学生论文
|
论文查询结果 |
返回搜索 |
|
|
|
| 论文编号: | 3389 | |
| 作者编号: | 1120070738 | |
| 上传时间: | 2011/6/10 10:27:07 | |
| 中文题目: | 信息瀑布的实验研究 | |
| 英文题目: | Experimental Study on Information Cascade | |
| 指导老师: | 李建标 | |
| 中文关键字: | 信息瀑布 钝化信念 贝叶斯法则 实验 | |
| 英文关键字: | Information Cascade Passivated Belief Bayes rule Experiment | |
| 中文摘要: | 序贯决策中,若最初几个人公布的决策结果一致,后续决策者就可能直接选择前人的决策结果,从而违背自己的私人信息,这种现象被称为信息瀑布行为。信息瀑布行为存在于社会生活的各个方面,网络集体事件、金融危机等群体一致行为的形成过程中,总能找到信息瀑布的踪迹,信息瀑布行为一旦被触发,往往发展迅猛且对其所依存的环境极具颠覆性。对信息瀑布问题的关注源于传统经济学对群体行为解释的无力,信息瀑布理论通过刻画序贯决策中个体信念的更新机理,预测了群体一致行为的形成过程。信息瀑布理论在填补传统经济学对群体行为研究不足的同时,也引发了人们对群体行为“理性”的思考。 1992年信息瀑布概念提出以来,这一理论正处在不断深化和完善的过程中。直接探讨信息瀑布发展机理与行为特征的研究不多。对信息瀑布中决策者行为特征、效率和动机的探讨,无论是增强人们对信息瀑布概念的认识和理解、完善理论体系,还是为决策者有效控制与合理疏导信息瀑布行为提供理论指导等方面均具有十分重要的意义。 本文在行为分类、理论框架和基本假设三个层面进行深化,并提炼了三个关键科学问题:第一,序贯决策中个体决策者行为类型体系的重构以及行为界定标准的建立;第二,信息瀑布中个体决策动机的测度以及动机测度实验框架的构建;第三,个体对不同信息赋予决策权重参数的测度以及参数测度方法的选择。论文沿着行为分析进路、运用实验经济学方法紧紧围绕以上问题,依次展开探讨。 第一个问题属于行为现象层面,现有信息瀑布研究对个体决策行为的界定缺乏系统性,本文按照个体行为是否遵循贝叶斯法则重构了行为分类体系,将序贯决策中全部决策行为界定为贝叶斯法则决策(信息瀑布属于该决策中的一种行为)、非贝叶斯法则决策和平衡决策三大类,基于此探讨了单决策信息瀑布中决策行为特征。这个层面的研究是在现有信息瀑布理论框架下的探讨。 第二个问题属于理论框架层面,针对序贯决策中个体决策动机问题,探讨了多决策信息瀑布中的个体决策动机与效率,为凯恩斯“动物精神”提供实验证据的同时,将现有“单决策信息瀑布”框架扩展到“多决策信息瀑布”框架,使理论更加符合实际情况。这个层面的研究突破了现有的信息瀑布理论框架。 第三个问题属于理论假设层面,本文探讨了信息瀑布中信息的决策权重,基于前景理论权重函数的思想,将不同信息“等决策权重”的现有研究假设推进到“异决策权重”假设,在实验基础上构建了信息瀑布“不同信息异权重模型”。基本假设层面上的修正和探讨,都将对信息瀑布理论的发展带来较为深远的影响。 此外,基于决策者“信息到信念传递敏感度降低”的特征,原创了“钝化信念”概念,提供了信息瀑布行为的识别标准,拓展了2002年诺贝尔经济学奖得主史密斯的“共同信念形成条件”的已有研究。论文主要创新如下: 1.将不同信息的“等决策权重”假设推进到“异决策权重”假设,修正了现有信息瀑布理论的基本假设; 2.将信息瀑布实验研究由现有的“单决策框架”拓展到“多决策框架”; 3.按贝叶斯法则将全部决策者行为界定为贝叶斯法则决策、非贝叶斯法则决策和平衡决策,重构了行为分类体系; 4.基于决策者“信息到信念传递敏感度降低”的特征,原创钝化信念概念,提供了信息瀑布行为区分标准。 论文主要结论如下: 1.实验结果证伪了现有信息瀑布理论的“不同信息等决策权重”假设。私人信息被赋予较高的决策权重,尤其是在损失框架下。实验数据支持了“信息瀑布触发和形成的阈值高于理论预测值”的推测; 2.信息瀑布决策中个体的行为动机是“动物精神”,而非个体策略理性。多决策信息瀑布实验中,第二次决策比第一次决策更容易发生信息瀑布,但决策正确率并不比第一次高; 3.单决策信息瀑布实验框架下,决策者的行为方式和效率会受到备选集对称性、损益框架、决策序位等因素的影响,与累积损益、性别等因素无关。 论文研究了信息瀑布中决策者的偏好和信念,讨论了群体一致行为特征,为信息瀑布行为的监管和治理提供理论基础。 | |
| 英文摘要: | An informational cascade occurs when it is optimal for an individual to follow the behavior of the preceding individual without regard to his own information after having observed the actions of those before him. Information cascade exists in social life widely. Information cascade plays an important role in many kinds of collective conformity behavior, such as network collective events and financial crisis. Once information cascade occurs, it will develop rapidly with destructive power. The information cascade event was concerned because of the weakness of explanation on collective behavior by traditional economics. The information cascade theory could predict the process of collective conformity behavior by depicting the individual belief updating in sequential decision making. The information cascade theory extended the traditional economics theory on collective behavior, but also led people to think over on the "rational" of collective behavior. Since the concept was introduced in 1992, information cascade was still on the stage of theory discussion. Researches on the development mechanism and behavioral characteristics directly were still not rich. The existing information cascade theory is poor on systemic. The studies on information cascade play an important role in understanding concept of information cascade, perfecting theory system, effective control and reasonably counseling information cascade behavior, and so on. The existing theory has three level of inadequate. All of them were refined to three key scientific topics in this dissertation: First, determined standard established and behavioral system reconstructed; Second, the measurement of decision-making motivation and the construct of experimental design for the measurement; Third, measurement of decision making weight parameters and the selection of measurement methods. All of the three questions were studied in this dissertation using behavior analysis approach and experimental economics method. The first question was about the behavioral phenomenon. Since the existing research ill-defined the behaviors, all the decision behaviors were reclassified into three types: Bayes Rule Decision (information cascade is one of this type behaviors), Non Bayes Rule Decision and Balance. Base on that, the characteristics of behavior were discussed. The study on this question was under the framework of existing information cascade theory. The second question was about theoretical framework. The one shot decision framework was extended to multi shots framework in order to discuss the decision motivation of individuals in the information cascade experiments. The motivation and efficiency in the multi-shots experiments were studied. Keynes’ “animal spirits” theory was supported by the evidence from the laboratory. This level of research has renewed the existing theoretical framework of information cascade. The third question was about theoretical assumptions. The decision weights in the information cascade were discussed in this dissertation. "Equal weight of decision-making towards different information" was advanced to "different weight of decision-making" hypothesis according to the Weighting Function of Prospect Theory. Any amendments and discussion on the basic assumptions level will bring further development on information cascade theory development. In addition, Base on the opinion of information passed to the belief is reducing; the concept of passivated beliefs was introduced firstly. The method for distinguishing information cascade to other behaviors was introduced. “The forming condition of common belief” by V. L. Smith (2002 Nobel Prize Winner) was extended in this dissertation. The main innovations of this dissertation are as follows: 1. Advanced "equal weight of decision-making towards different information" to "different weight of decision-making" hypothesis. In this sense, the fundamental assumption of information cascade was amendmented in this dissertation; 2. Expanded “one shot decision-making framework” of information cascade experiments to “mutli shots”; 3. Classified all the decision behaviors into three kinds: Bayes Rule Decision, Non Bayes Rule Decision and Balance; 4. Base on the opinion of information passed to the belief is reducing; the concept of passivated beliefs was introduced firstly. The method for distinguishing information cascade to other behaviors was introduced. The key conclusions of this dissertation are as follows: 1. The assumption of equal weight value of decision-making towards different information was falsificated by evidences from the laboratory. Private information was given higher weight in decision-making, especially in the loss domain. The speculation, the threshold of information cascade occurring is higher than the theoretical prediction, was supported by experimental data; 2. The motivation of individuals in decision-making is the "animal spirits" rather than rational individual strategies. The second time decision making is more like to follow an information cascade than the first time, but the accuracy of the second time decision was not better off than the first time in multi-shots information cascade experiment. 3. Symmetry & asymmetry, win & loss and decision-order have significant relationship with decision characters and accuracy in one shot information cascade experiment, while cumulative profit and gender do not. This dissertation will go through the agents’ preference and belief in information cascade behavior. It could provide theoretical basis on regulatory and management over collective conformity behavior. | |
| 查看全文: | 预览 下载(下载需要进行登录) |