Sure or unsure: How is confidence in perceptual decisions generated?

人类观察者经常面临对外部物体做出反应的需要，尽管对这些物体的感知是不完整的或扭曲的。在这些情况下，有必要使用对对象的感知来确定目前存在几个可能的对象中的哪一个。一般说来，这些判断具有三个特征：第一，人类可以做出正确或错误的判断，即存在哪些对象。其次，可能需要不同的时间段才能做出决定。最后，人类可能会或多或少地对自己对物体做出了正确的决定感到有信心。然而，现有的决策数学理论不能同时对准确性、决策时间和信心提供令人满意的解释：我们的预备研究表明，对错误决策的信心随着刺激的物理质量而增加，这一观察结果与现有理论中的任何一个都不能调和。因此，本项目的目标是提供一个关于选择、信心和决策时间的数学理论。为此，我们计划将我们先前提出的一个理论--加权证据和可见性模型扩展到时间域。其主要思想是，通过积累关于对象身份的感官证据来做出决定，直到达到决定阈值。信心来自两个输入，包括关于物体身份的感官证据，以及关于刺激的物理性质的感官证据。本项目的主要目的是测试这一新的动态版本的加权证据和能见度模型，这就是为什么需要一个独立的参与者样本、扩展的数学模型以及一系列对照实验。作为该项目的第二步，我们的目标是调查这一理论对使用置信度评级来衡量观察者区分不同刺激的能力的研究领域的影响。

Human observers are frequently faced with the need to respond to external objects although perception of these objects is incomplete or distorted. In these cases, it is necessary to use the percept of the object to make a decision which of several possible objects is present at the moment. In general terms, these decisions are characterized by three properties: First, humans can make a correct or an incorrect decision which of the objects is present. Second, it may take varying periods of time until a decision is accomplished. Finally, humans may feel a greater or lesser degree of confidence about having made the correct decision about the object. However, the existing mathematical theories of decision-making are not able to provide a satisfactory explanation for accuracy, decision time, and confidence at the same time: Our preparatory studies suggested that confidence in incorrect decisions increases with the physical quality of stimulation, which is an observation that cannot be reconciled with any of the existing theories. At a consequence, the goal of the present project is to provide a mathematical theory of choice, confidence, and decision time. For this purpose, we plan to extend a theory we proposed earlier, the weighed evidence and visibility model, in the temporal domain. The main idea is that a decision is made by accumulating sensory evidence about the identity of the object until a decision threshold is reached. Confidence is generated from two inputs, including the sensory evidence about the identity of the object, as well as sensory evidence about the physical quality of stimulation. The main purpose of the present project is to test this new dynamical version of the weighed evidence and visibility model, which is why an independent sample of participants, extended mathematical modelling, as well as a series of control experiments are required. As a second step of the project, we aim to investigate the consequences of that theory for research fields that use confidence ratings to measure observers’ ability to differentiate between different stimuli.