Computational Cognitive Models in Adaptive Automation: Supporting Operator Learning and Skill Retention

自动化辅助设备在工作场所和日常生活中变得越来越普遍。例如，自动驾驶系统可以驾驶飞机；预计自动驾驶汽车将在10年内上市。自动化使用的一个深刻的安全问题是，人类操作员可能会变得不那么积极参与，并逐渐失去技能熟练程度。在造成228人死亡的法航447航班事故中，一个主要原因是由于大量使用自动驾驶仪，“在高空缺乏任何手动飞机操作训练”（AF 447 Final Report, 2012）。为了提高人类自动化系统的安全性和可靠性，我将开发自适应自动化，使用计算认知模型来预测操作员的技能保留率和心理工作量。由此产生的自适应自动化将能够以“智能”的方式调整自动化辅助的水平，为操作员提供只有在工作量不太高的情况下练习技能的机会。

Automation aids are becoming increasingly prevalent in the workplace and everyday life. For example, autopilot systems can fly planes; autonomous cars are expected to hit the market within 10 years. A profound safety concern of automation use is that human operators can become less actively involved and gradually lose skill proficiency. In the Air France Flight 447 accident that killed 228 people onboard, a major causing factor is “the absence of any training, at high altitude, in manual aeroplane handling”, due to extensive use of autopilot (AF 447 Final Report, 2012). To improve the safety and reliability of human-automation systems, I will develop adaptive automation that uses computational cognitive models to predict operators’ skill retention rates and mental workload. The resulting adaptive automation will be able to adjust the level of automated aid in a “smart” way, providing operators with opportunities to practice their skills only when their workloads are not too high.