Design and Evaluation of Automated Decision Aids for Tactical Awareness

复杂系统中的自动化虽然通常是高度可靠的，但从来都不是完美的。这意味着人类操作员必须决定何时依赖自动化，何时拒绝其错误结论和不适当的建议。从直觉上看，当自动化更可靠时，操作员应该更多地依赖它——而当自动化不太可靠时，操作员应该减少对它的依赖。然而，以往的研究表明，随着自动化可靠性的变化，操作员对其依赖行为的调整不够。操作人员有时可能会怀疑并拒绝自动化，即使它提供了有用的信息、结论或建议。或者，他们有时可能过于信任，依赖于不适合手头情况的自动化。为了充分利用新的自动化技术——以及监督这些技术的人——我们需要更好地了解导致操作员依赖或拒绝自动化的因素。我们还需要开发方法来鼓励人工操作员适当地依赖自动化。我们的研究将在这两个领域创造新的知识。

Automation in complex systems, while often highly reliable, is never perfect. This means that human operators must decide when to rely on automation and when to reject its erroneous conclusions and inappropriate advice. Intuitively, operators should depend on automation more when it is more reliable - and less when it is less reliable. However, previous research shows that operators do not adjust their reliance behaviour enough as the automation reliability changes. Operators may sometimes be sceptical and reject automation even when it offers useful information, conclusions or recommendations. Alternatively, they may sometimes be too trusting and depend on automation that is not suitable for the situation at hand. To make the most of new automation technologies - and the humans who supervise them - we need to better understand the factors that lead operators to rely on or reject automation. We also need to develop methods to encourage human operators to rely on automation appropriately. Our research will create new knowledge in both of these areas. In earlier work, we developed a statistical method to determine how much operators should adjust their reliance behaviour as automation reliability changed. We will extend that method to consider factors beyond automation reliability, such as the operator's own ability to complete the task. This will allow us to understand the factors underlying reliance decisions in greater depth. We have also shown in earlier work that well-designed visual displays help operators better distinguish changes in automation reliability and rely on it more appropriately. We will pursue this by developing and testing new display concepts in increasingly realistic simulators. The research will emphasize combat identification technologies used by soldiers to identify and track friendly units on the battlefield. However, the results of this research can be applied to a wide variety of automated systems found in healthcare, transportation and energy systems. Supporting effective human interaction with automation will result in safer and more productive operation of all of these systems.