团队认知负荷实时监控自适应系统的认知增强研究

Maintaining high performance for teams which work in complex and dynamic environments, such as healthcare, nuclear power, and aviation, is challenging. Technological intervention with adaptive systems is one of the most promising approaches to support team performance in such environments. Recently, augmented cognition research showed that such adaptive systems can be effectively driven by human cognitive states, such as mental workload, which are assessed through real-time measurement and analysis of neurological and physiological signals. We propose that adaptive systems that aim to optimize team workload can support team performance. However, the few existing studies have three important limitations, including: workload was measured only through electroencephalograph; workload measurement was conducted only on the individual level; limited kinds of mitigation strategies were utilized and tested. Based on our prior research and experience, we propose a systematic approach to address those limitations. Specifically, this approach involves: (1) integrate signals from functional near-infrared spectroscopy, electrocardiography, and electrodermal activity measures to assess workload on the individual level; (2) utilize brain synchrony and physiological compliance indicators to assess workload on the team level; and (3) evaluate the performance of multiple mitigation strategies to determine their best use scenarios. This proposed project will lead to the Team wOrkload Monitoring and Adaptive Aids (TOMAA) framework. We anticipate that TOMAA can provide a viable and scientific path for supporting team performance in dynamic environments. We also anticipate that TOMAA can be further developed and integrated into future intelligent systems to support and optimize team performance in various domains.

如何支持和优化团队在复杂的、动态的环境中工作的绩效是许多国民经济重要领域——如医疗、核电、航空航天等——亟需解决的科学问题。认知增强的研究显示，基于认知状态实时监控的自适应系统，能优化个体的工作绩效。申请人根据自身的团队研究的经验，以及在人因学领域的理论和技术积累，提出针对团队认知负荷的认知增强技术是优化团队作业绩效的有效途径；但这方面的现有研究有以下的局限：团队认知负荷的测量指标单一、测量层面限于个体、控制手段单一。本项目拟整合使用近红外脑成像和多通道生理测量等多种指标、结合脑同步与生理同步等团队层面的分析技术，研发团队认知负荷的实时监测方法；并通过多种认知负荷控制手段的绩效研究，探索团队认知负荷实时监控自适应系统的性能和规律。本项目为解决智能系统如何实时地支持和优化团队工作的科学问题提供有效的科学途径，并为在未来能将相关的技术应用在团队作业智能环境的设计之中打下基础。

对团队的状态进行有效的测量是优化团队工作、从而提升团队作业绩效的基础。基于神经生理计算的人体状态测量技术是实现团队状态实时监测的可行途径。本研究主要采用近红外脑成像超扫描和机器学习的方法，开发了团队状态实时监测技术与团队干预原型系统，为解决智能系统如何实时地监测团队认知状态、支持和优化团队工作绩效的科学问题提供科学的途径。.本研究主要完成了三个方面的研究内容：第一，采用n-back范式构建个体认知负荷识别模型并探索模型在复杂任务场景下（包括视觉搜索任务和复杂人机合作决策任务）的可推广性；第二，通过对不同系统的多人协作典型任务的调研与分析构建了模拟工厂主控室监控协作任务平台，并采用近红外脑成像超扫描和机器学习的方法构建了在该任务场景下的团队状态识别模型；第三，基于团队状态实时识别，构建了团队作业基于神经生理计算的团队认知增强系统原型。.本研究的主要结果包括：第一，所构建的个体认知负荷识别模型可以有效地识别n-back任务中的个体负荷状态，但推广至视觉搜索任务和复杂决策任务时准确度不足；第二，以脑同步作为特征的机器学习模型可以有效地识别团队作业的协同性；第三，基于团队协同状态对团队进行自适应的干预（自纠正讨论）可以有效提升团队绩效。.本研究的主要科学意义是验证了神经生理参数与机器学习技术应用在团队层面的实时状态测量是可行且有效的，并为相关智能干预系统的开发提供了方法上的指引。