关于「飞行教官的错觉:批评为何看起来比表扬有效」,《均值回归与封面诅咒》给出的定量结论是什么?

结论:第二章重演飞行教官的训练季:反馈毫无作用时,教官仍会看到批评比表扬多带来 2.529 个单位的进步,只有随机对照能纠正。

数据口径:书中门店、达人、顾客与报告的参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议,涉及运程报告与原典之处只分析结构与概率,不作任何个人占断。 书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。

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: 2.529

landing-to-landing correlation rho = 0.3289
feedbacknmean changeshare bettershare worse
praise175-1.2980.1260.874
criticism189+1.2310.8150.185
silence586+0.0060.4880.512
apparent advantage of criticism over praise: +2.529
true causal effect of either word:+0.000
查看完整验算程序
# Chapter 2. p21: The flight instructor's season, with feedback that does
# nothing. 50 cadets fly 20 landings each. A landing's score is the
# cadet's skill (sd 0.7) plus a fresh draw of noise (sd 1). The instructor
# praises any landing above +1 and criticises any below -1; by
# construction neither word changes the next landing. We record what the
# instructor sees on the next landing. Fixed seed; parameters illustrative.
# 教官的一个训练季:表扬与批评都没有任何因果作用。
import math
import random

rng = random.Random(1974)
CADETS, LANDINGS = 50, 20
SKILL_SD, NOISE_SD, CUT = 0.7, 1.0, 1.0

after = {"praise": [], "criticism": [], "silence": []}
for _ in range(CADETS):
    s = rng.gauss(0.0, SKILL_SD)
    prev = s + rng.gauss(0.0, NOISE_SD)
    for _ in range(LANDINGS - 1):
        nxt = s + rng.gauss(0.0, NOISE_SD)
        if prev > CUT:
            key = "praise"
        elif prev < -CUT:
            key = "criticism"
        else:
            key = "silence"
        after[key].append(nxt - prev)
        prev = nxt


def mean(xs):
    return math.fsum(xs) / len(xs)


if __name__ == "__main__":
    rho = SKILL_SD ** 2 / (SKILL_SD ** 2 + NOISE_SD ** 2)
    print("landing-to-landing correlation rho = {:.4f}".format(rho))
    print("feedback    n    mean change   share better   share worse")
    for key in ("praise", "criticism", "silence"):
        d = after[key]
        up = len([v for v in d if v > 0])
        print("{:<10} {:>4}   {:>+9.3f}     {:>8.3f}      {:>8.3f}"
              .format(key, len(d), mean(d), up / len(d),
                      1 - up / len(d)))
    gap = mean(after["criticism"]) - mean(after["praise"])
    print("apparent advantage of criticism over praise: {:+.3f}"
          .format(gap))
    print("true causal effect of either word:            +0.000")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《均值回归与封面诅咒、运程报告后验校准与塞翁失马序列独立性母典》第 2 章「飞行教官的错觉:批评为何看起来比表扬有效」(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-202 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026202,
  author    = {Fan, Yafu},
  title     = {Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论