Designing Tools to Support Cognitive Decision Making Under Uncertainty

高风险行业（如航空、海军、核电、医疗保健）容易发生安全故障，因为操作人员与其决策支持工具之间的交互非常复杂。不确定性带来了额外的复杂性，因为当不确定性出现时，作业者必须与决策支持工具合作，以检测和适当地管理不确定性。了解如何在不确定的情况下支持作业者的认知决策（CDM），对于设计决策支持工具以在正确的时间产生正确的行为至关重要。目前的决策支持工具并不能完全支持作业者的决策。操作员继续经历认知超载，误解显示和不信任信号，导致不准确的决策。自动化对于定义良好的环境是有效的，在这种环境中，围绕最佳实践的知识和证据是被理解的。然而，在现实中，作业者经常面临他们必须适应的不断变化的情况，并且由于设计自动化方法的预期环境发生了变化，因此自动化方法不再适用。操作人员必须根据不确定程度在自动化和自适应CDM之间切换，不确定程度可能源于实际情况（例如，失去时间意识）、环境因素（例如，失去电力）和操作人员的技能/知识。当正常的节奏被打乱时，不确定性就会骤然产生，其范围可以从平凡的事件到灾难性的事件。我们必须设计工具，支持作业者在管理安全风险时有效地采用自动化和自适应CDM。我们如何控制不必要的变化，识别、重视和支持必要的变化？这是一个复杂的、多维的问题，是一个非常重要的领域，但基本上被忽视了。该研究计划的长期目标是通过引入决策支持工具来优化CDM质量，这些工具可以利用预测分析来检测不确定性并向运营商传达信息。短期目标是：(1)建立不同不确定性水平下的清洁发展机制分类模型；(2)开发新的决策支持显示概念，并评估其在支持CDM方面的有用性；(3)评估预测分析对缓解CDM不确定性的影响。许多高风险系统（如飞行甲板、潜艇、火场、应急控制中心和手术室）都需要研究不确定性下的决策。在这个项目中，我们将使用手术室的测试用例，因为它在竞争需求、数据过载、不确定性、短时间尺度和快速变化的条件方面代表了高风险系统。因此，从这个项目中学到的知识将被转移到其他环境中。

High-risk industries (e.g., aviation, naval, nuclear power, healthcare) are vulnerable to safety failures because of the complexity imposed by the interaction between human operators and their decision-support tools. Uncertainty introduces additional complexity because operators must collaborate with the decision-support tools to detect and appropriately manage the uncertainty when it arises. Understanding how to support operators' Cognitive Decision Making (CDM) during uncertainty is critical to designing decision-support tools to engender the right behaviours at the right time. Current decision-support tools don't fully support operator decision-making. Operators continue to experience cognitive overload, misinterpret displays, and mistrust signals, resulting in inaccurate decisions. Automation can be effective for well-defined circumstances where knowledge and evidence around optimal practices are understood. In reality, however, operators are frequently confronted with evolving situations to which they must adapt, and in which automated approaches are no longer appropriate because the anticipated context for which they were designed has changed. Operators must switch between automation and adaptive CDM depending on the level of uncertainty, which can stem from the situation (e.g., loss of time awareness), environmental factors (e.g., loss of power), and operators' skill/knowledge. Uncertainty is precipitated when usual rhythm is disrupted, and can range from mundane to catastrophic events. We must design tools to support operators to effectively engage automation and adaptive CDM when managing safety risks. How do we control for unnecessary variation and recognize, value, and support necessary variation? This is a complex, multi-dimensional question and is an incredibly important area that has by and large been ignored. The long-term objective of this research program is to optimize CDM quality with the introduction of decision-support tools that leverage predictive analytics to detect and communicate uncertainty to operators. The short-term objectives are to (1) develop a classification model of CDM under various levels of uncertainty; (2) develop novel decision-support display concepts and evaluate them for their usefulness in supporting CDM; and (3) assess the impact of predictive analytics to mitigate uncertainty in CDM. The study of decision-making under uncertainty is needed in many high-risk systems (e.g., operators working in flight decks, submarines, firegrounds, emergency control centres, and operation rooms). In this program, we will work with the test case of operating rooms because it is representative of high-risk systems in terms of competing demands, data overload, uncertainty, short timescales, and rapidly changing conditions. As such, learnings from this program will be transferable to other contexts.