关于「听其自述的边界:事后解释、出声思考与反应性」,《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:出声思考保住了 0.919 的步骤,与真实次序的肯德尔相关是 0.977,0.968 的报告仍能看出至少一处跳跃。事后回忆只记起 0.648 的步骤,相关降到 0.833,能看出跳跃的报告只剩 0.235。
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。
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: 0.235, 0.648, 0.833, 0.919, 0.968, 0.977
| synthetic people per condition: 2,000 | |||
| condition | coverage | tau(real) | leap kept |
| concurrent | 0.919 | 0.977 | 0.968 |
| retrospective | 0.648 | 0.833 | 0.235 |
| textbook order tau(real) = 0.786 |
查看完整验算程序
# Chapter 1. p12: does the report keep the order and the leaps?
# A reasoning episode has eight steps 0..7. The textbook order is 0..7;
# the real episode jumped twice: 4 came before 2 and 3, and 6 before 5.
# Concurrent think-aloud: each step reported with prob 0.92, neighbours
# swapped with prob 0.04. Retrospective account: each step recalled with
# prob 0.65; with prob 0.6 the person rebuilds the textbook order,
# otherwise keeps the real order with neighbour swaps at 0.10.
# Measures over 2,000 synthetic people per condition: coverage, Kendall
# tau against the real order, and the share of reports that still show
# a leap (4 before 2, or 6 before 5). Parameters are illustrative.
import random
import statistics
REAL = [0, 1, 4, 2, 3, 6, 5, 7]
N = 2000
def swaps(seq, p, rng):
s = list(seq)
for i in range(len(s) - 1):
if rng.random() < p:
s[i], s[i + 1] = s[i + 1], s[i]
return s
def tau(rep):
pos = {v: i for i, v in enumerate(REAL)}
conc = disc = 0
for i in range(len(rep)):
for j in range(i + 1, len(rep)):
if pos[rep[i]] < pos[rep[j]]:
conc += 1
else:
disc += 1
tot = conc + disc
return (conc - disc) / tot if tot else 0.0
def keeps_leap(rep):
where = {v: i for i, v in enumerate(rep)}
for a, b in ((4, 2), (6, 5)):
if a in where and b in where and where[a] < where[b]:
return 1
return 0
def concurrent(rng):
kept = [s for s in REAL if rng.random() < 0.92]
return swaps(kept, 0.04, rng)
def retrospective(rng):
kept = [s for s in REAL if rng.random() < 0.65]
if rng.random() < 0.6:
return sorted(kept)
return swaps(kept, 0.10, rng)
def main():
rng = random.Random(38512)
print("synthetic people per condition: {:,}".format(N))
print("condition coverage tau(real) leap kept")
for name, fn in (("concurrent", concurrent),
("retrospective", retrospective)):
cov, taus, leap = [], [], 0
for _ in range(N):
rep = fn(rng)
cov.append(len(rep) / len(REAL))
taus.append(tau(rep))
leap += keeps_leap(rep)
print("{:<14}{:>10.3f}{:>11.3f}{:>11.3f}".format(
name, statistics.fmean(cov), statistics.fmean(taus),
leap / N))
print("textbook order tau(real) = {:.3f}".format(tau(sorted(REAL))))
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 1 章「听其自述的边界:事后解释、出声思考与反应性」(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}
}