RUI: Causes and consequences of early quitting in visual search: Investigating the role of distractors

在现代生活中，分心无处不在，它们带来的后果从平凡到致命。众所周知，当观众第一次看到一个场景时，分散注意力的物体可以吸引他们的注意力。例如，路边闪烁的广告牌可能会让司机的视线暂时离开道路。但是，分心是否会以其他更微妙但同样重要的方式改变行为呢？视觉注意力的另一个关键方面是，当一个人搜索可能存在也可能不存在的东西时，必须决定他们是否找到了他们正在寻找的东西。一个重要的例子是，当放射科医生搜索医学图像时，扫描中可能存在也可能没有关注的区域。在这些搜索中，有一个战略决策组件，搜索者必须决定他们已经足够彻底地确定确实没有“目标”存在。然而，我们的知识有一个空白——很少有研究表明分散注意力的物体会如何影响这一决策成分。在当前的项目中，研究者探索了干扰诱发的退出现象，在这种现象中，分散注意力的物体改变了这一决策过程，导致人们比其他情况下更早地终止搜索。这种过早戒烟导致人们完全错过了他们本来可能找到的目标。从这个项目中获得的知识将促进我们对分心是如何被处理的理解，从而在注意力、分心和决策等领域对人类行为产生新的见解。此外，这些结果对涉及高风险目标搜索的任务具有潜在的现实意义，例如医学图像筛选或x射线行李检查。特别是，值得考虑的是，使用显著信号（例如，来自人工智能）向人类观察者传达信息可能会无意中引发这种确切的问题情况。例如，如果计算机系统被训练为扫描图像，并通过使用显著信号来突出放射学家（或安全人员）感兴趣的潜在区域，这些退出效应可能会抵消计算机引导系统可能提供的任何好处。最后，本科生参与这些实验的设计，数据的收集，并在包括医学影像学会议在内的会议上展示结果。其中一些学生来自康涅狄格学院的科学领袖项目，该项目致力于为历史上被排斥身份的学生提供科学机会。当人们寻找目标时，显著的信号可以改变搜索策略。更准确地说，在最近的研究中，研究人员发现，与任务无关的干扰因素会导致人们提前放弃搜索。因此，当这些干扰物存在时，人们更容易错过目标。在这个项目中，研究者使用各种实验协议来探索在目标可能存在或可能不存在的任务中显著干扰物对视觉搜索的影响。参与者在带有多个非目标的视觉显示中搜索简单目标，并按下一个键来指示目标是否存在。眼球追踪被用来研究导致参与者因视觉干扰而提前退出的具体机制——例如，当一个干扰物存在时，它是否会导致参与者不那么彻底地扫描显示器？或者它会导致参与者看每个项目的时间更短，使他们更有可能看一个目标，但不能正确处理它？接下来，研究人员建立了可以调节和潜在地消除这种干扰诱导的退出的因素，例如让参与者控制可能突出或不突出目标的显著线索的出现和消失。最后，研究者研究了显著信号的信息内容如何影响干扰诱导的退出——换句话说，如果显著信号有时会引起对目标的注意，那么这些信号在没有引起对目标的注意的情况下是否会引起同样多（或可能更多）的干扰？这些研究的结果可能有助于我们对人类视觉注意和视觉搜索的理解。研究结果与整个科学界分享，但也（更具体地说）与放射学的同事分享，以引发新的讨论，探讨研究人员如何将这项研究应用于帮助改善医疗环境中人工智能系统与人类观察者的沟通。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Distractions are everywhere in modern life, and they bring consequences ranging from the mundane to the deadly. It is well established that distracting objects can attract attention when a viewer first looks at a scene. For example, a flashing roadside billboard may take a driver’s eyes off the road for a moment. But might distractions change behavior in other, more subtle but equally critical ways? Another key aspect of visual attention arises when a person searches for something that may or may not be present and must decide whether or not they’ve found what they are looking for. An important illustration of this is that when a radiologist searches a medical image, there may or may not be an area of concern present in the scan. In these searches, there is a strategic decision component in which the searcher must decide that they’ve looked thoroughly enough to be confident that indeed no “target” is present. However, there is a gap in our knowledge – there has been little research on how distracting objects might affect this decision component. In the current project, the investigators explore the phenomenon of distractor-induced quitting, in which distracting objects alter this decision process and cause people to terminate search earlier than they otherwise would. This early quitting causes people to entirely miss targets that they would otherwise likely find. Knowledge gained from this project will advance our understanding of how distractions are processed, leading to new insights into human behavior in the fields of attention, distraction, and decision-making. Furthermore, these results have potential real-world implications for tasks that involve high-stakes searches for targets, such as medical image screening or x-ray baggage inspection. In particular, it is worth considering that the use of salient signals (e.g., from artificial intelligence) to convey information to a human observer may inadvertently trigger this exact problematic situation. For example, if a computer system is trained to scan images and highlight potential areas of interest for a radiologist (or security personnel) by using a salient signal, these quitting effects might offset any benefits the computer guidance system might otherwise afford. Finally, undergraduate students participate in the design of these experiments, collection of data, and the presentation of results at conferences including those focused on medical imaging. Some of these students are recruited from the Science Leaders program at Connecticut College, a program dedicated to providing opportunities in the sciences for students from historically excluded identities.Salient signals can alter search strategies when people are looking for targets. More precisely, in recent work, the investigators have discovered that task-irrelevant distractors can cause people to quit searching early. As a result, people more frequently miss targets when these distractors are present. In this project, the investigators use a variety of experimental protocols to explore the impact of salient distractors on visual search in tasks where targets may or may not be present. Participants search for simple targets in visual displays with multiple non-targets and press a key to indicate whether a target is present or not. Eye-tracking is employed to investigate the specific mechanisms that cause participants to quit early as a result of visual distraction – for example, when a distractor is present, does it cause participants to scan the display less exhaustively? Or does it cause participants to look at each item for a shorter time, making them more likely to look at a target but fail to process it correctly? Next, the investigators establish factors that can modulate and potentially eliminate this distractor-induced quitting, such as giving participants control over the appearance and disappearance of a salient cue that may or may not highlight the target. Finally, the investigators examine how the information content of salient signals can impact distractor-induced quitting – in other words, if salient signals sometimes draw attention to the target, do those signals cause as much (or perhaps more) disruption on the occasions where they do not draw attention to the target? Results from these studies may aid our understanding of human visual attention and visual search. Findings are shared with the scientific community at large, but also (more specifically) with colleagues in radiology in order to spark new discussion on how the investigators might apply this research to help improve communications from artificial intelligence systems to human observers in medical settings.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.