5D Library › 梦境意象二分图、确定性哈希牌卡与潜意识先验激活 · English

第六章 作为行动锚点的牌

每日一张牌实际能做什么

如果牌无法知道未来,它还能帮上忙吗?一条说得通的途径与牌面内容无关:牌缩小了选择,并且标记了开始。在许多相近的选项中做选择,所需时间随选项数增长(反应时研究中的希克-海曼定律 (Hick 1952)),而尚未决定的时间正是注意力溜走的时候。一个指定单一任务的指针,免去了大部分选择。两项元分析提供了参照:「如果出现情境 X,我就做 Y」形式的计划(执行意图)对目标达成的平均效应为中到大,d = 0.65 (Gollwitzer and Sheeran 2006);而常被引用的「选择过载」效应,跨研究平均接近于零,D = 0.02 (Scheibehenne et al. 2010)。下面的玩具模型把这个机制写清楚,参数是示意性的。

推导

# Chapter 6. p11: A toy model of action paralysis and the daily card.
# A work block starts with a choice among K similar tasks. Choosing takes
# T = a + b log2(K + 1) minutes (Hick-Hyman form, a = 5, b = 10), and
# every 10 undecided minutes the person drifts off with probability 0.15
# for the rest of the block. Aids: a card (or a coin) names one task and
# is followed with probability c, else the full choice is made after 2
# lost minutes; an if-then plan written the evening before is followed
# with probability 0.8. 25 blocks a week. All parameters illustrative.
# Reference points, not inputs: implementation intentions d = 0.65
# (Gollwitzer and Sheeran); choice overload mean D = 0.02 (Scheibehenne).
import math

A, B, DRIFT = 5.0, 10.0, 0.15

def p_start(k, lost=0.0):
    t = A + B * math.log2(k + 1) + lost
    return (1 - DRIFT) ** (t / 10)

def with_pointer(k, c):
    return c * p_start(1) + (1 - c) * p_start(k, 2.0)

def with_plan(k, adh=0.8):
    return adh * (1 - DRIFT) ** (A / 10) + (1 - adh) * p_start(k, 2.0)

if __name__ == "__main__":
    print("   K   free choice   card/coin c=0.5   c=0.8   if-then plan")
    for k in (2, 6, 12):
        print("   {:2d}      {:.3f}          {:.3f}         {:.3f}      "
              "{:.3f}".format(k, p_start(k), with_pointer(k, 0.5),
                              with_pointer(k, 0.8), with_plan(k)))
    base = 25 * p_start(6)
    card = 25 * with_pointer(6, 0.8)
    plan = 25 * with_plan(6)
    print("   K=6, blocks started a week: free {:.1f}, card {:.1f} "
          "(+{:.1f}), plan {:.1f}".format(base, card, card - base, plan))
#    K   free choice   card/coin c=0.5   c=0.8   if-then plan
#     2      0.713          0.737         0.765      0.876
#     6      0.584          0.675         0.740      0.851
#    12      0.505          0.636         0.725      0.835
#    K=6, blocks started a week: free 14.6, card 18.5 (+3.9), plan 21.3

有 6 个相近任务时,自由选择让一个工作时段顺利开始的概率是 0.584。一张牌或一枚硬币指定一个任务、80% 的时候被遵从,概率升到 0.740;前一晚写下的「如果-就」计划,升到 0.851。按每周 25 个时段计,不借助任何东西开始了 14.6 个,用牌 18.5 个(+3.9),用计划 21.3 个。模型没有赋予牌任何特殊力量:遵从率相同的硬币给出同样的数字。每日一张牌带来的任何行为增益,都应归功于缩小选择与承诺,而这两点一份普通计划做得更好。

随机指针是不好的决策规则吗

随机挑选听起来不理性。但当选项价值接近、思考又要花时间时,它并非不理性。在下一个模型中,六个任务的真实价值相差一个幅度 ss(以有效工作小时计),思考越久估计越准,但每分钟要花掉六十分之一小时。

推导

# Chapter 6. p12: When is a random pointer as good as deliberating?
# K = 6 tasks have true values drawn from Normal(0, s) (value in hours
# of good work). Deliberating for D minutes gives each a noisy estimate
# with error sd 1/sqrt(1 + D/10); the best estimate is chosen. Each
# minute spent costs 1/60 hour. A random pointer picks any task at no
# cost. 20,000 trials per cell, seed 89; mean net value is reported.
import math
import random

def net(s, D, rng, trials=20000, k=6):
    tot = 0.0
    err = 1 / math.sqrt(1 + (D or 0) / 10)
    for _ in range(trials):
        v = [rng.gauss(0, s) for _ in range(k)]
        if D is None:
            tot += v[rng.randrange(k)]
            continue
        est = [x + rng.gauss(0, err) for x in v]
        tot += v[est.index(max(est))] - D / 60
    return tot / trials

if __name__ == "__main__":
    rng = random.Random(89)
    print("   spread s   random   D=5 min   D=15 min   D=45 min")
    for s in (0.1, 0.3, 1.0):
        row = [net(s, None, rng)] + [net(s, D, rng) for D in (5, 15, 45)]
        print("   {:6.1f}   {:7.3f}   {:7.3f}   {:8.3f}   {:8.3f}".format(
            s, *row))
#    spread s   random   D=5 min   D=15 min   D=45 min
#       0.1     0.001    -0.069     -0.231     -0.720
#       0.3    -0.005     0.044     -0.088     -0.532
#       1.0    -0.005     0.893      0.823      0.413

当任务几乎相等时(s=0.1s = 0.1),随机挑选净得 0.001 小时,而思考 5 分钟净得 -0.069,思考 45 分钟净得 -0.720:思考只会亏。s=0.3s = 0.3 时,短暂思考略有收益(0.044),长时间思考则亏损。只有任务真正有差别时(s=1.0s = 1.0),思考才明显胜出:思考 5 分钟得 0.893,随机挑选得 -0.005。随机指针对「几乎打平」的选择是好规则,对真正的选择是坏规则;牌只应该用于前一种。

注

不主张预知。 本章的行为学论证不需要假定牌有任何含义,同样适用于骰子、硬币或编号清单。这正是它的长处:可以拿这些对照来检验(第八章)。

小结

在示意模型中,每日一张牌、80% 的时候被遵从,把一个工作时段开始的概率从 0.584 提高到 0.740,与硬币相同,且低于「如果-就」计划(0.851)。随机指针在选项几乎相等时是好规则,在选项有差别时是坏规则。

参考文献

Gollwitzer, Peter M., and Paschal Sheeran. 2006. “Implementation Intentions and Goal Achievement: A Meta-Analysis of Effects and Processes.” In Advances in Experimental Social Psychology, vol. 38. Academic Press.
Hick, William E. 1952. “On the Rate of Gain of Information.” Quarterly Journal of Experimental Psychology 4 (1): 11 to 26.
Scheibehenne, Benjamin, Rainer Greifeneder, and Peter M. Todd. 2010. “Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload.” Journal of Consumer Research 37 (3): 409 to 425.

Fan, Y. (2026). Dream Motif Bipartite Graphs, Deterministic Oracle Decks and Subconscious Prior Activation: REM Semantic Spreading, SHA-256 Fisher-Yates Permutations and Retroactive Coincidence Rates [梦境意象二分图、确定性哈希牌卡与潜意识先验激活] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090233 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).