5D Library › 说服文案代数、损失框架不对称性与结账微界面工程 · English

第七章 《韩非子·说难》《文心雕龙》与纵横家的利害之辞

为什么在这里读这些原典

前六章的模型都有古老的对应:第一章的损失框架就是「陈说利害」,第二章的精确与可信关乎「事信」,第四章的流畅度关乎「言文」,而第六章最大的空间在于读者对报价的兴趣,也就是「知所说之心」。本章核对这些文本,并把其中一条写成可以计算的模型。

韩非:难在知所说之心

原典

《韩非子·说难》

凡說之難,在知所說之心,可以吾說當之。

(陈奇猷 2000)

韩非把游说的难处放在了解对方,而不是辞藻本身。程序把这句话写成「信息的价值」:三类听者分别被收益、避损与事实打动(接受概率为示意值),说话者若分不清对方是哪一类,只能对所有人用同一种说法;若分得清,可以对每一类用其最有效的说法。

推导

# Chapter 7. p13: "Know the heart of the one you persuade" as the value
# of information. Three listener types, each moved by a different appeal
# (gain, loss avoided, plain facts). Acceptance probability by type and
# appeal below (illustrative). A speaker who cannot tell types uses the
# single best appeal for the mix; one who can tell uses each type's best.
TYPES = (("seeks gain", 0.40), ("fears loss", 0.35), ("wants facts", 0.25))
ACCEPT = {"seeks gain": {"gain": 0.30, "loss": 0.18, "facts": 0.20},
          "fears loss": {"gain": 0.15, "loss": 0.32, "facts": 0.20},
          "wants facts": {"gain": 0.10, "loss": 0.08, "facts": 0.35}}

if __name__ == "__main__":
    appeals = ("gain", "loss", "facts")
    for a in appeals:
        tot = sum(w * ACCEPT[t][a] for t, w in TYPES)
        print("   one appeal for everyone, {:5s}: {:.4f}".format(a, tot))
    blind = max(sum(w * ACCEPT[t][a] for t, w in TYPES) for a in appeals)
    known = sum(w * max(ACCEPT[t].values()) for t, w in TYPES)
    print("   best single appeal {:.4f}; matched to each type {:.4f}".format(
        blind, known))
    print("   value of knowing the listener: {:+.1%}".format(
        known / blind - 1))
    for acc in (1.0, 0.8, 0.6):
        # the speaker reads the type correctly with probability acc,
        # otherwise uses a random wrong appeal
        val = 0.0
        for t, w in TYPES:
            best = max(ACCEPT[t], key=ACCEPT[t].get)
            wrong = [ACCEPT[t][a] for a in appeals if a != best]
            val += w * (acc * ACCEPT[t][best] + (1 - acc) * sum(wrong) / 2)
        print("   reading accuracy {:.1f}: {:.4f}".format(acc, val))
#    one appeal for everyone, gain : 0.1975
#    one appeal for everyone, loss : 0.2040
#    one appeal for everyone, facts: 0.2375
#    best single appeal 0.2375; matched to each type 0.3195
#    value of knowing the listener: +34.5%
#    reading accuracy 1.0: 0.3195
#    reading accuracy 0.8: 0.2875
#    reading accuracy 0.6: 0.2556

对所有人用同一种说法,最好的是「讲事实」,接受率 0.2375;按类型匹配,升到 0.3195,即「知所说之心」的价值为 +34.5%。判断的准确率若只有 0.8,接受率降到 0.2875;0.6 时为 0.2556。值得注意的是,在这组示意参数下,单一说法中最好的不是恐惧,而是事实。

刘勰:事信而言文

原典

《文心雕龙·征圣》

志足而言文,情信而辭巧。

《文心雕龙·宗经》论经典之文有六义,其三曰「事信而不誕」。(范文澜 1958)

刘勰把「信」放在「文」之前:辞巧建立在情信之上,文采建立在事实可信之上。这与第二章的结论相合:精确的数字之所以可信,是因为它可以被核查;与第四章的提醒也相合:流畅只是「言文」,不能代替「事信」。

纵横家:陈说利害

《战国策》所载张仪游说诸国,常先陈述对方若不从将面临的危险,再给出从之的好处,是有记录的损失框架 (刘向 1978)。但纵横家所陈说的利害,是对方国家真实面临的军事与外交形势;一旦游说者所陈之害并不存在,信用便随之瓦解。这与第五章的账一致:「损失」必须是真实的。

小结

《说难》把说服的关键放在了解听者,程序给出这一了解的价值:示意参数下为 +34.5%。《文心雕龙》把「事信」置于「言文」之前,与精确数字的可核查性相合。纵横家的利害之辞以真实形势为前提。

参考文献

刘向. 1978. 战国策. 上海古籍出版社.
范文澜. 1958. 文心雕龙注. 人民文学出版社.
陈奇猷. 2000. 韩非子新校注. 上海古籍出版社.

Fan, Y. (2026). Persuasive Copywriting Algebra, Loss-Framing Asymmetry and Checkout Micro-Architecture: Prospect-Theory Valuation Ratios, Numerical Specificity Credibility and Cognitive-Fluency Conversion Bounds [说服文案代数、损失框架不对称性与结账微界面工程] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090205 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).