事故致因理论对自主系统开发与安全运行的启示

在现代工业中,自主智能系统处在创新与探索阶段。由于先验知识的缺乏,其自身所带来的风险问题,越发得到学术界与工业界的关注。本文对于如何应用国际主流的事故致因理论以提高自主系统开发与运营的安全性进行了探讨。指出,现有的事故致因理论可用来从不同角度识别已知风险源,提高对未知风险源的警惕性,并通过描述事故场景,来辅助设计技术安全屏障与运行策略提高自主系统的安全性。为更好的理解自主系统如何会做出错误或不恰当决策,需对现有事故致因理论未涉及到的脑科学与认知心理学进一步的探讨与研究。

Implications from Accident Causation Theories to Development and Operation of Autonomous Systems

Today, autonomous intelligent system is in the stage of innovation and revolution. The risk issue of these systems has gained more attention from both academia and industry due to the lack of prior knowledge of the technology and operating experience. This paper discusses how to apply mainstream accident causation theories to improve the safety of autonomous operations. The paper concludes that the existing accident causation theory can be used to identify the known risk from different perspectives, improve the vigilance of the unknown risk, and develop accident scenarios to assist in designing technical safety barriers and operational strategies. To further understand how accidents could happen due to wrong or improper decisions made by the autonomous system, more study should be directed to neuroscience and cognitive psychology which are not covered by the existing accident causation theory yet.