Interactions on the Move: Understanding Strategy Adaptation in Dynamic Multitask Environments

随着电脑从相对安全的桌面中解脱出来，人们越来越需要了解界面设计对人们如何在移动中与信息通信技术进行交互的影响。在人们与安全关键环境中的技术系统交互的情况下，例如在驾驶汽车时，这种需求最为迫切。在许多这样的多任务处理情况下，由于任务之间对有限注意力资源的竞争，人们通常一次只能积极地关注一项任务。与此同时，我们与技术系统的许多互动往往是由如何在该设备上执行常规程序任务的先验知识所塑造的。因此，在多大程度上，关于如何在任务之间交叉分配注意力的决定受到这种先前使用设备的经验的限制尚不清楚。如果人们不根据任务环境的要求调整他们的交互方式，这可能是潜在的危险。对于人们如何选择在任务之间穿插使用资源，有很多解释。一种可能性是任务交错被限制在任务执行中的自然断点上。例如，考虑一个司机拨打电话号码。在这种情况下，司机可能会选择只输入电话号码的区号部分（或者从交互式菜单中选择“地址本”选项），然后在完成次要任务的另一个小步骤之前，将注意力转移到监控前方道路上。通过这种方式，任务表征结构中的自然断点充当了从一个任务切换到另一个任务的线索。或者，司机可以简单地设定一个时间限制（或阈值），他们准备把目光从道路上移开，并在这段时间内尽可能多地完成次要任务。另一种可能性是，选择任务交错策略，以最佳方式权衡完成次要任务所需的时间和切换到主要驾驶任务所需的任何额外时间，以便在拨号时保持稳定的车道位置。这项研究计划列出了一系列计划中的实验，这些实验将用于调查人们在驾驶时如何在多个正在进行的任务之间分配资源。实验将在桌面驾驶模拟器中进行，使用专门的仪器设备进行次要任务交互。实验将通过各种计算计算来了解人们如何选择在任务之间安排资源，并将研究操纵次要车内任务的表征结构的后果，以及功能性任务环境对性能和策略适应的特征。与这些实验的运行相结合，将进行建模，以实现人类多任务调度的各种计算帐户，并为每个任务导出关键的定量性能预测。这项建模工作将旨在确定哪个帐户提供了人类行为的最佳特征，并在此过程中，将为未来的工作奠定基础，这些工作旨在开发设计工具，以快速预测支持多任务用户移动的设计替代方案的效率。这一研究项目将有助于更好地理解复杂多任务环境中的人类行为，所获得的知识将对移动交互系统的设计者具有潜在价值。这些经验数据将让我们了解如何重新设计车载设备的界面，以安全有效的方式支持用户的需求。这些结论对于理解人们在监视安全关键系统的同时必须在多个并发任务之间分配注意力的各种情况下的行为具有价值。

With computers having been untethered from the relative safety of the desktop there comes a growing need to understand the implications of interface design for how people interact with information communication technologies on the move. Nowhere is this need greater than in situations where people interact with technology systems in safety critical environments, such as when driving a car. In many such multitasking situations, people can often only actively attend to a single task at a time because of competition for limited attentional resources between tasks. At the same time many of our interactions with technology systems tend to be shaped by prior knowledge of how to perform routine procedural tasks on that device. It is therefore not clear to what extent decisions about how to interleave attention between tasks is constrained by this prior experience of using a device. If people do not adjust their interaction style to the demands of the task environment this could be potentially dangerous. There are a number of accounts for how people might choose to interleave resources between tasks. One possibility is that task interleaving is constrained to natural break points in the execution of a task. For example, consider a driver dialling a telephone number. In this situation, the driver might choose to enter only the area-code part of the telephone number (or indeed select the 'Address Book' option from an interactive menu), and then return attention to monitoring the road ahead before completing another small step of the secondary task. In this way, natural break points in the representational structure of the task act as a cue to switch from one task to another. Alternatively, drivers might simply set a limit (or threshold) on the amount of time they are prepared to look away from the road and complete as much of the secondary task as possible within this window of opportunity. A further possibility is that task interleaving strategies are selected that optimally trade the time required to complete the secondary task against any additional time taken to switch to the primary driving task in order to maintain a stable lane position while dialling.This research proposal sets out a series of planned experiments that will be conducted to investigate how people allocate resources between multiple ongoing tasks while driving. Experiments will be conducted in a desktop driving simulator using specially instrumented devices for secondary task interactions. The experiments will be informed by various computational accounts of how people might choose to schedule resources between tasks, and will investigate the consequences of manipulating the representational structure of secondary in-car tasks and features of the functional task environment on performance and strategy adaptation. In tandem with the running of these experiments, modelling will be conducted that will implement these various computational accounts of human multitask scheduling, deriving key quantitative performance predictions for each. This modelling work will be aimed at determining which account provides the best characterisation of human behaviour, and in doing so, will set the foundation for future work directed towards developing design tools for rapidly predicting the efficiency of design alternatives for supporting the multitasking user on the move.This programme of research will lead to greater understanding of human behaviour in complex multitasking environments and the knowledge gained will be of potential value to the designers of mobile interactive systems. The empirical data will give insights into how interfaces for in-car devices might be redesigned to support users' needs in a safe and efficient manner. These conclusions will be of value for understanding behaviour in a variety of contexts where people must allocate attention between multiple concurrent task while monitoring safety critical systems.