CAREER: Decision-induced Biases in Visual Percepts

认知科学早就证明，人类的决策往往会受到不受欢迎的偏见的影响。一个根本性的问题涉及这种偏见的根本基础，以及它们最有可能出现在什么类型的情况下。现实世界中的情况通常需要人类根据相同的视觉信息进行一系列的决策(例如，判断一个水果是苹果还是橙子，然后决定是吃那个水果还是吃它旁边的那个水果)。关于不同的知觉决定如何在这种视觉处理序列中相互作用，人们知之甚少，但最近几项研究的数据表明，基于不确定的感觉证据的知觉决定可以在很大程度上偏离一个人随后对这一证据的感知。在国家科学基金会的支持下，斯托克博士将开展和监督一项研究，使用计算建模和人类心理物理实验相结合的方法，以了解知觉决定如何以及为什么会影响随后的视觉感知。具体地说，研究人员的目标是测试一种假设，即大脑应用一种决策策略，确保在一系列知觉任务中解释感觉信息时的自我一致性。这项拟议的研究将是在更自然的条件下(决策不是独立做出的)理解知觉决策的重要一步。拟议的研究结果也有可能在将知觉和认知联系起来方面提供重大的理论进步，导致对人类决策策略的统一理解。该研究直接应用于在决策过程中强烈依赖人类专家对证据进行视觉分析的程序(例如法医科学、医学科学)。这项研究的一个关键特点是它专注于大脑功能的计算建模。斯托克博士的目标是在心理学和行为神经科学领域推广严格的量化方法。为此，为该项目开发的建模技术将直接纳入调查员的研究生教学。此外，调查员将为心理学和神经科学的研究生和博士后研究员组织年度建模讲习班，并将维护与他的建模方法有关的公开学习工具的在线储存库。总而言之，这些努力将有助于促进计算建模，并将其纳入主流神经科学和心理学课程。

Cognitive science has long established that human decision-making is often flawed by undesirable biases. A fundamental question concerns the underlying basis of such biases and in what types of situations they are most likely to appear. Real-world situations often require humans to perform sequences of decisions based on the same visual information (e.g., deciding whether a fruit is an apple or an orange and then deciding whether to eat that fruit or the one next to it). Very little is known about how different perceptual decisions interact in such visual processing sequences, yet data from a few recent studies suggest that a perceptual decision based on uncertain sensory evidence can substantially bias a person's subsequent percept of this evidence. With support from the National Science Foundation, Dr. Stocker will conduct and oversee research that uses a combined approach of computational modeling and human psychophysical experiments in order to understand how and why perceptual decisions affect subsequent visual percepts. Specifically, the investigator aims to test the hypothesis that the brain applies a decision strategy that ensures self-consistency in the interpretation of sensory information across a sequence of perceptual tasks. The proposed research will constitute a major step forward in understanding perceptual decision making under more natural conditions (in which decisions are not made independently). The results of the proposed research also have the potential to provide a major theoretical advance in linking perception and cognition, leading to a unifying understanding of human decision making strategies.The research has direct applications for procedures that strongly rely on human experts to perform visual analyses of evidence in their decision-making (e.g. forensic sciences, medical sciences). A key feature of the research is its focus on the computational modeling of brain functions. Dr. Stocker's goal is to promote a rigid quantitative approach to the fields of psychology and behavioral neuroscience. Toward this end, the modeling techniques developed for this project will be directly incorporated in the investigator's graduate teaching. Furthermore, the investigator will organize a yearly modeling workshop for graduate students and postdoctoral fellows in psychology and neuroscience, and will also maintain an online repository of publicly-available learning tools relating to his modeling methods. Together, these efforts will help promote and integrate computational modeling into the mainstream neuroscience and psychology curricula.