Evidential theoretic modeling and analyses of risk perception and information control in cognitive decision processes

本研究发展了一种证据理论来模拟和分析人类在不完全信息和证据的情况下对风险感知和推理的认知过程。给出了信息融合与更新的方法。基于稻垣的情境适应性自治（SAA）的概念，在人类和计算机之间动态地交换权威，一系列实验已经完成，以研究人类和自动化之间的权威交换控制的认知方面。三个微观世界模拟器已开发用于此目的：两个过程控制问题和一个去/不去起飞决策问题。与这些模拟器，SAA的自动化的可靠性，在控制过程中的故障的大小，人类的信任自动化，和人类的自信心，在他们的能力进行手动控制的影响。结果表明，SAA在时间临界条件下是有效的安全控制。SAA在安全控制方面表现出了上级性能，而在传统框架中，人类在任何情况下都被赋予了完全的决策和控制权。

This research project has developed an evidential theory to model and analyze human cognitive processes for risk perception and reasoning with incomplete information and evidence. Methods for fusing and updating information are given. Based on Inagaki's concept of situation-adaptive autonomy (SAA) where authority is traded between humans and the computer dynamically, 'a series of experiments has been made to investigate cognitive aspects of the trading control of authority between humans and automation. Three microworld simulators have been developed for this purpose : Two for a process control problem and one for a Go/No Go takeoff decision problem. With these simulators, the effect on human acceptance of SAA of reliability of automation, magnitude of fault in the controlled process, human trust in automation, and human self-confidence in their ability to perform manual control are investigated. It is shown that SAA is effective for safety- control under time-critical conditions. The SAA has exhibited superior performance in safety-control to the conventional framework in which humans are invested with full authority for decision and control at all times in every situation.