关于「时间预算的拉格朗日分配:人数是算出来的,不是定出来的」,《邓巴分层与私域半衰期》给出的定量结论是什么?

结论:第四章用拉格朗日乘子分配每周时间,最内 5 人占 0.406,平均分给 150 人只得到最优价值的 0.598。

数据口径:书中的层级下限、半衰期、权重与停止概率均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,合规规则以各地现行法规为准。 书中以标注框区分已发表的证据、可运行的 Python 计算(19 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: 0.598, 150

policyvaluevs LAYERpeoplemin/person(equal)
LAYER281.31.00050-
CAP150168.20.5981504.00
ALL50096.30.3435001.20
CAP50E236.50.8415012.00
LAYER gains 113.0 over CAP150 with the same 600 min
min/week CAP150 needs to match LAYER: 1133

核心公式

\[\max \sum_i n_i w_i \ln\left(1 + \frac{t_i}{c}\right) \quad \text{满足} \quad \sum_i n_i t_i = T,\ t_i \ge 0.\]
查看完整验算程序
# Chapter 4. p42: a head-count cap versus a layered time budget
# Same value model as p41: n_i people per layer, weight w_i, value
# w * ln(1 + t / c), c = 10 minutes, budget T = 600 minutes a week
# (all illustrative). Policies compared:
#   CAP150  the first 150 people, equal minutes each
#   ALL500  everyone in the list, equal minutes each
#   LAYER   the Lagrange-optimal split of p41 over all 500
#   CAP50E  the first 50 people, equal minutes each
# Reported: total value, value relative to LAYER, people reached.
import math

LAYERS = [("1-5", 5, 16.0), ("6-15", 10, 8.0), ("16-50", 35, 4.0),
          ("51-150", 100, 2.0), ("151-500", 350, 1.0)]
C, T = 10.0, 600.0


def allocate(T, layers=LAYERS, c=C):
    order = sorted(range(len(layers)), key=lambda i: -layers[i][2])
    best = None
    for k in range(1, len(order) + 1):
        act = order[:k]
        nw = math.fsum([layers[i][1] * layers[i][2] for i in act])
        nn = math.fsum([layers[i][1] for i in act])
        inv_mu = (T + c * nn) / nw
        t = [0.0] * len(layers)
        for i in act:
            t[i] = inv_mu * layers[i][2] - c
        if min(t[i] for i in act) < 0:
            break
        best = (1.0 / inv_mu, t)
    return best


def value(t):
    return math.fsum([n * w * math.log(1.0 + ti / C)
                      for (_, n, w), ti in zip(LAYERS, t)])


def equal_value(m):
    per = T / m
    terms, left = [], m
    for _, n, w in LAYERS:
        take = min(n, left)
        left -= take
        terms.append(take * w * math.log(1.0 + per / C))
    return math.fsum(terms)


def main():
    _, t_opt = allocate(T)
    v_opt = value(t_opt)
    reach = int(math.fsum([n for (_, n, _), ti in zip(LAYERS, t_opt)
                           if ti > 0]))
    rows = [("LAYER", v_opt, reach)]
    for name, m in (("CAP150", 150), ("ALL500", 500), ("CAP50E", 50)):
        rows.append((name, equal_value(m), m))
    print("policy   value   vs LAYER  people  min/person(equal)")
    for name, v, m in rows:
        per = "-" if name == "LAYER" else "{:.2f}".format(T / m)
        print("{:7s} {:7.1f} {:9.3f} {:7d}  {}".format(
            name, v, v / v_opt, m, per))
    gain = v_opt - equal_value(150)
    print("LAYER gains {:.1f} over CAP150 with the same 600 min".format(
        gain))
    print("min/week CAP150 needs to match LAYER: ", end="")
    lo, hi = T, 50.0 * T
    for _ in range(80):
        mid = 0.5 * (lo + hi)
        per = mid / 150.0
        terms = []
        left = 150
        for _, n, w in LAYERS:
            take = min(n, left)
            left -= take
            terms.append(take * w * math.log(1.0 + per / C))
        if math.fsum(terms) < v_opt:
            lo = mid
        else:
            hi = mid
    print("{:.0f}".format(hi))


main()

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《邓巴数分层、关系维护时间预算与私域客户衰减半衰期母典》第 4 章「时间预算的拉格朗日分配:人数是算出来的,不是定出来的」(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-231 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026231,
  author    = {Fan, Yafu},
  title     = {Dunbar's Layered Numbers, Relationship-Maintenance Time Budgets and the Decay Half-Life of Private-Channel Customers},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论