关于「四种解释的指纹:资源、过程、信念与排序」,《自我损耗与饥饿法官》给出的定量结论是什么?

结论:资源、过程、信念与排序四个故事可以产生完全相同的日内衰退曲线,在一个普通的日子里,它们之间的差距为 0.000。打乱顺序的设计区分力最强,能把排序伪影与其余三者分开;激励设计能区分资源与过程。随机安排休息时点是一个更易执行的替代方案:在 p52 的模拟中,它把纯疲劳世界的休息距离斜率正确估计为 −0.0092,而朴素分析只看到 −0.0029。下一章假定衰退效应存在且大小未知,计算一位老板应当怎样安排每天的决策顺序与休息。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是心理或医疗建议,书中流程与工具为非个人化经营建议。 本书把两项证据放到预注册复制、元分析偏差检验与案件排序模拟面前,看它们还剩多少。

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: 0.000

first-to-last fall in quality under each design
story plain day break at 12 late incentive shuffled order
resource0.230 0.164 0.184 0.230
process0.230 0.120 0.000 0.230
belief0.230 0.120 0.115 0.230
artefact0.230 0.230 0.230 0.000
largest spread between stories: shuffled order (0.230)
the plain day separates nothing: spread 0.000
stories that predict a fall after shuffling3 of 4
mean fall across stories, break at 120.1585
查看完整验算程序
# Chapter 5. p51: Four stories, one curve, four different fingerprints.
# Each story produces the same observed fall in a day of 24 decisions:
# 0.010 per decision, 0.23 from the first to the last. They differ in
# what a break after decision 12, a late incentive and a shuffled order
# do to that fall. Resource: depletion, a break restores 0.6 of the loss,
# an incentive 0.2. Process (motivation and attention shift): a break or
# an incentive restores all. Belief: half the deciders believe willpower
# is limited and fall 0.020 a decision, the rest do not fall. Artefact:
# quality is flat; case mix is sorted easy-last. Illustrative values.
# 四种解释,同一条曲线,四种不同的「指纹」。
import math

N, SLOPE = 24, 0.010
STORIES = {
    # name: (break restore share, incentive restore share, shuffle keeps)
    "resource": (0.6, 0.2, True),
    "process": (1.0, 1.0, True),
    "belief": (1.0, 0.5, True),
    "artefact": (0.0, 0.0, False),
}


def fall(story, design):
    br, inc, keeps = STORIES[story]
    total = SLOPE * (N - 1)
    if design == "plain day":
        return total
    if design == "break at 12":
        # loss accumulated before the break is partly restored
        before = SLOPE * 11
        return total - br * before
    if design == "late incentive":
        return total * (1 - inc)
    if design == "shuffled order":
        return total if keeps else 0.0
    raise ValueError(design)


DESIGNS = ("plain day", "break at 12", "late incentive", "shuffled order")
print("first-to-last fall in quality under each design")
print("story      " + "  ".join("{:>14s}".format(d) for d in DESIGNS))
for s in STORIES:
    print("{:9s}  ".format(s) + "  ".join(
        "{:14.3f}".format(fall(s, d)) for d in DESIGNS))
spread = {}
for d in DESIGNS:
    vals = [fall(s, d) for s in STORIES]
    spread[d] = max(vals) - min(vals)
best = max(DESIGNS, key=lambda d: (spread[d], d))
print("largest spread between stories: {} ({:.3f})".format(
    best, spread[best]))
print("the plain day separates nothing: spread {:.3f}".format(
    spread["plain day"]))
cnt = len([1 for s in STORIES if STORIES[s][2]])
print("stories that predict a fall after shuffling: {} of {}".format(
    cnt, len(STORIES)))
print("mean fall across stories, break at 12: {:.4f}".format(
    math.fsum(fall(s, "break at 12") for s in STORIES) / len(STORIES)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《自我损耗大规模复制失败、法官饥饿效应排序伪影与决策疲劳日程母典》第 5 章「四种解释的指纹:资源、过程、信念与排序」(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-196 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026196,
  author    = {Fan, Yafu},
  title     = {The Failed Replication of Ego Depletion, the Ordering Artefact behind the Hungry Judges and Decision-Fatigue Scheduling},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论