“君子博学而日参省乎己，则知明而行无过矣”，在日常生活中，每个人每天都会做出上百个决定，决策过程常常伴随着对决策的监控和调节，这种对自身认知的认知与调节在心理学中被称为“元认知”。元认知与决策之间存在紧密联系与相互作用，元认知不仅仅对决策进行评价，有效的元认知还会促进更好的决策。近期的研究发现，元认知对决策的监控存在偏差，对决策的调节也会导致决策的偏差，这说明元认知也会对决策产生负向的影响。因此，对这两种偏差的研究可以为决策与元认知的关系提供新的认识。

作为一种高级认知功能，元认知可能不是人类独有的能力，动物也可能具备与人类相似的元认知能力，探究非人灵长类动物的元认知不仅可以回答动物是否具备反思性思维的问题，还可以为人类元认知的研究提供动物模型。以往的研究提供了动物具备元认知能力的行为证据，但如何对比人类和非人灵长类动物的神经信号一直存在挑战，这很大程度上是因为对人类和非人灵长类的研究来自于不同的研究技术。功能磁共振成像技术能够弥合不同研究方法带来的差异，在定位大脑功能网络方面具有独特的优势。

本论文使用功能磁共振成像方法，从行为、神经机制和动力学模型三个方面系统性地探索了人类与非人灵长类的决策与元认知相互影响的认知神经机制。本论文从以下三个部分进行了四个研究：

第一部分，探索人类决策对元认知的影响。近期研究发现，元认知对决策的监控存在偏差，这可能是由决策机制引起的。个体对于决策正确性的主观信念，即决策信心，是元认知在决策中的体现。研究一的结果表明，在知觉决策和价值决策中，不同选项的证据以同等的权重影响决策的选择，但支持所做选择的证据对决策信心的影响权重显著地大于支持替代选择的证据。功能磁共振成像结果显示，在两种决策任务中，背侧前扣带回皮层均参与决策信心的表征，并且它的神经活动同样表现出对选择一致性证据的偏好，且偏好的强度与决策信心对选择一致性证据偏好的强度在被试间有显著的相关，这说明，背侧前扣带回皮层的神经活动与决策信心的选择一致性偏差有关。不同决策机制的动力学模型结果显示，两个选项决策证据的相互抑制和决策边界随着时间逐渐下降共同导致决策信心对选择一致性证据的偏好，这支持了这两种机制在决策过程中的存在。

第二部分，探索人类元认知对后续决策的影响。研究二探究在独立的决策序列中，先前的决策信心对后续决策和决策信心的影响，结果表明，无论在知觉决策还是在价值决策中，前一次的决策信心都会促使个体在下一次决策中更多地做出与前一次决策相同的选择并且报告更高的决策信心。在不同机制的动力学模型中，前一次的决策信心调节后续决策的初始边界的模型有最好的表现，这说明，边界更新是元认知调节后续决策的方式，前一次的决策信心以更大的权重下调与前一次决策的选择一致的选项的初始决策边界，从而导致个体的重复选择偏差。功能磁共振成像结果显示，初级运动区参与初始决策边界的表征，这说明运动区携带了决策相关的信息，它可能参与决定决策应该在何时终止。研究三探究独立决策序列中先前的决策信心对后续决策和决策信心的影响是否跨领域存在。交替决策实验结果表明，先前的决策信心只在后续的相同决策中加强个体的重复选择偏差，但在后续相同与不同的决策中都会引起更高的决策信心，这说明，元认知对后续决策的调节具有领域特异性。

第三部分，探索非人灵长类猕猴的决策与元认知的相互影响。研究四发现，在知觉决策中，猕猴的反应时间符合决策信心的行为特征，能够反映猕猴决策的正确性。功能磁共振成像结果表明，前额叶皮层和内顶叶皮层有可能是参与猕猴元认知功能的候选脑区。在决策序列中，元认知对后续决策的调节作用与人类具有一致的方向，先前的决策信心会促使猕猴在下一次决策中更多地做出与前一次决策相同的选择。

综上，本论文的研究发现：决策机制驱动了元认知在对决策的监控中表现出的对选择一致性证据的偏好，这是决策信心形成方式的一般特征；元认知以更大的权重下调了在后续决策中与先前决策选择一致的选项的初始边界，从而导致个体更容易重复先前的选择，这样的作用在人类和非人灵长类动物中具有一致性。这些研究结果表明元认知与决策是不可分割的，存在紧密的联系和相互作用，决策机制中需要包含元认知调控的过程。

“A noble person is learned and constantly reflects on oneself, thus gaining knowledge and acting without fault,” is a quote from Confucius. People make hundreds of decisions a day in their daily lives, and decision-making frequently involves self-monitoring and self-regulation. In psychology, such cognition about cognition is called “metacognition.” Metacognition and decision-making are interdependent. In addition to assessing decision-making, metacognition also facilitates better decision-making. However, recent studies have found that there are biases in metacognitive monitoring of decision-making and that metacognitive regulation can lead to decision biases, indicating that metacognition may possess negative effects on decision-making. Thus, it is worth studying these biases to provide new insights into the relationship between decision-making and metacognition.

Metacognition, as a higher-level cognitive activity, may not be unique to humans; animals may have similar metacognitive capacities. Exploring metacognition in non-human primates can yield answers to the question of whether animals have reflective thinking as well as animal models for investigating human metacognition. Previous studies have found behavioral evidence for animals' metacognitive capacities, but comparing neural signals between humans and non-human primates has proven difficult due to the use of different neural recording techniques. Functional magnetic resonance imaging (fMRI) could bridge the gap between different techniques and has unique advantages in localizing brain functional networks.

To systematically investigate the cognitive neural mechanisms underlying the interaction between decision-making and metacognition in humans and non-human primates, we combine the perspectives of behavior, neural mechanisms, and dynamical models using fMRI technology. This dissertation included four studies organized into three parts:

In the first part, we investigated how human decision-making influences metacognition. Recent studies have found confidence biases that may be caused by decision mechanisms. Confidence, or an individual’s subjective belief in the correctness of their choice, is a reflection of metacognition in decision-making. Study 1 showed that the participants’ confidence ratings were systematically biased towards the chosen option across both the perceptual and value-based decision-making tasks. By contrast, their choices were equally accounted for by the evidence of both options. We also found that confidence bias was not only present in behavior but also in neural activity. The dorsal anterior cingulate cortex (dACC) activities, which represented confidence commonly across both tasks, were also biased toward the chosen option. The strength of neural activity bias towards chosen option evidence in dACC significantly correlates with the strength of confidence bias towards chosen option evidence across participants, indicating that the neural activity in dACC is related to the confidence bias towards evidence supporting the chosen option. To capture this confidence bias, we used variants of drift diffusion models to fit with the participants’ behavioral data. The best-fitting model was the one equipped with both mutual inhibition and a collapsed decision boundary, and only this model could replicate the confidence bias, which is confirmed by simulation, implying that these two mechanisms are indispensable components of neurocomputational models accounting for the decision-making process.

In the second part, we investigated how human metacognition influences subsequent decision-making. Study 2 focuses on the impact of previous confidence on subsequent decisions and confidence in independent decision sequences. The results showed that in both perceptual and value-based decisions, previous confidence prompted individuals to make choices consistent with their previous choice and report higher confidence in the subsequent decisions. The best-fitting model was the one with modulation of the initial boundary by previous confidence, suggesting that boundary updating is most likely the mechanism by which metacognition regulates subsequent decisions. Previous confidence downweighed the initial decision boundary for the chosen options in the previous decision with a greater weight, leading to a repetition bias in individual choices of subsequent decisions. fMRI results showed that the primary motor cortex is involved in representing the initial decision boundary, indicating that the motor cortex also carries decision-related information and may be involved in determining when the decision should terminate. Study 3 focuses on whether the effect of previous confidence on subsequent choices and confidence exists across domains in independent decision sequences. The results of an alternating decision-making task showed that previous confidence only strengthens the repetition bias in subsequent identical decisions but leads to higher confidence in both subsequent identical and different decisions, suggesting that metacognition's regulation of subsequent decisions is domain-specific.

In the third part, we investigated the interaction between metacognition and decision-making in non-human primates (macaques). Study 4 found that in perceptual decisions, macaque reaction times conform to the behavioral characteristics of decision confidence and can reflect the correctness of their choices. fMRI results showed that the prefrontal cortex and intraparietal cortex may be involved in metacognition of macaques. In the decision sequence, the effect of metacognition on subsequent decisions was consistent with that of humans, and previous confidence lead macaques to make more choices in the subsequent decision that are same to the previous decision.

In summary, we provided a comprehensive investigation of the interaction between decision-making and metacognition in humans and non-human primates. The findings highlight biases in metacognitive monitoring of decision-making and demonstrate the influence of metacognition on subsequent decisions. Decision-making mechanisms drive confidence bias toward choice-consistent evidence, which is a general characteristic of the formation of confidence; confidence downweighs the initial boundary of chosen options of previous decisions in subsequent decisions with greater weight, making it easier for individuals to repeat previous choices. This effect is consistent in both humans and non-human primates. Overall, this research provides valuable insights into the cognitive and neural mechanisms underlying decision-making and metacognition, shedding light on the role of metacognition regulation in decision-making.