系统透明度信息对人机信任和协同决策的影响

Impact of System Transparency Information on Human-Machine Trust and Collaborative Decision Making

中文关键词: 人机协同  自动化信任  透明化设计  信号检测论

英文关键词: human-machine collaboration  automation trust  transparency design  signal detection theory

中文摘要:

      目的 以信号检测论为切入点，探究将系统可靠性信息进行透明化设计后，其在不同系统可靠性和风险情境下对操作者的信任水平和对人机协同绩效的影响。方法 采用2(透明度)×2(风险)×3(可靠性)被试内设计，通过TNO信任任务收集了32位被试的数据，记录各条件下对自动化系统的信任水平、判断标准倾向和人机协同绩效，并对数据进行重复测量方差分析。结果 在不同可靠性水平和风险情境下，透明化设计产生的效果不同。相对于低透明化设计，当自动化系统可靠性较高时，高透明化设计可显著提升操作者对系统的主观信任评分、屈从率，更接近最优判断标准，带来更高的正确率增益、更小的人机最优绩效差异。风险水平影响人机协同绩效，高风险下的人机协同绩效显著低于低风险下人机协同绩效，并且当系统可靠性高时，操作者在高风险下对系统的屈从率、正确率增益较低风险下均显著下降。透明化设计可以改变不同风险下操作者的信任行为和绩效。高风险下，高透明度可显著提升操作者对系统的屈从率，减小与人机最优绩效的差异。结论 透明化设计有利于提高操作者的人机信任适配水平、优化决策判断标准、促进人机协同绩效的提升，尤其在系统高可靠性和高风险情境下，透明化设计的优势更加明显。

英文摘要:

      From the standpoint of signal detection theory, the work aims to explore the impact of transparency design that provides information on system reliability levels on operator trust levels and human-machine collaboration performance under different system reliability levels and risks. In experiment, a within-subjects design of 2 (transparency of the system:low, high) × 2 (risk:low, high) × 3 (reliability level of the system:95%, 70%, 45%) was used, and TNO trust task was adopted to collect the data of 32 subjects. The trust level, judgment standard tendency and human-machine collaboration performance of automation system under various conditions were recorded, and the data were analyzed by repeated measurement of variance. Under different reliability levels and risks, transparency design had different effects. Compared with low transparency design, when automation system reliability was high, the high transparency design could significantly improve the yield rate and subjective trust score of the operator to the system, which was closer to the optimal judgment standard, brought higher correctness gain and smaller optimal performance difference between human and machine. Risk level affected the performance of human-machine collaboration. The performance of human-machine collaboration under high risk was significantly lower than that under low risk, and when the system reliability was high, the operator's yield rate and correctness gain to the system under high risk were significantly reduced under low risk. Transparency design could change the trust behavior and performance of operators under different risks. Under high risk, high transparency could significantly improve the operator's yield rate to the system and reduce the difference with the optimal human-machine performance. Transparency design is beneficial to improving the adaptability of the operator to human-machine trust, optimizing the decision making criteria and promoting the performance of human-machine collaboration, especially under high reliability and high risk of the system.