关于「饥饿的法官还是排好的案卷:排序伪影的三种生成方式」,《自我损耗与饥饿法官》给出的定量结论是什么?

结论:第三章把「饥饿法官」拆成律师优先排序、控制代理后的剩余斜率与格勒克纳的时间管理三种机制;排序伪影的斜率可达真实疲劳的 4.68 倍。

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: 4.68

worldnaive slopeadjustedlawyer
random order-0.0013-0.00190.413
half the sessions sorted-0.02070.00110.410
all sessions sorted-0.0437-0.00160.381
random order, true fatigue-0.0093-0.00920.414
all sorted, first to 12th case: naive fall 0.481
artefact slope over true-fatigue slope: 4.68
查看完整验算程序
# Chapter 3. p32: Does adjusting for representation remove the slope?
# Same board as p31 (lawyer 0.35; grant 0.55 with, 0.15 without). In a
# share s of sessions the cases are sorted lawyers-first, otherwise
# heard in random order. A fourth world adds a true fatigue effect of
# -0.01 per position on random order. Linear probability models:
# naive  grant = a + b * position
# adjusted grant = a + b * position + c * lawyer
# Slopes are per position. Illustrative parameters; seed fixed.
# 控制「有无律师」之后,位置斜率是否消失。
import math
import random

rng = random.Random(20261112)
SESSIONS, P_REP = 2000, 0.35


def solve(a, b):
    # Gaussian elimination with partial pivoting
    n = len(b)
    m = [row[:] + [b[i]] for i, row in enumerate(a)]
    for c in range(n):
        p = max(range(c, n), key=lambda r: abs(m[r][c]))
        m[c], m[p] = m[p], m[c]
        for r in range(c + 1, n):
            f = m[r][c] / m[c][c]
            for j in range(c, n + 1):
                m[r][j] -= f * m[c][j]
    x = [0.0] * n
    for r in range(n - 1, -1, -1):
        s = math.fsum(m[r][j] * x[j] for j in range(r + 1, n))
        x[r] = (m[r][n] - s) / m[r][r]
    return x


def ols(rows, y):
    k = len(rows[0])
    a = [[math.fsum(r[i] * r[j] for r in rows) for j in range(k)]
         for i in range(k)]
    b = [math.fsum(r[i] * yi for r, yi in zip(rows, y)) for i in range(k)]
    return solve(a, b)


def world(sort_share, fatigue):
    xs1, xs2, y = [], [], []
    for _ in range(SESSIONS):
        n = rng.randint(5, 12)
        reps = [rng.random() < P_REP for _ in range(n)]
        if rng.random() < sort_share:
            reps.sort(reverse=True)
        for pos, r in enumerate(reps, start=1):
            p = (0.55 if r else 0.15) + fatigue * (pos - 1)
            g = 1.0 if rng.random() < max(0.0, p) else 0.0
            xs1.append([1.0, float(pos)])
            xs2.append([1.0, float(pos), 1.0 if r else 0.0])
            y.append(g)
    return ols(xs1, y)[1], ols(xs2, y)[1], ols(xs2, y)[2]


res = {}
print("world                         naive slope  adjusted  lawyer")
for label, s, f in (("random order", 0.0, 0.0),
                    ("half the sessions sorted", 0.5, 0.0),
                    ("all sessions sorted", 1.0, 0.0),
                    ("random order, true fatigue", 0.0, -0.01)):
    b_naive, b_adj, c = world(s, f)
    res[label] = b_naive
    print("{:28s}   {:8.4f}   {:8.4f}  {:6.3f}".format(
        label, b_naive, b_adj, c))
srt = res["all sessions sorted"]
tru = res["random order, true fatigue"]
print("all sorted, first to 12th case: naive fall {:.3f}".format(-11 * srt))
print("artefact slope over true-fatigue slope: {:.2f}".format(srt / tru))

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 3 章「饥饿的法官还是排好的案卷:排序伪影的三种生成方式」(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