IB内部评估 科学IA 数学反思 经济评价
IB 的六门学科都要交一份内部评估，但同样叫评价的那一节，在科学、数学、经济里分量完全不同。科学组的一份报告 24 分，评价占 6 分。数学探索 20 分，担当评价角色的反思只占 3 分。经济的三条评论加起来 45 分，每条评论里的评价占 4 分。这篇文章把三套评分准则摊开对照，说明各自要写什么，以及哪些句子会被判到低档。准则名称与分值均以 IBO 公布的科目指南为准。
Internal assessment looks identical across the Diploma Programme until you open the rubrics. In the sciences the evaluative criterion carries a quarter of the marks, in mathematics reflection carries three points out of twenty, and in economics every commentary has its own evaluation mark. Those differences decide what you write and where you spend your words.
内部评估由任课教师打分，再由 IBO 复核
IA 的第一道分数来自你的任课教师，教师按 IBO 公布的准则逐条给分。学校把分数录入 IBIS 之后，IBO 会抽取一部分作业交给校外评审员复核。抽样的份数按报考人数决定，五人以内全部寄出，六到二十人抽五份，二十一到四十人抽八份，四十人以上抽十份。评审员如果认为整所学校给分偏高或偏低，就会对全部考生的这一科分数做统一调整。调整不是逐份改分，而是按教师分数与评审员分数之间的差值整体移动。很多学生以为老师打完分就定了，其实最终分数由 IBO 在复核之后确定。错得最多的是把复核当成运气，却忘了它检验的是教师有没有按同一把尺子打分。
An IB teacher marks the work first against the published criteria, and the school then submits those marks through IBIS. The IB selects a sample, five to ten pieces depending on the size of the entry, and a moderator re-marks them with the same descriptors; where a school's marking sits consistently above or below the worldwide standard, the adjustment is applied to every candidate in that subject and level. Because the shift moves a cohort rather than individuals, a teacher who is generous with everyone costs the whole class, not only the sampled students.
科学组三个科目共用四个准则，评价独占 6 分
生物、化学、物理从 2025 年首次评估起改用同一份内部评估结构，任务叫科学探究。报告总分 24 分，研究设计、数据分析、结论、评价四个准则各 6 分。评价与研究设计同分，这一点让擅长做实验却不会收尾的学生吃亏。评价准则要的是对方法的判断，不是把实验步骤复述一遍。考生需要指出具体的方法弱点，说明它怎样影响数据，再给出可操作的改进。IBO 的 Biology Guide 把报告上限定在 3,000 字，图表、数据表、公式与参考文献不计入。字数不多，评价那一节通常只有四百到六百字的空间。
Biology, chemistry and physics now share a single internal assessment, the scientific investigation, first examined in 2025. It is marked out of 24, six marks each for research design, data analysis, conclusion and evaluation, and the report is capped at 3,000 words with tables, graphs, equations and references left out of the count. Evaluation carries the same weight as the design, so a well-executed experiment with a thin final section cannot reach the top band no matter how clean the data look.
科学 IA 的评价要排出影响大小，不是罗列缺点
评价准则的最高档要求解释方法弱点对结果的相对影响。相对两个字决定了写法：先比较，再排序。举例来说，如果总百分比不确定度是 3%，而你与文献值的偏差是 15%，那么随机误差解释不了这个差距，系统误差才是主要嫌疑。写到这里还要指名可能的来源，比如水浴温度漂移两度导致反应速率整体偏低。评审员看的是这条推理链，而不是你列了几条缺点。只写设备不够精确、操作不够熟练，通常停在低档。把弱点与自己数据里的具体数字挂钩，才有机会拿到五六分。
The top band asks for the relative impact of each weakness, which means ranking them rather than listing them. If your total percentage uncertainty is 3 per cent while your result sits 15 per cent away from the literature value, random error cannot explain the gap and something systematic is at work; naming the likely culprit, such as a water bath drifting two degrees, is the reasoning that earns the mark. Generic complaints about inaccurate equipment stay in the bottom band however long the list grows.
科学报告里最贵的四个字是人为误差
阅卷时最常见的减分句是「人为误差」。这四个字没有信息量，因为它既不说明误差来自哪里，也不说明它有多大。评价一节要区分随机误差与系统误差，前者影响精密度，后者影响准确度。随机误差可以通过重复测量与取平均压缩，系统误差不会因为多测几次而消失。常见的错误是把两者混在一起写，然后用「改进操作」一句话带过。有效的改进要能被别人照做，比如把秒表换成每 0.1 秒自动记录的传感器，或者把量筒换成移液管以减小读数不确定度。写清换什么器材、改哪一步，比写十句态度端正更有用。
Human error is the most expensive phrase in a science report because it says nothing about where the error came from or how large it was. Random error affects precision and shrinks when you repeat measurements; systematic error affects accuracy and survives every repeat, so the improvement you propose has to match the type you identified. Swapping a stopwatch for a sensor that logs every 0.1 seconds, or a measuring cylinder for a pipette, is an improvement another student could copy, while resolving to be more careful is not.
数学 AA 与 AI：20 分里反思只有 3 分
数学的分析与方法（AA）与应用与解释（AI）两门课都要求交一份数学探索，满分 20 分，占总成绩 20%。五个准则的分值并不平均：表达 4 分，数学沟通 4 分，个人投入 3 分，反思 3 分，数学运用 6 分。反思只有 3 分，比数学运用的一半还少。这不代表反思可以敷衍，因为它和表达、沟通一起决定整篇读起来是否像一个人在做数学。数学探索没有科学那样的 3,000 字上限，指南只要求简洁。简洁的意思是删掉重复计算与无关背景，而不是把推理压成结论。
Both mathematics courses require a single exploration worth 20 marks and 20 per cent of the final grade. Presentation and mathematical communication take four marks each, personal engagement and reflection three each, and use of mathematics six, so the reflective criterion is worth half of the mathematical one. That does not make reflection optional, because it is where a reader sees whether the mathematics belongs to the student or was lifted from a textbook.
数学探索的反思要回到模型本身
数学里的反思不是感想，而是对模型与方法的再检查。可用的写法有三种：说明结果为什么合理，指出模型在什么条件下失效，比较两种解法各自的代价。比如用指数模型拟合冷却曲线时，可以讨论当温差趋于零时模型预测的冷却速率是否还成立。再比如同一个积分用换元法与数值方法各算一次，比较误差与可解释性。这样的句子把反思钉在数学对象上，评审员能看出你知道自己在做什么。很多学生以为反思要写学习感受，结果写成了心得，3 分里只能拿 1 分。
Reflection in mathematics is a second look at the model, not a diary entry. Say why the result is reasonable, describe the conditions under which the model fails, and compare what two methods cost you in accuracy and in clarity; a cooling curve fitted with an exponential, for instance, invites the question of what happens as the temperature difference approaches zero. Reflections attached to the mathematics read as ownership, and the three marks follow.
经济 IA 是三条评论，评价在每条里占 4 分
经济的内评不是一篇长文，而是三条基于新闻的评论组合成的作品集。三条评论要取自大纲里不同的单元，微观、宏观、全球经济各一条，导论单元不能用来写。每条评论上限 800 字，评审员读到第八百个词就停。每条评论按五个准则打分，满分 14 分：图表 3 分，术语 2 分，应用 2 分，核心概念 3 分，评价 4 分。三条加起来 42 分，作品集整体再按格式要求给 3 分，总分 45。评价在每条评论里占 4 分，是单项里最重的一项，比图表还多 1 分。这份作品集在 SL 占 30%，在 HL 占 20%。
Economics asks for a portfolio of three commentaries on news extracts, one from each of the three main units, with every commentary capped at 800 words and moderators instructed to stop reading at the limit. Each commentary is marked out of 14, three for diagrams, two for terminology, two for application, three for the key concept and four for evaluation, and a further three marks cover the rubric requirements of the portfolio, giving 45 in total. Evaluation is the heaviest single criterion in each piece, which is why a commentary that describes a policy without ever weighing it stalls near half marks.
经济评论的评价要写取舍与受影响的人
经济里的评价不是复述文章观点，而是权衡。可用的落点是政策代价、时间维度、受益者与受损者的分布、以及替代方案。比如一则关于最低工资上调的新闻，评价可以讨论小企业成本上升与低薪工人收入改善之间的取舍，再指出通胀与就业弹性会改变结论。如果只写这项政策好或不好，就只是立场，不是评价。评审员的说明里写得很清楚：完全没有评价意图，或者只是把文章换个说法，这一项可以给 0 分。常见的错误是把结论段写成摘要，读起来完整，实际上一个判断都没有。想让 4 分里的高点成立，就要给出条件式的判断，比如在劳动需求弹性较小时效果更明显。
Evaluation in economics means weighing, not summarising. Take a rise in the minimum wage: the judgement lies in the trade-off between higher costs for small firms and better pay for low-paid workers, and in how the conclusion shifts once you allow for the elasticity of labour demand and for inflation. A commentary that calls the policy good or bad without naming who gains, who pays and under what conditions earns one mark out of four, not three.
三套准则对字数的态度完全不同
科学报告有硬上限，3,000 字之外的内容不读。经济评论更紧，每条 800 字，超出部分评审员直接跳过。数学探索反过来，没有字数规定，只有对简洁的要求。这个差别改变了写作策略：科学要把方法写成可复现，数学要把推导留足，经济要在 800 字里塞进图表、术语、应用、概念与评价五样东西。经济那一科的压缩最考验取舍，一条评论里塞两个政策工具，通常两边都写不深。字数限制不是排版问题，它决定了你能容纳几个论点。
The three subjects treat length in three different ways. Science stops at 3,000 words, economics stops at 800 per commentary and moderators simply do not read past that point, while mathematics sets no word count and asks only for concision. That changes what you can attempt: a commentary trying to analyse two policy instruments inside 800 words usually does neither well, whereas a science report with room to spare should spend it on method detail that makes the work reproducible.
同一句话在三个科目里的分数不一样
拿一句常见的收尾来对照：这个方法还有改进空间。在科学 IA 的评价里，这句话几乎不得分，因为没说改进什么、影响多大。在数学探索的反思里，它也只是最低档，除非你接着说明模型在哪个区间失效。在经济评论里，这句话甚至可能被算作没有评价意图。三个科目对同一个动作的要求一致：把判断落到具体的对象与条件上。空话不是错话，它只是没有内容。评分准则奖励的是内容，不是态度。
Consider one sentence: the method could be improved. In a science evaluation it scores almost nothing, because it names neither the weakness nor its effect; in a mathematics exploration it stays at the lowest reflection level unless you specify where the model breaks; in an economics commentary it may not count as evaluation at all. All three rubrics want the same move, which is to attach the judgement to a named object and a stated condition.
复核为什么会让全班分数一起移动
复核的作用是校准，不是抽查作业质量。评审员拿到的是几份被抽中的作业，看到的是教师给分与准则描述之间的偏差。偏差如果是普遍偏高，整所学校的这一科分数会被下调同样的幅度，个人写得再好也跟着动。这也是为什么同一篇 IA 在不同学校的得分率会有差别，标准在复核环节被拉回全球口径。学校的内部统一评分因此变得重要，多位教师教同一科时，先对齐样卷再打分。考生能做的部分有限，把准则逐条对上，比猜测评审员的口味可靠。
Moderation calibrates a school rather than inspecting individual scripts. A moderator sees a handful of sampled pieces, compares the teacher's marks with the descriptors, and looks for a pattern rather than a single disagreement. Where that pattern runs consistently high or low, the adjustment is applied to every candidate in that subject and level, which explains why identical work can score differently in two schools and why departments that standardise their marking first tend to move less at results.
交稿之前把准则当清单逐条对一遍
交稿之前，把准则打印出来，用铅笔在自己稿子上标出每一分来自哪一句。科学 IA 要检查四件事：研究问题是否点明了自变量与因变量，不确定度是否一路带到结论，结论是否与文献值比过，评价是否排过影响大小。数学探索要检查反思是否落在模型上，而不是落在感受上。经济评论要检查每条评论的核心概念是否不同，以及评价段是否有条件式判断。这一步花二十分钟，通常能改出三到五分。评审员不会替你补上没写的东西。
Before you submit, print the rubric and mark in pencil which sentence earns which mark. In the sciences, check that the research question names both variables, that uncertainties are carried into the conclusion, that the result is compared with a published value, and that the evaluation ranks its weaknesses; in mathematics, check that reflection sits on the model rather than on your feelings; in economics, check that the three key concepts differ and that each evaluation offers a conditional judgement. Twenty minutes of this usually finds three to five marks, because a moderator cannot award what is not written down.
牛津高阶英汉双解词典 第10版 商务印书馆¥168📱 去京东购买京东精选
A-Level / GCSE / IB / KS2-KS3 全科辅导
更多咨询请联系16621398022（同微信）
IB 拓展论文与内部评估的写作资料在 aleveler.com
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图