关于「检测器为什么不能当证据:误报、偏差与人眼近随机」,《机器文气与句长节律》给出的定量结论是什么?
结论:检测器不可靠,这是已发表研究反复给出的结论。p31 说明,即便排序能力看上去不错,基础比例低时被标记的稿子大半是冤枉的;p32 说明,同一个阈值落在词汇面较窄的写作者身上,误报率可以翻十倍以上;p33 说明,人眼判别要几百次才能和随机区分开。
数据口径:书中示例均为作者自写的合成文字或公有领域文本,参数为示意,全部计算由书中程序生成。 第一至六章依次讲句长指纹与对数正态、突发度与顺序量、检测器的贝叶斯账与偏差、套话密度的泊松检验与群体超额词汇、三元连发与压缩冗余、只动句读的动态规划与真实细节账本:合成草稿只改 5 处句读,变异系数就从 0.123 升到 0.546,而字词一个不动。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库
| TPR | FPR | prevalence | PPV | wrong flags per 1,000 honest |
| 0.80 | 0.10 | 0.02 | 0.140 | 100.0 |
| 0.80 | 0.10 | 0.10 | 0.471 | 100.0 |
| 0.80 | 0.10 | 0.30 | 0.774 | 100.0 |
| 0.90 | 0.05 | 0.02 | 0.269 | 50.0 |
| 0.90 | 0.05 | 0.10 | 0.667 | 50.0 |
| 0.90 | 0.05 | 0.30 | 0.885 | 50.0 |
| 0.95 | 0.01 | 0.02 | 0.660 | 10.0 |
| 0.95 | 0.01 | 0.10 | 0.913 | 10.0 |
| 0.95 | 0.01 | 0.30 | 0.976 | 10.0 |
| 30 honest texts, FPR 0.05: P(at least one flag) = 0.785 | ||||
| expected honest writers flagged in 30: 1.50 | ||||
| log-odds a flag adds at TPR 0.90, FPR 0.10: 2.197 |
查看完整验算程序
# Chapter 3. p31: what a detector flag is worth (Bayes' rule).
# A detector with true-positive rate TPR and false-positive rate FPR is
# run on a pool in which a share pi of texts is machine-written.
# PPV = TPR pi / (TPR pi + FPR (1 - pi)) is the chance that a flagged
# text really is machine-written; "wrong per 1,000 honest" counts
# texts by human writers flagged per 1,000 such texts. The three
# operating points are illustrative, not the figures of any product.
import math
POINTS = [(0.80, 0.10), (0.90, 0.05), (0.95, 0.01)]
PREV = [0.02, 0.10, 0.30]
print("TPR FPR prevalence PPV wrong flags per 1,000 honest")
for tpr, fpr in POINTS:
for pi in PREV:
ppv = tpr * pi / (tpr * pi + fpr * (1 - pi))
print("{:.2f} {:.2f} {:.2f} {:.3f} {:>6.1f}".format(
tpr, fpr, pi, ppv, 1000 * fpr))
# accusation arithmetic for one class of 30 honest writers
n, fpr = 30, 0.05
p_any = 1 - (1 - fpr) ** n
print("30 honest texts, FPR 0.05: P(at least one flag) = {:.3f}".format(
p_any))
print("expected honest writers flagged in 30: {:.2f}".format(n * fpr))
odds = math.log(0.9 / 0.1)
print("log-odds a flag adds at TPR 0.90, FPR 0.10: {:.3f}".format(odds))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《机器生成文本统计指纹剥离、句长突发律动重构与宗师级文气呼吸感母典》第 3 章「检测器为什么不能当证据:误报、偏差与人眼近随机」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-394 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026394,
author = {Fan, Yafu},
title = {Stripping Statistical Fingerprints of Machine Prose, Sentence-Length Burstiness Reconstruction and Master-Level Stylistic Cadence},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}