关于「推理图:节点、跳跃、回溯与反例修正」,《推理轨迹识人与结果运气剥离》给出的定量结论是什么?

结论:十二条边时,区间从 0.063 到 0.483,半宽 0.210,几乎什么都说明不了。边数翻到 48 条,区间收到 0.111 到 0.333;96 条时是 0.132 到 0.291;192 条时半宽 0.056;384 条时是 0.163 到 0.243,半宽 0.040。

数据口径:事后解释常常是一套通俗理论:合成实验里,事后自述与真实权重的相关只有 -0.013,与旁观者凭常识的猜测却高达 0.826;同步出声思考保留了次序与跳跃,要求边做边解释则会改变被测过程。 核对文献后本书做了蓝图修正:人能表演推理,旁人判别真假的正确率只比抛硬币高一点,因此不主张「不可伪造的思维指纹」,只主张伪造成本随追问深度上升,并写出上下界与推翻条件;甄选效度以修订后的估计为锚,结构化访谈居首也只有 0.42 上下,单一方法有限,须按规则多方法合成。

English version

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 标准库与结论吻合的打印值: 0.040, 0.056, 0.063, 0.111, 0.132, 0.163

leap rate 0.20, Wilson 95% interval
edges low high half-width
120.063 0.483 0.210
240.087 0.396 0.154
480.111 0.333 0.111
960.132 0.291 0.079
1920.150 0.262 0.056
3840.163 0.243 0.040
edges per person to tell 0.15 from 0.30121
episodes of 12 edges11
within-episode icc 0.05: design effect 1.55, episodes 16
within-episode icc 0.10: design effect 2.10, episodes 22
查看完整验算程序
# Chapter 3. p32: how many coded edges before a leap rate means much?
# Wilson 95 per cent interval for a leap rate of 0.20 at several edge
# counts; then the edges per person needed to tell a leap rate of 0.15
# from 0.30 (two-sided alpha 0.05, power 0.80, normal approximation for
# two proportions). One episode like p31 yields about a dozen edges.
# Edges inside one episode are not independent, so the counts below
# are lower bounds (see the design-effect line). Illustrative values.
import math
from statistics import NormalDist

Z = NormalDist()


def wilson(p, n, z=1.959964):
    den = 1 + z * z / n
    mid = (p + z * z / (2 * n)) / den
    half = z * math.sqrt(p * (1 - p) / n + z * z / (4 * n * n)) / den
    return mid - half, mid + half


def n_two(p1, p2, alpha=0.05, power=0.80):
    za = Z.inv_cdf(1 - alpha / 2)
    zb = Z.inv_cdf(power)
    pb = (p1 + p2) / 2
    num = za * math.sqrt(2 * pb * (1 - pb)) + zb * math.sqrt(
        p1 * (1 - p1) + p2 * (1 - p2))
    return math.ceil((num / (p1 - p2)) ** 2)


def main():
    print("leap rate 0.20, Wilson 95% interval")
    print("edges   low    high   half-width")
    for n in (12, 24, 48, 96, 192, 384):
        lo, hi = wilson(0.20, n)
        print("{:>5}{:>7.3f}{:>8.3f}{:>9.3f}".format(n, lo, hi,
                                                      (hi - lo) / 2))
    n = n_two(0.15, 0.30)
    print("edges per person to tell 0.15 from 0.30: {}".format(n))
    print("episodes of 12 edges: {}".format(math.ceil(n / 12)))
    for icc in (0.05, 0.10):
        deff = 1 + (12 - 1) * icc
        print("within-episode icc {:.2f}: design effect {:.2f}, "
              "episodes {}".format(icc, deff, math.ceil(n * deff / 12)))


main()

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《思维脉络溯源识人、过程追踪博弈与剥离意图结果假象母典》第 3 章「推理图:节点、跳跃、回溯与反例修正」(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-385 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026385,
  author    = {Fan, Yafu},
  title     = {Reasoning-Trajectory Profiling, Process-Tracing Games and Separating Cognitive Structure from Stated Intentions and Outcome Luck},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论