二进制信号模型：序列个体行为采纳决策机制的技术分析问询

Alright, let's break down this sequential decision-making problem step by step—this is a classic information cascade (herding) setup, so we'll start by grounding ourselves in the core rules, then walk through each individual's logic based on their position in the queue. First, I need to clarify a standard implicit assumption of this binary signal model (since it's critical to the behavior):

Core Setup (Including Standard Implicit Rule) #

First, let's restate your given parameters clearly, plus the key unstated assumption that makes this model meaningful:

• Sequential Individuals: Everyone decides in a fixed, publicly known order, and each can see all prior decisions (adopt/reject).
• Cost & Payoff: Adopting costs C = 1/2. The true value of the behavior V is either 0 or 1, with equal prior probability (P(V=1) = P(V=0) = 1/2).
• Private Signals (Critical Implicit Rule): Each individual gets a private binary signal s ∈ {0,1} with accuracy p > 1/2. That means:
  • If V=1, the chance of getting s=1 is p, s=0 is 1-p
  • If V=0, the chance of getting s=0 is p, s=1 is 1-p
    All individuals know this accuracy rate, and signals are independent across people.

Decision Logic by Position #

Let's go through each individual's reasoning one by one:

Position 1 (First Person) #

This person has no prior decisions to reference, so they rely solely on their private signal and the prior probability:

• If they get s=1:
  Using Bayes' theorem, P(V=1|s=1) = (p × 1/2) / (p × 1/2 + (1-p) × 1/2) = p
  Expected net payoff: p × 1 + (1-p) × 0 - 1/2 = p - 1/2 > 0 (since p>1/2), so they adopt.
• If they get s=0:
  P(V=1|s=0) = 1-p, expected net payoff is (1-p) - 1/2 = 1/2 - p < 0, so they reject.
  In short: The first person's decision perfectly reveals their private signal.

Position 2 (Second Person) #

They can see the first person's decision, plus have their own signal:

• If Person 1 adopted (so s₁=1) and Person 2 gets s₂=1:
  P(V=1|s₁=1,s₂=1) = p² / (p² + (1-p)²) > p > 1/2, net payoff positive → adopt.
• If Person 1 adopted and Person 2 gets s₂=0:
  P(V=1|s₁=1,s₂=0) = (p(1-p)) / (p(1-p)+(1-p)p) = 1/2, net payoff is 0. We typically assume they reject in this indifference case.
• If Person 1 rejected (so s₁=0) and Person 2 gets s₂=0:
  P(V=1|s₁=0,s₂=0) = (1-p)² / ((1-p)² + p²) < 1/2, net payoff negative → reject.
• If Person 1 rejected and Person 2 gets s₂=1:
  P(V=1) = 1/2, net payoff 0 → reject (or random, but reject is standard).

Position 3 (Third Person) #

This is where the magic (and problem) starts—information cascades kick in:
Suppose the first two people both adopted (meaning s₁=1, s₂=1). Now, no matter what signal Person 3 gets:

• If Person 3 gets s₃=1: Obvious to adopt, same as Position 2.
• If Person 3 gets s₃=0:
  Calculate P(V=1|s₁=1,s₂=1,s₃=0) = (p²(1-p)) / (p²(1-p) + (1-p)²p) = p / (p + (1-p)) = p > 1/2
  Expected net payoff is still positive, so Person 3 will ignore their own signal and adopt.
  The same logic applies if the first two rejected: Person 3 will reject even if they get s=1, because the prior two signals outweigh their own.

Positions ≥4 (All Subsequent People) #

Once we have two consecutive identical decisions (two adopts or two rejects), every person after that will completely ignore their private signal and follow the crowd:

• For two prior adopts: No matter their signal, the updated P(V=1) is always >1/2 → adopt.
• For two prior rejects: Updated P(V=1) is always <1/2 → reject.
  This is the information cascade: the group locks into a decision, even if it's wrong (e.g., if the first two people happened to get incorrect signals).

Key Takeaways #

• Position Matters: The earlier you are in the queue, the more your decision influences the entire group. Later individuals have no incentive to use their own information.
• Cascade Trigger: Two identical consecutive decisions are the tipping point—after that, herding is inevitable.
• Potential for Error: The cascade could be based on wrong initial signals, leading the entire group to a suboptimal decision.

内容的提问来源于stack exchange，提问作者Jess