Eye Movements and the Dynamics of Adaptive Learning

项目总结/摘要 本研究探讨主观信念分布的建构与更新机制。人类 决策者必须经常根据新的观察结果更新他们的信念。信仰的程度 应该更新的是高度依赖于上下文：在稳定但嘈杂的环境中，新的观察应该 在不稳定的环境中需要更快速的更新。这个工作主要集中在 决定决策者（和他们的大脑）如何代表主观信念的精确度（或宽度） 分布，尚未明确证明。通过使用眼动追踪和神经成像 方法，我们可以在空间预测任务中获得连续的，隐含的和多方面的预测集 从个体的潜在信念分布中更密集地采样。这将使我们能够调查 是否以及如何代表信仰的整个分布，包括其中心趋势以及 他们的宽度。此外，很少有人知道感觉不确定性如何与中枢觉醒和更高的- 有序的认知过程来支持适应性学习。结合eye丰富的行为数据流 功能磁共振成像跟踪将为理解这些相互作用及其神经机制提供新的工具。 本研究的第一个目的是探讨不确定性和唤醒程度对自由观看的影响 内隐空间预测任务。眼动追踪将使我们能够监测个人预测能力的演变， 分布在一个试验一个试验的水平，并调查奖励和惩罚对学习率和 扫视动力学此外，瞳孔测量将使我们能够探索在一个人的信念更新的机制。 注视空间预测任务我们将评估奖励和惩罚是否能提高学习率 通过共同的激励机制。这将开始帮助我们解开是否变化， 唤醒水平是动态学习率适应机制的一部分。第二，功能性 磁共振成像将使我们能够检查不同类型的不确定性（噪声和 波动性），同时保持不确定性的数量不变。我们将首先调查是否大胆 额叶区域（即前额叶皮层、前扣带皮层和前额叶皮层）的反应是 差分依赖于由波动性而不是噪声主导的条件的预测误差。我们 接下来，我将测试这些因素如何改变额叶脑区域和视觉之间的功能连接。 皮质最终，这项研究将帮助我们了解信念分布是如何表示和更新的， 以及大脑如何在不断变化的环境中动态地适应这一过程。

PROJECT SUMMARY/ABSTRACT This project examines the mechanisms of construction and updating of subjective belief distributions. Human decision-makers must often update their beliefs in response to new observations. The degree to which beliefs should be updated is highly dependent on context: In stable but noisy environments, new observations should be given limited weight, while volatile environments require more rapid updating. This work focuses on determining how decision makers (and their brains) represent the precision (or width) of subjective belief distributions, which has not yet been explicitly demonstrated. By using eye tracking and neuroimaging approaches, we can obtain a continuous, implicit and multifaceted set of predictions in a spatial prediction task that samples more densely from individuals' underlying distributions of beliefs. This will allow us to investigate whether and how entire distributions of beliefs are represented, including both their central tendency as well as their width. Additionally, little is known about how sensory uncertainty interacts with central arousal and higher- order cognitive processes to support adaptive learning. Combining the rich behavioral data stream of eye tracking with fMRI will provide new tools for understanding these interactions and their neural mechanisms. The first aim of the project is to explore the influence of amount of uncertainty and arousal in a free-viewing implicit spatial prediction task. Eye tracking will allow us to monitor the evolution of individuals' predictive distributions on a trial-by-trial level, and investigate the effects of reward and punishment on learning rate and saccade dynamics. Additionally, pupillometry will allow us to probe the mechanisms of belief updating in a fixational spatial prediction task. We will assess whether reward and punishment enhance learning rate through common arousal/incentive mechanisms. This will begin to help us disentangle whether changes in arousal level are part of the mechanism of dynamic learning rate adaptation. For the second aim, functional magnetic resonance imaging will allow us to examine the effects of differing types of uncertainty (noise and volatility), while holding the amount of uncertainty constant. We will first investigate whether the BOLD response in frontal regions (i.e. anterior prefrontal cortex, anterior cingulate cortex, and anterior insula) is differentially dependent on prediction error for the condition dominated by volatility as opposed to noise. We will next test how these factors change the functional connectivity between frontal brain regions and visual cortices. Ultimately, this research will help us understand how belief distributions are represented and updated, and how the brain dynamically adapts this process in a changing environment.