贝叶斯推理中概率信息作用的认知神经机制
发布时间：2018-07-16 21:38
【摘要】：日常生活中人们经常碰到许多不确定的信息(即具有概率性质的信息),并在此基础上根据新信息或证据调整已有观点得出适当结论并做出决定,如“气象部门根据气候资料做出天气预报”、“医生根据检查和化验结果做出病情诊断”、“房屋购买者根据市场信息做出是否购房决策”等等,这一过程即贝叶斯推理过程。 到目前为止,有关贝叶斯推理的行为研究仍未得到一致的结论。其争论主要集中在以下两个方面：(1)人们在推理中是否遵循贝叶斯定理；(2)是什么原因导致了贝叶斯推理偏差?其核心的问题是,贝叶斯推理过程中基础比率和击中率这两种关键的概率信息,其作用机制是怎样的。关于这一问题,以往大多数研究是将后验概率估计值与贝叶斯规则计算的标准值相比较,从而得出结论；即使有研究谈及推理过程和反应时,其结论也大多是根据推理结果来间接地反推其推理过程,缺乏时间进程数据的支持。基于以往的研究,本研究设计两个实验来进一步探讨贝叶斯推理中概率信息的作用及其认知神经机制。 实验一采用三因素混合设计考察了基础比率、击中率信息和时间压力对贝叶斯推理的影响。结果表明,基础比率和击中率信息对后验概率估计值有显著影响,而时间压力对后验概率估计值的影响不显著。 实验二采用两因素被试内设计考察了基础比率和击中率这两个概率信息在贝叶斯推理中的作用,这是首次使用事件相关电位(ERP)技术探讨贝叶斯推理的实时进程,为进一步揭示贝叶斯推理的认知加工机制提供电生理学证据。结果表明,与高击中率的推理任务相比,低击中率的推理任务在额、顶叶诱发了一个更明显的N1(100-200ms),而在额中央、中央顶、顶枕联合区诱发了一个更明显的N300(250-350ms)；与低基础比率的推理任务相比,高基础比率的推理任务在额、颞、顶叶诱发了一个更明显的晚期正成分LPC(350-700ms)。 结论：本研究的实验表明在贝叶斯推理过程中,人们并没有遵循贝叶斯定理,而是运用了锚定-调整启发式。产生推理偏差的原因也并非忽视基础比率信息,而是在锚定-调整启发中存在的认知偏差和左右脑的认知功能的差异。同时,也表明推理过程主要激活了颞叶、额叶和顶叶区域,与演绎推理所激活的脑区一致。
[Abstract]:In daily life, people often encounter a lot of uncertain information (that is, information with probabilistic nature), and on this basis, according to the new information or evidence to adjust the existing views to draw appropriate conclusions and make decisions, such as "weather department based on climate data to make weather forecast", "the doctor based on the examination and test results to make a diagnosis of the disease." This process is called Bayesian inference process.
So far, the research on the behavior of Bias's reasoning has not been agreed. The debate is mainly focused on the following two aspects: (1) whether people follow the Bias theorem in reasoning; (2) what causes Bias's reasoning deviation? The core problem is that the basic ratio and hit rate in Bias's reasoning process are the key problem. What is the mechanism of the two key probability information; in the past, most of the previous studies have compared the standard values of the posterior probability estimation to the Bayes rule, and then draw a conclusion. Even if there is a study on the reasoning process and reaction, the conclusion is mostly based on the inference results to indirectly push back its push. Based on previous studies, two experiments were designed to further explore the role of probability information and its cognitive mechanism in Bayesian inference.
In Experiment 1, the influence of the basic ratio, hit rate information and time pressure on Bias's reasoning was investigated with three factor mixing design. The results showed that the basic ratio and the hit rate information had a significant influence on the posterior probability estimation, while the time pressure had no significant influence on the posterior probability estimation.
In experiment two, the role of the two probability information of basic ratio and hit rate in Bayesian inference was investigated with the two factor interior design. This is the first time to use the event correlation potential (ERP) technology to explore the real time process of Bayesian reasoning, and to provide electrophysiological evidence for further revealing the recognition processing mechanism of Bayesian inference. Compared with the reasoning task of the high hit rate, the inferential task of the low hit rate in the forehead, the parietal lobe induces a more obvious N1 (100-200ms), while in the middle of the forehead, the central top, the top occipital region induces a more obvious N300 (250-350ms); compared with the inferential task of the low base ratio, the reasoning task of the high base ratio is induced in the frontal, temporal, parietal lobe. A more obvious late positive component LPC (350-700ms).
Conclusion: This study shows that in the process of Bias's reasoning, people do not follow the Bias theorem, but use the anchoring and adjustment heuristic. The reason why the reasoning deviation is not ignored is not the basic ratio information, but the difference between the cognitive bias and the cognitive function of the left and right brain in the anchoring adjustment enlightenment. The reasoning process mainly activates the temporal lobe, frontal lobe and parietal lobe, and is consistent with the brain regions activated by deductive reasoning.
【学位授予单位】：湖南师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：B842.1
本文编号：2127739
[Abstract]:In daily life, people often encounter a lot of uncertain information (that is, information with probabilistic nature), and on this basis, according to the new information or evidence to adjust the existing views to draw appropriate conclusions and make decisions, such as "weather department based on climate data to make weather forecast", "the doctor based on the examination and test results to make a diagnosis of the disease." This process is called Bayesian inference process.
So far, the research on the behavior of Bias's reasoning has not been agreed. The debate is mainly focused on the following two aspects: (1) whether people follow the Bias theorem in reasoning; (2) what causes Bias's reasoning deviation? The core problem is that the basic ratio and hit rate in Bias's reasoning process are the key problem. What is the mechanism of the two key probability information; in the past, most of the previous studies have compared the standard values of the posterior probability estimation to the Bayes rule, and then draw a conclusion. Even if there is a study on the reasoning process and reaction, the conclusion is mostly based on the inference results to indirectly push back its push. Based on previous studies, two experiments were designed to further explore the role of probability information and its cognitive mechanism in Bayesian inference.
In Experiment 1, the influence of the basic ratio, hit rate information and time pressure on Bias's reasoning was investigated with three factor mixing design. The results showed that the basic ratio and the hit rate information had a significant influence on the posterior probability estimation, while the time pressure had no significant influence on the posterior probability estimation.
In experiment two, the role of the two probability information of basic ratio and hit rate in Bayesian inference was investigated with the two factor interior design. This is the first time to use the event correlation potential (ERP) technology to explore the real time process of Bayesian reasoning, and to provide electrophysiological evidence for further revealing the recognition processing mechanism of Bayesian inference. Compared with the reasoning task of the high hit rate, the inferential task of the low hit rate in the forehead, the parietal lobe induces a more obvious N1 (100-200ms), while in the middle of the forehead, the central top, the top occipital region induces a more obvious N300 (250-350ms); compared with the inferential task of the low base ratio, the reasoning task of the high base ratio is induced in the frontal, temporal, parietal lobe. A more obvious late positive component LPC (350-700ms).
Conclusion: This study shows that in the process of Bias's reasoning, people do not follow the Bias theorem, but use the anchoring and adjustment heuristic. The reason why the reasoning deviation is not ignored is not the basic ratio information, but the difference between the cognitive bias and the cognitive function of the left and right brain in the anchoring adjustment enlightenment. The reasoning process mainly activates the temporal lobe, frontal lobe and parietal lobe, and is consistent with the brain regions activated by deductive reasoning.
【学位授予单位】：湖南师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：B842.1
【参考文献】
相关期刊论文 前10条
1 刘金平;李红锋;;时间压力下的决策策略和决策理论[J];河南大学学报(社会科学版);2008年06期
2 姚双雁;钟毅平;;贝叶斯推理的影响因素研究述评[J];社会心理科学;2007年Z3期
3 张凤华;邱江;杨群;张庆林;;传递性推理的ERP研究[J];心理发展与教育;2009年04期
4 赵晓东,傅小兰;贝叶斯推理的改进方法——以频率格式代替概率格式进行信息表征[J];心理科学;2002年01期
5 张向阳;刘鸣;张积家;;主体关联性对贝叶斯推理概率估计的影响[J];心理科学;2006年04期
6 张向阳;刘鸣;张积家;;基础概率在贝叶斯推理中的作用[J];心理科学;2007年03期
7 史滋福;邱江;张庆林;;任务情境和证据信息对青少年贝叶斯推理的影响研究[J];心理科学;2008年01期
8 杨娟;邱江;张庆林;;传递性推理的认知与脑机制探讨[J];心理科学;2008年03期
9 王宝玺;向玲;邱江;张庆林;;条件推理中无效推断被抑制的ERP研究[J];心理科学;2008年05期
10 史滋福;张庆林;;“锚定参照错误”偏向对贝叶斯推理成绩的影响[J];心理科学;2009年02期
相关博士学位论文 前1条
1 杨莉;两种概率表征的贝叶斯推理的加工过程[D];江西师范大学;2007年
，本文编号：2127739
本文链接：https://www.wllwen.com/shekelunwen/xinlixingwei/2127739.html
