关于「实证闭环:黑名单共享协议、保证金与代理人信用评分卡」,《商帮多边惩罚与买手信任网络》给出的定量结论是什么?

结论:第八章给出八条黑名单共享条款、预注册检验方案与蒸馏小工具「代理人信用评分卡」;模拟中,核实式共享名单把有违约动机的代理人比例降到 0.245,不核实的名单则回升到 0.417。评分卡只用交易行为数据,不得用于歧视性用途。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

数据口径:书中代理人、订单与网络参数均为示意;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。 第四章求临界群规模:链状熟人网络在示意参数下只有几十到一百多人,均匀混合网络要到 33742 人。

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: 0.245, 0.417

protocolsedefaultwrongly listed
none0.00.0000.9400 of 24
verified0.90.0050.24534 of 302
unverified0.90.0500.417177 of 233
defaults counted among 400 agents; listing term 12 months
查看完整验算程序
# Chapter 8. p83: three blacklist protocols on one simulated market.
# Model: 400 agents, each with a private discount d_i ~ U(0.80, 0.98)
# and a typical order at stake G_i ~ U(0.5, 6) rent units (fixed seed).
# An agent defaults iff G_i exceeds his G_max under the protocol, using
# the p52 formulas: listing term T = 12, outsider rehire q0 = H (1 - s),
# false-report rate e. Protocols: none (s = 0, e = 0), verified sharing
# (s = 0.9, e = 0.005) and unverified sharing (s = 0.9, e = 0.05). We
# also count honest agents wrongly listed at least once in 24 months.
# 模型:无共享、核实后共享、未经核实共享三种协议的违约与误伤。
import random

R, W0, H, T = 1.0, 1.0, 0.5, 12


def g_max(d, s, e):
    q0 = H * (1.0 - s)
    a, b = 1.0, 0.0
    for _ in range(T):
        a, b = q0 + (1.0 - q0) * d * a, (1.0 - q0) * (W0 + d * b)
    w = W0 + R
    v = (w + d * e * b) / (1.0 - d * (1.0 - e) - d * e * a)
    return v - w - d * (a * v + b)


def market(rng):
    return [(rng.uniform(0.80, 0.98), rng.uniform(0.5, 6.0))
            for _ in range(400)]


if __name__ == "__main__":
    rng = random.Random(20261112)
    agents = market(rng)
    print("  protocol        s     e      default  wrongly listed")
    for name, s, e in (("none", 0.0, 0.0), ("verified", 0.9, 0.005),
                       ("unverified", 0.9, 0.05)):
        bad = [g > g_max(d, s, e) for d, g in agents]
        honest = [i for i, x in enumerate(bad) if not x]
        hit = 0
        for _ in honest:
            if any(rng.random() < e for _ in range(24)):
                hit += 1
        print("  {:<12s}  {:.1f}   {:.3f}  {:.3f}    {:3d} of {}".format(
            name, s, e, len([x for x in bad if x]) / 400.0, hit,
            len(honest)))
    print("defaults counted among 400 agents; listing term 12 months")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《商帮多边惩罚声誉机制、马格里布商人代理博弈与买手社群信任网络母典》第 8 章「实证闭环:黑名单共享协议、保证金与代理人信用评分卡」(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-222 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026222,
  author    = {Fan, Yafu},
  title     = {Multilateral Punishment in Merchant Networks: Maghribi Agency Games and Trust Networks in Buyer Communities},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论