1 Techniques: Principles Topics | 试卷一技巧：原理类题目
Paper 1 covers data representation, communication and networking, hardware, processor fundamentals, system software, security and ethics. The questions are short and factual, which means precision matters more than volume.
试卷一涵盖数据表示、通信与网络、硬件、处理器基础、系统软件、安全与伦理。题目短小、偏事实性，这意味着"精准"比"写得多"更重要。
When asked to describe a process, such as the fetch-decode-execute cycle, write it as an ordered sequence with the register names. When asked to state a benefit, do not add an explanation that might contradict what you wrote earlier.
当题目要求描述某个过程（例如取指-译码-执行周期）时，请按顺序写出步骤并带上寄存器名称。当题目只要求"陈述一个优点"时，不要额外添加可能与你前面所写内容相矛盾的说明。
- Name registers exactly: PC, MAR, MDR, CIR, ACC, ALU. 中文：寄存器名称必须准确，如 PC、MAR、MDR、CIR、ACC、ALU。
- Use the vocabulary from the syllabus, not casual synonyms. 中文：使用大纲中的术语，不要用日常口语化的同义词。
- If a mark is for a definition, do not give an example instead. 中文：如果该分是给定义的，就不要用举例替代。
5. Paper 2 Techniques: Algorithms and Programming | 试卷二技巧：算法与编程
Paper 2 rewards structure. Algorithm questions usually ask you to design a solution, complete a trace table, or write pseudocode. Examiners follow your logic line by line, so keep control flow visible.
试卷二看重结构。算法题通常要求你设计解法、补全跟踪表或编写伪代码。考官会逐行跟随你的逻辑，因此必须让控制流程清晰可见。
For a design question, always declare your variables and constants before using them, give each one a sensible name, and state the data type. For a completion question, check carefully what has already been given - you must continue the existing style rather than rewrite it.
对于设计类题目，务必先声明变量与常量再使用，为每个变量取一个合理的名称，并注明数据类型。对于补全类题目，要仔细看清题目已给出什么——你必须沿用现有写法，而不是重写一遍。
Use the standard CAIE pseudocode syntax consistently:
请一致地使用 CAIE 标准伪代码语法：
- Assignment: Count ← Count + 1 中文：赋值语句使用左箭头。
- Selection: IF ... THEN ... ELSE ... ENDIF 中文：分支语句要以 ENDIF 结束。
- Iteration: FOR i ← 1 TO 10 ... NEXT i, WHILE ... ENDWHILE, REPEAT ... UNTIL 中文：三种循环结构必须写全结束关键字。
- Arrays: DECLARE Marks : ARRAY[1:30] OF INTEGER 中文：数组声明必须带上下界与数据类型。
- Subroutines: PROCEDURE / FUNCTION with ENDPROCEDURE / ENDFUNCTION 中文：子程序必须写出结束标识。
6. Data Representation Questions | 数据表示类题目
Conversion questions are pure method marks. Always write the working: for binary to denary, show the place values and the addition; for denary to binary, show the repeated division or subtraction by place values. Never place a bare answer on the line.
进制转换题是纯粹的方法分题。务必写出过程：二进制转十进制时，写出各位权值及求和；十进制转二进制时，写出反复除以 2 的过程或按权值相减的过程。绝不要只写一个光秃秃的答案。
1 byte = 8 bits = 2⁸ = 256 possible values
Denary range of an 8-bit unsigned integer: 0 to 255
Also remember that two's complement negative values run from −128 to +127, that a hex digit maps to exactly 4 bits, and that a sign-and-magnitude representation has two zeros. These small facts appear repeatedly as one-mark items.
同时要记住：二进制补码负数范围是 −128 到 +127；一个十六进制位正好对应 4 个二进制位；原码表示中存在两个零。这些细小事实会反复以 1 分题的形式出现。
7. Trace Tables and Algorithm Tracing | 跟踪表与算法跟踪
Trace tables are marked generously but only if the column headings are right. First identify every variable and every output column, then work through the algorithm one line at a time, updating values in the correct order.
跟踪表的给分相对宽松，但前提是表头正确。先列出所有变量列和输出列，然后逐行推演算法，按正确顺序更新数值。
Common traps to watch for:
需要警惕的常见陷阱：
- Updating a variable before the line that changes it. 中文：在改变变量的那一行之前就更新了它的值。
- Forgetting that REPEAT ... UNTIL always runs at least once. 中文：忘记 REPEAT ... UNTIL 至少会执行一次。
- Missing the final output row after the loop ends. 中文：漏掉循环结束后的最终输出行。
- Confusing integer division DIV with remainder MOD. 中文：混淆整除 DIV 与取余 MOD。
If the algorithm contains an array, add a separate small table for the array contents, or annotate each cell clearly. An examiner cannot award a mark for a value they cannot find.
如果算法中涉及数组，请为数组内容单独画一个小表，或清晰标注每个格子。考官无法为一个"找不到"的数值给分。
8. Structured Answers for Explain and Describe Questions | Explain 与 Describe 题的结构化作答
The fastest way to secure full marks on a multi-mark explanation is to plan the number of points before writing. Draw a tiny tick list in the margin: three marks means three bullets.
在多分解释题上拿满分的快捷方法，是动笔前先规划要写几个要点。在页边画一个小小的勾选清单：3 分就是 3 个要点。
Each point should contain a technical term plus its consequence. For example, rather than writing 'RAM is faster', write 'RAM has a shorter access time than secondary storage, so frequently used instructions can be fetched quickly by the CPU'.
每个要点都应包含一个专业术语加上它的后果。例如，不要只写"RAM 更快"，而应写"RAM 的访问时间比二级存储更短，因此 CPU 可以快速取用频繁使用的指令"。
Connectives that examiners like include: because, therefore, which means that, as a result, whereas. Each of these signals a developed point rather than a repetition.
考官偏好的连接词包括：because、therefore、which means that、as a result、whereas。每一个这样的词都表明这是一个展开的要点，而不是重复表述。
9. Comparison, Discussion and Evaluation Questions | 比较、讨论与评价类题目
Higher-tariff questions in Paper 1 often ask you to discuss a technology or evaluate an approach. These carry 4 to 6 marks and are marked using a levels-of-response style rubric, where a coherent, balanced argument scores more than a list of loose facts.
试卷一的高分题常要求讨论某种技术或评价某种方案，通常 4 到 6 分，采用"分级评分"（levels of response）标准：结构连贯、观点平衡的论述比零散事实的罗列得分更高。
A safe structure is:
一个稳妥的结构是：
- Point 1 - advantage, with a reason. 中文：第一个优点，并说明理由。
- Point 2 - a second advantage or deeper detail. 中文：第二个优点或更深入的细节。
- Point 3 - a disadvantage or limitation. 中文：一个缺点或局限。
- Point 4 - a short judgement that answers the question directly. 中文：用一句简短的判断直接回答问题。
Never sit on the fence. If the question says 'justify', you must commit to a position and defend it, even if both sides have merit.
绝不要模棱两可。如果题目说 justify（论证），你就必须明确站定一个立场并为其辩护，即使双方都有道理。
10. Programming Question Checklist | 编程大题检查清单
Longer programming questions in Paper 2 are marked against a bespoke scheme that usually awards marks for input, validation, processing, output, and correct use of constructs. You do not need a perfect program - you need to hit as many of those checkpoints as possible.
试卷二的分值较高的编程题采用专门制定的评分方案，通常对输入、验证、处理、输出以及结构正确使用分别给分。你不需要写出完美的程序——你需要在尽可能多的评分点上得分。
- Read the question twice and underline every required feature. 中文：读两遍题，把所有要求的功能划出来。
- Declare variables with types before use. 中文：使用前先声明变量及其类型。
- Include validation if the question mentions valid data or ranges. 中文：只要题目提到有效数据或范围，就写验证。
- Produce output with clear labels and units. 中文：输出要带清晰的标签和单位。
- Add brief comments only where they clarify logic. 中文：只在能说明逻辑的地方加简短注释。
If you run out of time, leave the structure visible: written input, a loop header and a closing statement can still earn marks even if the middle section is incomplete.
如果时间不够，也要让结构可见：写出的输入语句、循环头与结束语句，即使中间部分未完成，仍可能拿到分数。
11. Common Mistakes That Cost Easy Marks | 常见的"送分"失误
Most lost marks in AS Computer Science are not caused by ignorance but by poor exam technique. Removing these habits alone can raise a grade.
AS 计算机科学中丢掉的分数，大多不是源于不会，而是源于答题技巧不佳。仅改掉以下习惯就能提升一个等级。
- Writing more than the question asks, then contradicting yourself. 中文：写得比题目要求的多，结果自相矛盾。
- Giving an example when a definition was required. 中文：题目要定义，却举了个例子。
- Leaving calculation working out of the answer booklet. 中文：把计算过程留在草稿纸上，没写进答题本。
- Using indentation-free pseudocode that hides the loop structure. 中文：伪代码不缩进，导致循环结构看不清。
- Ignoring the last line of the question, which often carries a separate mark. 中文：忽略题目最后一句，而它常常单独占一分。
- Spending 15 minutes on a 2-mark question at the start. 中文：一开场就在一道 2 分题上花掉 15 分钟。
Before the exam ends, spend five minutes checking units, variable names, spelling of technical terms, and whether every part of every question has been attempted.
考试结束前，用五分钟检查单位、变量名、专业术语拼写，以及每道题的每个小问是否都已作答。
12. Revision Strategy That Matches the Mark Scheme | 与评分标准对齐的复习策略
Effective revision for 9618 is not rereading the textbook - it is practising with real mark schemes in front of you. After each past paper, mark your own work strictly, then compare your wording with the published answers.
9618 的高效复习不是重读课本，而是把真实评分方案摆在面前进行练习。每做完一套真题，严格自评，然后把自己的措辞与官方答案逐条比对。
Build a personal error log with three columns: the topic, the command word you misread, and the exact wording the mark scheme accepted. Reviewing twenty of these entries before an exam is worth more than reading two more chapters.
建立一份个人错题本，分三列：知识点、你误读的指令词、评分方案实际接受的措辞。考前回顾二十条这样的记录，比再读两章书更有价值。
Finally, practise under timed conditions at least four times. You need to know not only what you know, but how quickly you can write it.
最后，至少进行四次限时模拟训练。你不仅要清楚自己会什么，还要清楚自己能多快把它写出来。
Published by TutorHao | AS 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