📚 Human imperfection | 数学中的人性不完美
Human imperfection is not a flaw that mathematics tries to hide; it is the very reason why mathematics has developed powerful theories of approximation, error and risk. In A-Level Mathematics, we learn that exact answers are often impossible and that our measurements, calculations and conclusions are always accompanied by uncertainty. Understanding these limitations is part of becoming a competent and honest mathematician.
人类的不完美并不是数学试图掩盖的缺陷，而恰恰是数学发展出近似、误差和风险等强大理论的原因。在 A-Level 数学中，我们认识到精确答案往往不可能获得，而我们的测量、计算和结论总是伴随不确定性。理解这些局限是成为合格而诚实的数学家的部分。
1. The Illusion of Exactness | 精确性的错觉
Exact numbers exist only in abstract mathematics. In the real world, every measurement is limited by the instrument, the human eye, and the environment. For example, the value of π was never discovered; it was approximated over centuries. Writing π ≈ 3.14 is a human choice that makes calculation convenient while sacrificing absolute precision.
精确的数只存在于抽象数学中。在现实世界里，每一次测量都受到仪器、人眼和环境的限制。例如，π 的值从来不是被“发现”的精确数，而是经过数百年近似得到的。写 π ≈ 3.14 是人类为了方便计算而作出的选择，它以牺牲绝对精度为代价。
In A-Level Mathematics, we learn to express answers to a specified number of significant figures or decimal places, such as 4.56 (3 s.f.). This notation is an open admission that the unrounded value is known, but the reported value is what a sensible human will use. The difference between the exact value and the rounded value is a measure of human imperfection.
在 A-Level 数学中，我们学会将答案写成指定位数的有效数字或小数，例如 4.56（3 位有效数字）。这种记号公开承认未舍入的值是存在的，但报告值是理智的人会采用的。精确值与舍入值之差，就是人类不完美的一种度量。
2. Rounding and Truncation Errors | 舍入与截断误差
Rounding replaces a number by a nearby, shorter decimal, while truncation simply cuts off the tail. Suppose a true value is 7.856. Rounded to two decimal places, it becomes 7.86, and the rounding error is 0.004. If we truncate to two decimal places, we get 7.85, and the truncation error is 0.006. In both cases, the human operation creates a difference from the original value.
舍入是用一个更接近的短小数替换原数，而截断则是直接切掉尾数。假设真值为 7.856。四舍五入到两位小数后为 7.86，舍入误差为 0.004。如果截断到两位小数，得到 7.85，截断误差为 0.006。无论哪种操作，人为处理都会产生与原值的差异。
When a rounded number is used, its true range is bounded by half the unit of rounding. For example, a value given as 12.3 cm to one decimal place lies between 12.25 cm and 12.35 cm. This error bound protects us from false certainty. Without such bounds, human mistakes would silently accumulate in multi-step calculations.
当使用舍入后的数字时，其真实范围被限制在舍入单位的一半以内。例如，给定位 12.3 cm（精确到一位小数）的值，实际在 12.25 cm 与 12.35 cm 之间。这种误差界限保护我们免受虚假确定性的误导。如果没有这些界限，人类的错误会在多步计算中悄无声息地累积。
3. Percentage Error | 百分误差
To compare errors fairly, we use percentage error. If an approximate value v is accepted instead of the true value t, then the percentage error is the absolute difference divided by the true value, multiplied by 100%. This standardises the size of the error so that a 0.2 cm error on a 5 cm measurement is recognised as much more serious than the same 0.2 cm error on a 200 cm measurement.
为了公平比较误差，我们使用百分误差。若用近似值 v 代替真值 t，则百分误差为绝对差除以真值再乘以100%。这样将误差大小标准化，使 5 cm 上的 0.2 cm 误差与 200 cm 上的 0.2 cm 误差相比，前者更严重这一事实变得显而易见。
percentage error = (| true value − approximate value | / | true value |) × 100%
For example, a student estimates the area of a rectangle as 24 cm², but the correct area is 25 cm². The percentage error is (|25 − 24| / 25) × 100% = 4%. This small number tells us that the human estimation, while not perfect, is reasonably close. Percentage error turns our imperfection into a useful, comparable statistic.
例如，一个学生估计矩形的面积为 24 cm²，而正确面积为 25 cm²。百分误差为 (|25 − 24| / 25) × 100% = 4%。这个小数告诉我们，人类的估算虽然不完美，但已相当接近。百分误差将我们的不完美转化为一种有用且可比较的统计量。
4. Binomial and Small-Value Approximations | 二项式与小量近似
For a rational n, the binomial expansion of (1+x)ⁿ is an infinite series. When x is very small, we often use the first two terms only: (1+x)ⁿ ≈ 1 + nx. This approximation ignores terms involving x² and higher powers. The choice to stop early is a human decision, a deliberate trade-off between simplicity and accuracy.
对于有理数 n，(1+x)ⁿ 的二项式展开是无穷级数。当 x 很小时，我们常常只使用前两项：(1+x)ⁿ ≈ 1 + nx。这个近似忽略了含 x² 及更高次幂的项。提前截断是人类作出的决定，是在简洁与准确之间进行的取舍。
The omitted terms are the error of the approximation. For instance, with n = 4, the full expansion is 1 + 4x + 6x² + 4x³ + x⁴. If we write (1+x)⁴ ≈ 1 + 4x, the error is about 6x² for small x. A-level students learn to judge whether such an error is acceptable in context. This practical judgement is exactly where human imperfection meets mathematical sophistication.
被省略的项就是近似的误差。例如，当 n=4 时，完整展开为 1 + 4x + 6x² + 4x³ + x⁴。如果写 (1+x)⁴ ≈ 1 + 4x，那么对于小 x，误差约为 6x²。A-Level 学生学会判断这种误差在具体情境下是否可接受。这种实际判断正是人类不完美与数学精细程度相遇之处。
5. Newton–Raphson: A Human Starting Guess | 牛顿-拉弗森法：人类的初始猜测
The Newton–Raphson iteration is a beautiful numerical method for solving f(x) = 0. The formula is xₙ₊₁ = xₙ − f(xₙ) / f′(xₙ). Yet the method cannot begin without a human-chosen starting value x₀. The choice of x₀ reflects our intuition about the shape of the curve, and that intuition can be imperfect.
牛顿-拉弗森迭代是求解 f(x) = 0 的优美数值方法。公式为 xₙ₊₁ = xₙ − f(xₙ) / f′(xₙ)。然而，如果没有人为选择的初始值 x₀，这个方法就无法开始。x₀ 的选择反映了我们对曲线形状的直觉，而这种直觉可能不完美。
If the starting point is too close to a stationary point, f′(x₀) is very small, and the next estimate can leap far away. A poor guess may lead to convergence towards the wrong root, or none at all. This shows that human error is not removed by formulas; rather, a sound mathematical method should be paired with good understanding.
如果初始点过于靠近驻点，f′(x₀) 就会非常小，下一个估计值可能远远跳离。糟糕的猜测可能导致收敛到错误的根，甚至完全不收敛。这说明人类错误不会被公式自动消除；相反，可靠的数学方法应配以良好的理解。
xₙ₊₁ = xₙ − f(xₙ) / f′(xₙ)
6. Statistical Sampling: A Glimpse of the Whole | 统计抽样：窥见整体
Human beings cannot inspect every element of a large population. We must rely on samples, but a sample is never a perfect picture of the population. For instance, the mean height of 10 randomly selected students is unlikely to equal the mean height of all 1,000 students in a college. This mismatch is called sampling error, and it is an unavoidable companion of human decision-making.
人类不可能检查大型总体中的每一个元素。我们只能依靠样本，但样本永远不是总体的完美图像。例如，随机抽取 10 名学生的平均身高不太可能等于学院里 1000 名学生的平均身高。这种不一致被称为抽样误差，它是人类决策中不可避免的伴侣。
Random sampling is a way to reduce bias, but it cannot remove chance variation. A-Level Statistics introduces the idea that different samples from the same population give different means; the spread of those means follows a distribution. We can then calculate a confidence interval, which is a range of plausible values for the true population mean. The interval itself is a numerical acknowledgement of human imperfection.
随机抽样可以减少偏倚，但无法消除随机变异。A-Level 统计学介绍了同一个总体中不同样本给出不同均值的思想，而这些均值的分布遵循某一分布。我们可以随后计算置信区间，即真实总体均值可能落入的范围。这个区间本身就是对人类不完美的数值化承认。
7. Hypothesis Testing and the Probability of Error | 假设检验与错误的概率
In hypothesis testing, a human researcher chooses a significance level, often α = 0.05. This means there is a 5% chance of rejecting the null hypothesis when it is actually true. This is not a proof that the conclusion is correct; it is a deliberate acceptance of a certain probability of being wrong.
在假设检验中，研究者会选择显著性水平，通常取 α=0.05。这意味着存在5%的概率在零假设实际上为真时拒绝它。这不是结论正确的证明，而是有意接受一定程度出错概率。
There are two kinds of human error in testing. A Type I error happens when we reject a true null hypothesis; a Type II error happens when we fail to reject a false null hypothesis. No test can eliminate both simultaneously. The table below summarises this unavoidable imperfection:
检验中存在两种人为错误。第一类错误是拒绝了真的零假设；第二类错误是未能拒绝假的零假设。没有任何检验能同时消除两者。下表总结了这种不可避免的不完美：
| Decision / Reality | H₀ true | 更多咨询请联系16621398022（同微信）
CommentsMore posts |
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导Cancel reply