概率扭曲的动态神经编码机制

Humans distort probability information in many judgment and decision-making tasks in a similar way. That is, subjective probability deviates systematically from objective probability, with the typical pattern of overestimation of small probability and underestimation of large probability. Probability distortion in different tasks can be well captured by the same linear transformation of log-odds with two parameters, implying that it may arise from some general cognitive process. Though the phenomenon has received much attention, its origin is still largely unknown. The proposed project is intended to reveal the dynamic neural coding underlying probability distortion. In three studies, we will investigate two representative behavioral tasks—judgment of relative frequency and decision under risk—and use Electroencephalography and Magnetoencephalography to record the brain activities of human subjects during their encoding of probability information. Combining the steady state response, temporal response function, and computational modeling analyses for the neural data, we will answer the following questions: (1) Does a coding of the distorted subjective probability exist from the very beginning, or transformed from that of objective probability in a later stage? (2) Are there any neural codes for intermediate variables? (3) How may the brain regions for probability coding evolve with time? (4) What common or different neural mechanisms may underlie the probability distortion in different domains? Our findings will provide theoretical foundations for the development of training procedures that help people to reduce probability distortion and make better decisions.

人类在各种判断和决策任务中都表现出相似的概率扭曲，即主观概率与客观概率的系统偏离，典型模式是高估小概率、低估大概率。不同任务中的概率扭曲可以被同一个双参数的方程精确描述，暗示概率扭曲可能源于某种普适的认知加工过程。虽然概率扭曲现象广受关注，其产生机制却尚属未知。本项目力图揭示概率扭曲背后的动态神经编码过程。在三个研究中，我们将针对两种代表性的判断和决策任务，用脑电图和脑磁图记录人类被试编码概率信息时的神经活动。通过对神经数据进行稳态响应、时间响应函数和计算建模分析，我们将回答下列问题：（1）对扭曲的主观概率的编码是初始就存在，还是在加工过程中由客观概率的编码转化而来？（2）有哪些中间阶段的神经编码？（3）编码概率信息的脑区在加工进程中如何演化？ （4）不同领域中概率扭曲的神经机制有何异同？本项目的研究成果将为干预和改变概率扭曲构建理论基础，帮助人们在经济生活中更好地作出决策。

人类在各种判断和决策任务中都表现出相似的概率扭曲，即主观概率与客观概率的系统偏离，典型模式是高估小概率、低估大概率。不同任务中的概率扭曲可以被同一个双参数的方程精确描述，暗示概率扭曲可能源于某种普适的认知加工过程。虽然概率扭曲现象广受关注，其产生机制却尚属未知。在本项目中，我们针对两种代表性的判断和决策任务开展了一系列研究，结合行为实验、计算建模以及脑电图和脑磁图，来理解概率扭曲产生的认知和神经基础。主要发现包括：看似非理性的概率扭曲现象可能是大脑在有限的认知资源下优化信息传递的结果；概率扭曲的神经基础之一是大脑对概率信息的表征不确定性的自动而快速的编码。这些发现提示，概率扭曲与认知资源和不确定性之间存在密切关系；这为干预和改变概率扭曲、帮助人们在经济生活中更好地作出决策提供了理论基础。