IB IA 研究问题 不确定性 评价 结论 24分
IB 的 Internal Assessment（IA）由任课老师评分，再交给 IBO 做外部调整（moderation），它在最终成绩里的比重从 20% 到 30% 不等：科学组与数学是 20%，Economics 在 SL 是 30%、HL 是 20%，History 在 SL 是 25%、HL 是 20%。篇幅上限写在各科的 subject guide 里，科学组报告不超过 3,000 词，数学 exploration 是 12 到 20 页，History 的 historical investigation 是 2,200 词，Economics 每篇 commentary 不超过 800 词。评分标准的名字各不相同，从研究问题到数据、分析、评价、结论这条主线在所有科目里都成立。下面按写作顺序拆成八个环节，每一节给一个通用写法、一个各科差异和一个当下就能做的检查动作。
The internal assessment is marked by your teacher and then moderated externally by the IB, and its weight runs from 20% to 30% of the final grade: sciences and mathematics carry 20%, Economics counts 30% at SL against 20% at HL, and History counts 25% at SL against 20% at HL. Each subject guide sets the length, with science reports capped at 3,000 words, mathematics explorations at 12 to 20 pages, the History investigation at 2,200 words and each Economics commentary at 800 words. Criterion names differ from subject to subject, yet the same path runs through all of them, and the eight stages below each arrive with one check you can run straight away.
把兴趣缩成一个可测的研究问题
研究问题是整篇 IA 的地基，评分标准从第一条读到最后一条，都在回头看它有没有被回答。一个能用的研究问题通常满足三个条件：变量可以测量、工作量能在一到两周内完成、结果有可能推翻你自己的猜想。很多学生以为题目写得大一点显得有水平，实际上 how does temperature affect enzyme activity 这种问题的得分上限很低，因为温度范围、实验对象和测量方式全都没有限定，别人无法判断你的数据够不够用。收窄的写法是给出具体对象、范围和测量方式，例如把马铃薯组织里过氧化氢酶的浓度与 60 秒内产生的气体体积对应起来，浓度取 0 到 5 个百分点，水浴温度固定在 20 摄氏度。这样的句子自带自变量、因变量、单位与时间窗，方法部分几乎不用再解释一遍。常见的错误是把研究问题写成一句描述现象的话，句子以句号收尾而不是问号，评分老师读不出你要测什么。写完之后把问题念给同学听，如果他还要追问你到底测哪个量，就说明范围还没收窄。Economics 的路径不一样，问题来自你自己挑的那篇新闻，文章必须在动笔前一年内发表。
The research question sits under every criterion in the report, so a vague one limits the mark long before the experiment starts. Three tests are worth applying: the variables can be measured, the work fits into one or two weeks, and the outcome could genuinely surprise you. A question such as how does temperature affect enzyme activity fails the first test, because the range, the tissue and the measurement method are all unspecified. Narrowing means naming them, as in catalase in potato tissue read from gas volume over 60 seconds across 0 to 5 percent hydrogen peroxide held at 20 degrees Celsius, which leaves little for the method section to explain. Economics reverses the order: the question grows out of a news article published within twelve months of writing and has to sit inside one syllabus section.
引言与背景：先给理论再谈操作
引言不写学科史，也不写堆砌的赞美句，它要做的是把读者带到你的变量关系上。通用结构分三层：这个现象在学科理论里怎么解释，前人的工作给出了什么量级的数据，你的研究问题在这个背景下补哪一格。生物与化学要写出反应机理或生理过程的定性说明，物理要写出公式与它的适用条件，数学要用符号定义每一个变量并在第一次出现时给出取值范围。数学 exploration 的评分标准里，Presentation 占 4 分、Mathematical communication 占 4 分，这两块的分几乎全部从引言与方法段落里读出来，符号在正文、表格与图表中必须是同一个写法。很多学生以为背景写得越长越显得用功，实际上评分标准看的是相关性，与 aim 无关的两段背景会把 Presentation 拉下来。检查动作很直接：给引言里每一句话标一个指向，说明它对应研究问题的哪一部分，指不上的句子删掉即可。
The introduction has one job, which is to put the reader in front of your variables. Set the phenomenon in the theory of the subject, quote the order of magnitude earlier work has established, then say which gap your question fills; biology and chemistry need the qualitative mechanism, physics needs the formula and its conditions of use, and mathematics needs every symbol defined with its range at first mention. In the mathematical exploration, presentation and mathematical communication are worth four marks each and are read straight off these opening paragraphs, so one symbol has to survive from the text into every table and graph. Background that never touches the aim pulls the presentation mark down, and the test is mechanical: each introductory sentence should point somewhere in the research question, while anything that points nowhere can be deleted.
方法设计：让别人照着做一遍
方法部分的可复现性是科学组 Research design 那 6 分的依据。IBO 的 Biology、Chemistry、Physics 指南在 2023 年 9 月起施教、2025 年首次评估的版本里，把科学组 IA 改成四条标准：Research design、Data analysis、Conclusion、Evaluation，每条 6 分合计 24 分，占最终成绩 20%，报告上限 3,000 词。旧版里的 Personal engagement 与 Communication 两条标准已经并入这四条，表达是否清楚在每个指标里分别计分。可复现的意思是：读你报告的同学能买到同样的材料、配出同样的浓度、在同样的时刻读数。需要写清楚的量包括仪器型号与精度、每次读数的时刻、控制变量的手段、重复次数，评分时通常期望自变量至少 5 个取值、每个取值约 5 次重复。错得最多的是只写 I measured the rate，却没有说明速率是怎么从原始读数算出来的。涉及人体反应的实验要写知情同意与安全处理，用电压或腐蚀性药品时写明防护措施。把方法交给没做过这个实验的同学，让他口述一遍操作，他卡住的位置就是缺掉的细节。
Research design is where reproduction lives, and the science guides first assessed in 2025 mark the report against four criteria of six marks each, with communication folded into all of them instead of carrying its own score. Reproducible means a classmate can buy the same materials, mix the same concentrations and read the instrument at the same moments, so instrument precision, timing, controls and repeat counts all belong on the page, and examiners generally expect at least five values of the independent variable with around five repeats at each. Writing I measured the rate without showing how the rate came out of the raw readings is the most common gap. Work with human volunteers needs consent and safety notes, and the test is to hand the method to someone who has never done the experiment and listen for where their retelling stalls.
原始数据、不确定性与有效数字
数据处理这条标准在各科都占大约 6 分，考的是记录、处理与呈现三件事是否都清楚且精确。原始数据表要带单位、仪器精度与重复次数，处理后的表要写出所用的公式或软件，图表要有标题、标好的坐标轴，并在需要时加误差棒。不确定性的处理最容易丢分：单次读数用仪器最小刻度的二分之一或标称精度，重复读数用极差或标准差，再做传递，最后把结果写成 1.24 ± 0.05 mol dm-3 这样的形式。有效数字跟着最不确定的那个量走，中途多留一位、最后一步才修约。数学 exploration 的 B 分同样要求符号与精度全篇一致，把圆周率 3.1416 和 π 混着用会被读成不一致。常见的错误是把误差只写在结论里，或者用 0.1 克的称量精度去报告 0.001 克的结果。任选表格里的一行，从原始读数一路算到最后结果，中间每一步都写出公式与单位，缺哪一步一眼就能看出来。
Recording, processing and presenting data carry their own marks in every subject, and uncertainty is where most of them disappear. Raw tables need units, instrument precision and repeat counts; processed tables need the equation or software behind them; graphs need titles, labelled axes and error bars where they apply. Half the smallest scale division, or the stated instrument accuracy, serves a single reading, while repeat readings are usually handled with range or standard deviation and then combined through the formula, ending as a value plus or minus an uncertainty in the same unit. Significant figures follow the least certain quantity, with one spare digit kept during the work and rounding left to the end, and mathematics explorations penalise notation that shifts halfway down a page. Pick any row of your table and recompute it to the final answer in writing, formula by formula, to see which step is missing.
分析：把结果接回理论
分析不是把表格再念一遍，而是说明数据与理论之间的关系。写法上先给出定量结论，例如斜率对应的活化能、拟合曲线的决定系数，或者两组数据的差异是否超过不确定度范围，然后解释这个数字在学科上意味着什么。物理常问 your value compared with the accepted value，答题时要把两个数的差写成百分比并指出偏大还是偏小。数学 exploration 的 E 分（Use of mathematics，6 分）看的是数学是否与课程水平相当、每一步是否被解释过，把计算器上的结果直接抄进正文拿不到这一分。很多学生以为数据点越多分析越强，实际上五个点画出的趋势如果没有理论解释，分析依然是空的。常见的错误是发现数据不合预期就悄悄删掉某个点，评分老师在重复次数的表格里能看出这个缺口。给每张图表写一句结论，句子里必须有一个数字和一个理论名词。
Analysis is not a second reading of the table; it states what the numbers mean in the language of the subject. Lead with the quantitative result, whether that is an activation energy from a gradient, a coefficient of determination from a fitted curve, or a difference between two groups weighed against their uncertainty, and then interpret it. Physics questions regularly ask how your value compares with the accepted one, which needs a percentage difference and its direction, and the six-mark use of mathematics criterion rewards mathematics at the level of the course with each step explained rather than results copied off a calculator screen; saying why the model was chosen earns more than a long calculation. Dropping an outlier that spoils the trend is easy to spot in a table with repeats, and one sentence under every graph should contain a number and a named concept.
评价：局限、改进与延伸
评价那 6 分几乎全部来自具体性。可写的局限有系统误差来源、量程选择不当、样本量偏小、外部条件未受控、理论假设被破坏这几类，每一条都要接一个改进方案，方案要落在下次能真实执行的范围内。写 use more accurate equipment 会被判为空泛，写成把 0.1 克的台秤换成 0.001 克的电子天平、每个浓度点的重复从 2 次提高到 5 次，才算到位。延伸方向通常是改变自变量范围、换一种材料，或者用另一个模型验证同一组数据。历史与经济的评价更看重判断：Economics 的 Evaluation 一栏 3 分要求给出有理由的结论，罗列正反两面拿不满。常见的错误是把改进写成检讨，只承认做得不好却不给出动作。每条局限后面跟一个能在下次实验里落地的动作，这一栏的分就稳了。
Six marks in evaluation are bought almost entirely with specificity. Name the source of systematic error, the badly chosen range, the small sample, the uncontrolled external condition or the assumption the method broke, and pair each one with a change you could actually carry out next time. The phrase use more accurate equipment reads as filler, while swapping a 0.1 g balance for a 0.001 g one and raising repeats from two to five per concentration is a real plan. Extensions usually stretch the independent variable, change the material or test the same data against another model. Judgement matters as much as listing in Economics, where the evaluation criterion expects a reasoned verdict rather than two balanced paragraphs, and turning this section into an apology gives away marks.
结论、字数与引用规范
结论直接回答研究问题，给出数值结果与不确定度，并说明它在多大范围内支持你的假设。三到五句通常够用，不需要复述方法，也不该出现前面没有出现过的数字。字数与格式是白送的分：Economics 每篇 commentary 上限 800 词，三篇合计 2,400 词，45 分里有 3 分专门给 rubric 要求，包括三篇必须落在不同的大纲单元、每次使用不同的 key concept、引用一年内发表的文章；History 的 investigation 上限 2,200 词；科学组报告上限 3,000 词，表格、图表与计算过程不计入。引用统一用同一种格式，正文里的每一个外部数字都要有出处，图表借自外部资料时标注来源。提交前把页面与文件格式按学校要求固定下来，超出的部分放进附录，不要留到最后一晚改排版。IBO 在每个科目的 subject guide 里都列出了每条标准的分数与层级描述，对着改比反复润色句子有用。
Conclusions answer the question directly with a value, its uncertainty and the range over which the hypothesis survives; three to five sentences is usually enough, with no method recap and no numbers that never appeared earlier. Word limits are free marks that students hand back. Economics caps each commentary at 800 words and the portfolio at 2,400, and three of its 45 marks reward nothing but rubric compliance, including three different syllabus sections, a different key concept every time and articles published within the previous year; History allows 2,200 words, and science reports stop at 3,000 words with tables, graphs and calculations excluded. Citations follow one style throughout, every borrowed figure needs a source, and layout should be settled before the final night rather than during it, since the IB publishes every criterion and its descriptors for you to work down.
提交前的六步检查
提交流程做成一张清单，按顺序走完一遍通常能救回 2 到 4 分。先把研究问题、结论和每一张图表摆在一起，确认图表回答的正是问题本身。再检查每个变量是否在第一次出现时定义过，图表的横纵轴标签有没有说明。接着从任意一行数据反向验算，看公式、单位与有效数字是否对得上。数一遍引用条数，确认正文里的每个外部数字都能找到出处，图片与表格也有来源。核对字数与页数，超出的内容放进附录。换一个人读一遍，请他用自己的话复述你的 aim 与结论。这六步花不了一小时，逐条对着 subject guide 里的评分标准改，比再润色一晚上句子更划算。
A submission pass written as a checklist usually recovers two to four marks. Line up the research question, the conclusion and every graph together and check that the graphs answer the question itself, then confirm each variable is defined where it first appears; recompute one row of data backwards to see whether formulas, units and significant figures still agree. Count citations so every borrowed number and image can be traced, check word and page limits with the overflow parked in an appendix, and ask someone else to summarise your aim and conclusion in their own words. The whole pass takes under an hour, and working down the criteria in your subject guide beats another evening of polishing sentences.
TI-Nspire CX II CAS 图形计算器 AP/IB/A-Level¥1320📱 去京东购买京东精选需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
16621398022 同微信
aleveler.com
或截图后用「扫一扫 → 相册」选择本图