关于「回弹与软化:纠正的衰减、间隔与半衰期」,《先验消解与反逆火沟通》给出的定量结论是什么?

结论:纠正有效,但会褪色。半衰期 3 天时,四周只纠正一回,平均降幅只有 0.0484;天天纠正,总降幅最大,可每次只换来 0.0185。拉力弱、半衰期 10 天时,每周一回就有 0.3118 的降幅。半衰期要从自己的记录里估,十轮数据估出的区间从 1.41 到 2.26 轮,并不窄。纠正还要成套,撤回结论的同时给出可核对的替代解释,否则空缺会把旧说法拉回来。

数据口径:书中参数为示意,全部计算由书中程序生成;本书非心理咨询、不提供任何临床服务、非个人化经营建议,例子只用商品与消费认知,不涉及政治、宗教与族群议题。 第六章把追问、肯定、证据与思想实验写成马尔可夫决策过程,示意参数下最难起点的期望收益 10.555,参数扰动两千次,完整路线不变的比例为 0.8270,沿用基准策略的平均相对遗憾只有 0.0003,所以只称模拟最优。

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: 0.0185, 0.0484, 0.3118, 10

relapse half-life 3 days, ALPHA = 0.4
everynmean xlast xgaingain/correction
1280.18280.16540.51720.0185
3100.38890.30000.31110.0311
740.52950.62060.17050.0426
1420.60580.68580.09420.0471
2810.65160.69950.04840.0484
relapse half-life 10 days, ALPHA = 0.4
everynmean xlast xgaingain/correction
1280.09280.06390.60720.0217
3100.24020.15420.45980.0460
740.38820.41250.31180.0779
1420.49340.56040.20660.1033
2810.57210.65690.12790.1279
single correction, half-life 3: day-0 x 0.42, mean 0.6516
查看完整验算程序
# Chapter 3. p31: corrections work, then decay; spacing decides the rest.
# Misperception x in [0, 1] (share of confidence in the false claim).
# Each day the environment (old posts, friends' remarks) pulls x back to
# X_ENV with half-life H days: x <- X_ENV + (x - X_ENV) * 2^(-1/H).
# A correction on a scheduled day removes a share ALPHA of x at once:
# x <- (1 - ALPHA) x. Horizon 28 days, start at X_ENV. Illustrative.
import math

X_ENV, ALPHA, DAYS = 0.7, 0.4, 28


def run(spacing, half_life):
    keep = 2.0 ** (-1.0 / half_life)
    x, path, n_corr = X_ENV, [], 0
    for day in range(DAYS):
        if spacing and day % spacing == 0:
            x *= 1.0 - ALPHA
            n_corr += 1
        path.append(x)
        x = X_ENV + (x - X_ENV) * keep
    return math.fsum(path) / DAYS, path[-1], n_corr


if __name__ == "__main__":
    for h in (3.0, 10.0):
        print("relapse half-life {:.0f} days, ALPHA = {}".format(h, ALPHA))
        print("  every  n   mean x   last x   gain   gain/correction")
        base = run(0, h)[0]
        for s in (1, 3, 7, 14, 28):
            m, last, n = run(s, h)
            gain = base - m
            print("  {:4d}  {:2d}   {:.4f}   {:.4f}   {:.4f}   {:.4f}".format(
                s, n, m, last, gain, gain / n))
    one = run(28, 3.0)
    print("single correction, half-life 3: day-0 x {:.2f}, mean {:.4f}"
          .format(
        X_ENV * (1 - ALPHA), one[0]))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《顽固先验偏见消解、解释深度错觉破除与反逆火效应沟通母典》第 3 章「回弹与软化:纠正的衰减、间隔与半衰期」(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-387 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026387,
  author    = {Fan, Yafu},
  title     = {Dissolving Entrenched Priors, Shattering the Illusion of Explanatory Depth and Anti-Backfire Persuasion Mechanics},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论