关于「推理路径图:命题节点、补全概率与分岔点」,《前提植入与零阻抗说服》给出的定量结论是什么?

结论:推理路径图把「说服」拆成可估的两类数:读者选哪条路,读者能否自己补全这一步。在本章的示意图上,最值钱的前提出现在信息不全导致误会的分岔点,增量是基线的 2.222 倍;能压低正当退出的前提一律不说;能帮助读者正当退出的前提即使降低目标概率,也照样说。

数据口径:核对原文以后,本书写下蓝图修正:O'Keefe 的元分析汇总 13,754 名被试,明示结论平均更有说服力,平均相关 0.110;省略结论只在高卷入、听者有能力自行推出时占优;书名里的「零阻抗」不是可证事实,只能写成逆反分量的上下界与模拟分位数。 书中参数为示意或合成数据,全部计算由书中程序生成。

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: 2.222

baseline P(T) 0.0557P(C) 0.2430P(N) 0.1945
edgedP(T)dP(C)dP(N)verdict
S->A-0.0078+0.1170-0.0273allowed
S->B+0.0427-0.1215+0.1492blocked: erodes honest exit
A->B+0.0306-0.1215+0.1069blocked: erodes honest exit
B->D+0.1238+0.0000-0.0972allowed
D->E+0.0239+0.0000+0.0000allowed
E->T+0.0139+0.0000+0.0000allowed
best allowed premise: B->D adds 0.1238 to P(T)
ratio best / baseline P(T): 2.222
查看完整验算程序
# Chapter 2. p22: premise value = gain in P(target) from one true premise
# Same graph as p21. Stating a premise on edge u -> v sets its completion
# to 1 (the reader no longer has to supply it) and, at a branch point,
# moves a share ETA of the other branches' choice weight onto u -> v.
# Ethics filter: a premise is allowed only if it is true, checkable and
# does not lower P(C), the honest "free answer is enough" exit.
import math

EDGES = [
    ("S", "A", 0.60, 0.90), ("S", "B", 0.40, 0.70),
    ("A", "C", 0.50, 0.90), ("A", "B", 0.50, 0.60),
    ("B", "D", 0.45, 0.50), ("B", "N", 0.55, 0.80),
    ("D", "E", 1.00, 0.70), ("E", "T", 1.00, 0.80),
]
ENDS = ("T", "C", "N")
ETA = 0.5
TRUE_AND_CHECKABLE = {("S", "A"), ("S", "B"), ("A", "B"), ("B", "D"),
                      ("D", "E"), ("E", "T")}


def outcome(edges, node):
    if node in ENDS:
        return {k: (1.0 if k == node else 0.0) for k in ENDS}
    out = [e for e in edges if e[0] == node]
    sub = {v: outcome(edges, v) for _, v, _, _ in out}
    return {k: math.fsum(w * c * sub[v][k] for _, v, w, c in out)
            for k in ENDS}


def seed(edges, u, v):
    out = [e for e in edges if e[0] == u]
    new = []
    for a, b, w, c in edges:
        if a != u:
            new.append((a, b, w, c))
        elif b == v:
            w2 = w + ETA * (1.0 - w) if len(out) > 1 else w
            new.append((a, b, w2, 1.0))
        else:
            new.append((a, b, w * (1.0 - ETA), c))
    return new


base = outcome(EDGES, "S")
print("baseline P(T) {:.4f}  P(C) {:.4f}  P(N) {:.4f}".format(
    base["T"], base["C"], base["N"]))
print("edge     dP(T)    dP(C)    dP(N)   verdict")
rows = []
for u, v, _, _ in EDGES:
    if v in ("C", "N"):
        continue
    r = outcome(seed(EDGES, u, v), "S")
    d = {k: r[k] - base[k] for k in ENDS}
    if (u, v) not in TRUE_AND_CHECKABLE:
        verdict = "blocked: not verifiable"
    elif d["C"] < -1e-12:
        verdict = "blocked: erodes honest exit"
    else:
        verdict = "allowed"
    rows.append((d["T"], u, v, verdict))
    print("{}->{}   {:+.4f}  {:+.4f}  {:+.4f}   {}".format(
        u, v, d["T"], d["C"], d["N"], verdict))
ok = sorted((r for r in rows if r[3] == "allowed"), reverse=True)
print("best allowed premise: {}->{} adds {:.4f} to P(T)".format(
    ok[0][1], ok[0][2], ok[0][0]))
print("ratio best / baseline P(T): {:.3f}".format(ok[0][0] / base["T"]))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《推演中途前提植入、省略三段论自发归结与零阻抗顺势说服母典》第 2 章「推理路径图:命题节点、补全概率与分岔点」(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-386 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026386,
  author    = {Fan, Yafu},
  title     = {Mid-Stream Premise Seeding, Enthymematic Self-Derivation and Zero-Reactance Persuasion Along Reasoning Paths},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论