随机序列生成中赌徒谬误的神经学习机制

The Neural Learning Mechanism of the Gambler’s Fallacy Bias in Random Sequences Generation

摘要: 赌徒谬误指人们在产生随机序列时更偏好于使用交替模式，即认为在类似于投无偏硬币事件中，如果出现了硬币某一面，那么接下去则更可能出现另一面，而不是继续出现同一面。赌徒谬误一般被认为是一种人脑对随机序列的错误知觉，是由于“表征偏见” (representativeness bias)引起的。但Sun等(2015)发现即使在一个p = 0.5的经典随机过程中，也存在一定的潜在结构(latent structure)则表明赌徒偏见的神经基础可以通过神经网络模型来解释。这一结果为赌徒谬误偏见提供了一种理性的解释，也为人脑对不确定性和随机性的认知过程提供了一种新的数学描述。

Abstract: The gambler’s bias refers to the preference for alteration patterns when people generate random sequences. That is, for example, when tossing fair coins, if one side appears, people would prefer to anticipate the other side to be more possible to appear next. The gambler’s bias is generally thought to be human brain’s misperception of random sequences which results from the repre-sentativeness bias. However, Sun et al. (2015) uncovered the latent structure in random sequen- ces, and provided a neural learning mechanism for the gambler’s bias using this statistical struc-ture. This finding not only gives a rational explanation for the bias but also provides a mathematical description for the cognitive processing of uncertainty and randomness in human mind.