【作者】 沈强；

【导师】 马庆国；

【作者基本信息】 浙江大学 ， 管理科学与工程， 2011， 博士

【摘要】 决策神经科学(decision neuroscience)是一门非常新兴的前沿学科,是管理科学新的学科生长点。它是在传统的行为决策科学研究的基础上发展起来的一门交叉学科,该学科试图利用神经科学的成像技术,结合认知科学和心理学的最近进展,与管理科学中的决策科学相结合,来研究人的决策行为背后的加工机制,打开大脑这个决策加工的“黑匣子”。通过对大脑水平决策加工过程的解读,在神经科学层面理解决策加工过程的机制,来重新审视传统的行为决策模型,为重塑决策模型的理论框架与规范基础提供实证依据。风险决策是决策科学研究的一个重要主题。2002年,行为决策科学家丹尼尔·卡尼曼因其在风险和不确定决策领域的重要贡献而获得了当年的诺贝尔经济学奖。而在决策神经科学领域,2007年,亚利桑那大学决策神经科学实验室主任Alan Sanfey在《决策神经科学：判断与决策研究的新方向》一文中,提出决策神经科学的两大主题：一是对个人决策机制的研究；二是对社会环境中的个体之间互动的研究；其中对个体决策机制最重要的研究领域就是对风险决策的研究。本研究以风险决策的加工处理过程为研究对象,利用功能核磁共振(functional magnetic resonance imaging, fMRI)和事件相关电位(event-related potential, ERP)两种神经成像技术为研究手段,开展了两个实验研究,来观察风险决策处理过程的神经基础。充分利用功能核磁共振高的高空间分辨率和事件相关电位的高时间分辨率特点来研究风险决策的预期、决策与反馈评估脑神经活动特征。本研究发现,预期情绪在决策的三个阶段——决策的预期、决策和结果的反馈中都起到了重要的作用,但是在三个阶段所起的作用又存在着差异,三个阶段的加工过程也存在不同。本研究在过去相关的决策科学、认知科学、神经科学和最近的行为决策科学等的研究成果的基础上,根据核磁磁共振空间定位和事件相关电位时间进程的信息,结合行为学数据的结果,从决策神经科学这一交叉学科的视角深入探讨了风险决策的加工机理,得到了如下的研究结论：(1)在风险决策的预期阶段,存在着预期的决策情感。利用功能核磁共振发现,无论是在获益还是损失线索下,都由奖赏相关的脑区中脑、腹、背侧纹状体的激活和惩罚相关脑区岛叶的同时激活来进行表征,而且随着线索唤醒度水平的增加,奖赏、惩罚脑区的信号也相应增加,体现了对未来大获益(小损失)的期望与小获益(大损失)的担心交织并存的预期情感状态,反映了决策者在决策预期阶段的决策动机。(2)在风险决策的反馈阶段,反馈结果的评估对效价敏感。与在决策预期阶段正负线索有类似的大脑激活模式不同,大脑对正负反馈不同的结果产生不同模式的应答,奖赏反馈结果主要由内侧前额叶和腹侧纹状体来进行表征,而损失反馈结果主要由岛叶和扣带回来体现。并且在预期阶段由腹、背侧纹状体的同时激活来表征对预期可能结果的期望,而在反馈阶段在正性反馈结果下由腹侧纹状体来表征预测奖赏偏差,说明预期与反馈阶段的加工是分离的。(3)与预期阶段存在预期的决策情感一样,在风险决策的决策阶段也存在着决策情感,这种情感帮助决策者协调决策情绪,高效、准确地完成风险决策的决策选择。利用事件相关电位发现在决策阶段错误相关负波(error realted negativity, ERN)与错误正波(error positivity, Pe)都参与了对风险情形和投资决策选择的调节,其中ERN反映了对未来可能风险与奖赏之间权衡利弊的预期情感评估,而之后的Pe成分反映了对决策阶段决策选择好坏的评价。起源于大脑前扣带回的ERN在决策阶段产生的对未来的预期情感使得决策者能够做出适应性的决策选择(adaptive choice),而且从大脑双系统理论的视角来看,这种加工可以在大脑意识水平下的系统1即可完成。(4)在风险决策的反馈阶段,对反馈结果的评估反映了决策者对决策预期的再评估,为决策的学习和下一次决策行为的调整做好准备。与决策阶段不同的是,与ERN一样同样起源于大脑前扣带回的反馈相关负波(feedback related negativity, FRN)没有参与对风险情形的调节,但是其对决策阶段选项概率大小非常敏感。决策阶段决策选项的概率越大,在反馈评估阶段诱发更加明显的FRN,反映了反馈阶段对决策阶段的决策预期与反馈实际结果之间差异的表征,即表征了预测情感偏差,与功能核磁共振研究中反馈阶段腹侧纹状体激活的研究结论相一致。本研究的创新点如下：(1)本研究发现预期阶段预期情感的存在,而且这种情感反映了对未来可能结果的动机。本研究利用功能核磁共振技术验证了过去Knutson & Greer提出的预期情感假说的基本结论,并发现在预期阶段大脑奖赏系统和惩罚系统同时参与了对获益和损失线索的加工,而且是唤醒度敏感的,反映了对不确定后果的担心和期望并存的预期情感状态,这种状态反映了在决策预期阶段的决策动机。这一结论进一步延伸了过去的研究报告奖赏系统如伏隔核的激活可以预测接下来的风险偏好的决策行为,而惩罚系统如岛叶的激活可以预测接下来的风险规避行为的研究结论。(2)本研究发现在决策阶段大脑的决策加工反映了决策者的适应性选择。本研究利用事件相关电位发现ERN参与了对风险决策情形与风险投资选择的动态加工,反映了对未来可能风险与奖赏权衡的预期情感评估,指引决策者在动态的风险决策环境中作出有利于自身的决策,为金融学的均值—方差模型提供了神经学水平的实证依据。起源于前扣带回的ERN对决策选择的表征说明情绪在个体决策者高效的决策中起到的关键作用,为Damasio提出的情绪对决策影响的躯体标记学说提供了支持,也从神经科学的层面验证了Kahneman总结的大脑决策加工的双系统模型,并说明其中的系统1在准确、快速的风险决策中起到了重要作用。(3)预期情感在风险决策的预期、决策的执行和反馈评估三阶段中都起到了重要作用。如创新点1所述,预期情感在决策的预期阶段起到了重要作用,反映了对未来奖惩结果的预期动机；而在决策的执行阶段前扣带回(通过ERN来反映)作为一个协调情绪的重要脑区,参与了对未来预期风险与奖赏之间协调的表征。而在反馈阶段,前扣带回(通过FRN来反映)参与了对预期违反的表征。因此,决策的预期情感参与了风险决策的预期阶段、决策阶段和反馈评估阶段的加工。预期情感不只是在预期阶段参与了对未来结果的预期和表征,同样在风险决策的决策阶段与反馈评估阶段起也扮演者重要的角色,进一步丰富了基于决策神经科学研究提出的预期情感假说的研究内容。(4)预期阶段、决策阶段与反馈阶段的加工存在差异,预期效用与体验效用是分离的,同样,决策效用与体验效用也是分离的。首先在预期阶段获益和损失的线索都表现为奖赏和惩罚系统的同时激活,反映了决策的动机；而在反馈阶段负性和正性的反馈结果有明显的区别,前者有奖赏系统的参与,而后者由惩罚相关脑区来表征,说明预期效用与体验效用是分离的。在决策阶段主要由协调奖赏和惩罚情绪的前扣带回来进行表征,反映对未来风险与奖赏预期的适应性表征；反馈阶段同样由起源于前扣带回的FRN参与对反馈结果表征,但是其与决策阶段ERN参与对风险情形的调节不同,FRN对风险情形不敏感,而是参与了对预期违反的表征。因此决策效用与体验效用也是分离的。从空间和时间的两个视角在神经科学的水平证实了Kahneman等提出的不同阶段的效用理论。更多还原

【Abstract】 Decision neuroscience is an emerging frontier discipline and a new growing point in the field of management science. It is an interdisciplinary field that derives from the recent development of the research in traditional behavioral decision making. By combining with the recent advance of the decision science in management science, cognitive science as well as psychology, decision neuroscience trys to adopt the neuroimaging technique to study the processing mechanism of human being’s decision behavior, thus opening the "black box" of the brain. By virtue of understanding the neural processing of the decision making to slove the mechanism of decision making on the level of neuroscience, decision neuroscience will revisit the traditional behavioral decision making model and provide some empirical evidence for the reconstruction of the theoretical framework and normative basis of the decision making model.Risk decision making or decision making under risk is an important theme of decision science research. Behavioral decision making scientist Daniel Kahneman won the 2002’s Nobel Laureate for his great contribution to the risk and uncertainty study. In the field of decision neuroscience, the director of the decision neuroscience lab in Unviersity of Arizona, America put forward in the "Decision neuroscience: New directions in studies of judgment and decision making" that there are two topics that deserve exhaustive exploration, one is the individual decision making and the other is social decision making, and the risk decision making is the main topic for individual decisions.The current studies take the processing procedure of the risk decision making as an object and utlize two neuroimaging techniques, the functional magnetic resonance imaging and even-related potential as the methods. These studies trys to observe the neural basis of the risk decision making by carrying out two experimental studies. The studies make full use of fMRI that possesses a high spatial resolution and the ERP with a high temporal resolution to explore the neural characteristics of the three stages of risk decision making—anticipation, decision making and outcome evaluation. This study found that the anticipation affect plays a key role in all three stages:anticipation, the decision choice and outcome evaluation. But the role it plays is not the same across the three stages and the processing procedures of these three stages per se are different from each other.The studies were carried out on the basis of the recent advance of decision science, cognitive science, neuroscience and etc. to intensively investigate the principle of the risk decision making according to the fMRI’s spatial data, ERP’s temporal information as well as the behavioral results.It draws the conclusion as follows:(1) At the stage of anticipation, the anticipatory affect exsits. With the method of functional magnetic resonance imaging, we observed that both gain and loss cues could induce the activation of reward-related regions including ventral, dorsal striatum and midbrain as well as the punishment-related areas insula at the same time. The signal of both reward and punishment areas increase with the raise of the cue levels. Such an activation mode indicate the complicated anticipatory affect consists of both the hope for large gain (small loss) and fear of small gain (large loss), reflecting the motivation of the decision maker.(2) At the stage of outcome evaluation, the brain is sensitive to the valence of the outcome. The positive and negative results have a discrepant activation, the medial prefrontal cortex and ventral striatum represent the win outcome, while the insula and cingulate cortex represent the loss results, different from the similar activation mode toward both gain and loss cues at the stage of anticipation. In addition, both the ventral and dorsal parts of the striatum is engaged in the process of the hope for possible gain in the stage of anticipation, but only the ventral striatum is involved in the positive outcome evaluation in representing the reward prediction error, further supporting the distinct processing of the anticipation and outcome evaluation.(3) Similar with the exsitence of the anticipatory affect at the stage of the anticipation, the decision affect exist at the decision stage, which helps the decision makers coordinate the affect to finish the decision making rapidly and effectively. With the adoption of the event-related potential, we observed both the ERN and Pe are involved in the modulation of the risky situation and investment choice. The ERN reflects the ancipatory affect evaluation that derives from the tradeoff the possible risk and reward in near future, and the subsequent Pe indicate the goodness or badness of the choice option. The ancipatory affect for the future that the ERN which orgined from anterior cingulate cortex conveys made the subject do the adaptive decisions. Additionally, such a processing is accomplished at the sysem 1 subconsciously from the perspective of the dual system hypothesis.(4) At the stage of outcome evaluation, the outome results reflect the re-evaluation of the decision anticipation, which makes a preparation for the learning and behavioral adjustment for the next round of decision making. Different from the processing that happened at the stage of decision making, although same as the ERN that is also orgined from ACC, the FRN is not engaged in the modulation of the risky situation,instead it’s sensitive to the probability of choice at the stage of decision making. The larger the probability of the choice at the stage of decision making, the more obvious the FRN is at the stage of outcome evaluation, which represents the discrepancy between the decision anticipation at the stage of decision and the revealed results at the outcome stage. Again, the FRN represent the reward prediction error, consistent with the conclusion draw from the fMRI’s ventral striatum activation at the outcome evaluation stage. The innovation points of this study are as follows:(1) In this study, we observe the existence of the anticipatory affect which reflects the motivaition toward the possible outcome in future. This study validates the basic conclusion of Knutson & Greer’s anticipatory affect hypothesis with the fMRI technique. This study observe that both the reward and punishment-related system simultaneously involved in the processing of the gain and loss cues and such an engagement is sensitive to the arousal of the cues. This activation mode reflects the anticipatory affect for the fear and hope for the uncertain outcome and further embodys the decision motivation of the anticipation stage. Such a conclusion further extends the recent reports that the activation of the reward related system such as the necleus accumbens could predict the risk taking behavior while the activation of the punishement related system such as insula could predict the loss aversion behavior.(2) In this study, we observe the neural processing at the decision stage reflect the decision makers’ adaptive choice. This study observe the ERN is involved in the dynamic processing of the risky decision situation and the risky investment choices, which reflects the ancipatory evaluation of the tradeoff of risk and reward, guiding the decision maker make the optimal choice in the dynamic risky settings. The guidance of the ERN that origned from the anterior cingulate cortex indicates that the affect play a key role at the efficient decision making of the individual decision maker, which gives a direct support for Damasio’s somatic marker hypothesis and dual system model that Kahneman summarized in behavioral decision making from the neuroscience level. So from the perspective of the dual system model of the brain, the study suggests that the system 1 is quite important in the quick and accurate decision makings.(3) The ancipatory affect plays an important role at all the three stage of decision making:ancipation, decision selection and outcome evaluation stage. As mentiond in the first innovation point, ancipatory affect play a vital part at the anticipation stage, reflecting the anticipation motivation toward future reward and punishment outcome. At the decision selection stage, the anterior cingulate cortex worked as an important part to coordinate the emotion to represent the anticipated risk and reward through ERN. At the outcome evaluation stage, anterior cingulate cortex represents the anticipatory violation through the FRN. Therefore, the ancipatory affect involves in all three stage of decision making. The anticipatory affect not only play a role at the stage of the anticipation, but also at the stage of decision making and outcome evaluation, further enrich the content of the anticipatory affect hypothesis put forward by the decision neuroscientists.(4) The processing of anticipation, decision making and outcome evaluation are different, not only the anticipated and experienced utility are dissociated, but also the decision and experienced one. First, at the stage of the anticipation, the gain and loss anticipation activate both the reward and punishment-related systems at the same time reflect the decision motivation. At the stage of outcome evaluation, the positive and negative outcome has respective activation; the former has the reward related regions engaged, while the later has the punishment related regions involved. Therefore, the anticipated and experienced utilities are not the same. The decision stage is mainly represented by the activation of the ACC to reconcile the reward and risk tradeoff, suggesting the adaptive representation of the decision making. The outcome evaluation stage is also represented by the ACC orgined component, FRN. But different from the ERN that is sensitive to the risky situation, the FRN is not sensitive to the risky situation, but merely response to the anticipation violation. Therefore, decision and experienced utilities are distributed. In general, from the spatial and temporal levels of neuroscience, this study validates the difference of the utility at different stage of decision that was initially put forward by Kahneman & Wakker.更多还原