论文部分内容阅读
Celen和Kariv(2004a)针对不完美信息下序贯决策的社会学习场景,建立连续信号-离散行动的贝叶斯学习模型,但该模型不能很好地解释全部样本的实验结果。本文尝试通过实验进一步考察不完美信息下序贯决策的个体行为是否符合贝叶斯推断。通过设计两种不完美信息(“仅见行动”和“仅闻建议”)下序贯决策实验,对广义贝叶斯模型进行了估计检验,检验结果表明在仅有决策者的前一个人的行动信息下,序贯决策者不可能依贝叶斯模型推断自己的最优解;决策行为大致可以用正态分布描述;处于前后不同决策顺序位置的个体行动选择的分布没有显著差异,他们只能简单地进行一步贝叶斯推断。
Celen and Kariv (2004a) set up a Bayesian learning model for continuous signal-discrete operations for a series of social learning scenarios with imperfect information, but the model does not explain the experimental results well for all samples. This paper tries to further investigate whether the individual behavior of sequential decision under imperfect information is in accordance with Bayesian inference through experiments. By designing sequential decision experiments under two imperfections ( “see only action ” and “smell only suggestion ”), we test the generalized Bayesian model. The test results show that in the former Under the action information of a person, it is impossible for a sequential decision-maker to deduce his optimal solution according to the Bayesian model. The decision-making behavior can be roughly described by a normal distribution. There is no significant difference in the distribution of individual action choices between different decision- , They can only make one-step Bayesian inference simply.