Prevalence effects in visual research: Theoretical and practical implications

在医学检查中，“偶然发现”是应该注意但对检查结果是偶然的发现。 检查的目的（例如，寻找肺炎和意外发现癌症）。遗漏的偶然事件 调查结果可能会产生重大的医疗和法律的后果。这些错误可以被认为是 在基础视觉中，“搜索满足”错误与“不注意视盲”密切相关 research.这种错误在日常生活中经常发生，例如，在搜索超市后， 带回家的是番茄而不是罗勒本提案的目的是了解 在社会重要的搜索任务中产生偶然发现错误的条件，并使用 理解，以产生干预措施，可以提高这些搜索的效率，并减少 失败率我们将开发一个基于搜索方面的偶然发现的实验室模型 行为研究在我们的实验室广泛在先前的赠款期间：首先，我们已经研究了“混合搜索” 观察者同时搜索多种类型的目标的任务，结合视觉搜索 记忆搜索其次，我们开创了人类“觅食”搜索的研究--搜索一个 一个目标类型的未知数量的实例（例如采摘浆果）。结合这两个 范式，我们已经研究了“混合觅食”，其中观察员寻找未知数量的 多个目标的实例（类似于某些医疗情况）。最后，我们广泛地 研究了“流行效应”;发现罕见的项目在搜索中不太成功，只是 因为他们的低流行率。我们已经在实验室中记录了这些流行效应， 行李检查和癌症检查。在这一先前工作的基础上，有三个具体目标： 1)我们将把联合收割机混合搜索、觅食和低流行率方法结合到一个模型系统中进行测试 假设一些临床搜索任务的结构更可能是偶然的， 调查结果将被遗漏。我们将把我们的模型系统与放射科医生的行为进行比较， 评估干预措施以减少这些假阴性错误。 2)我们将使用第二个模型系统来研究警戒/任务时间和中断的影响。 在这里，我们使用信号检测方法来区分从标准偏移的可检测性的变化。 3)我们的引导搜索（GS）模型一直是经典搜索性能的领先帐户。 搜索任务。第三个目标是有一个GS模型，可以预测的条件，将产生 偶然发现错误的比率较高。更重要的是，该模型将成为一个可以使用的工具 以确定可能最小化这些错误的可测试条件。 我们的实验将为视觉搜索的基本过程提供见解，并可能导致 可用于减少危险类型的医疗错误的干预措施。

In medical exams, “incidental findings” are findings that should be noted but that are incidental to the purpose of the exam (e.g. looking for pneumonia and unexpectedly finding cancer). Missed incidental findings can have significant medical and legal consequences. These errors can be considered as “satisfaction of search” errors and are closely related to “inattentional blindness” in basic vision research. Such errors occur regularly in daily life, as when, after a search of the supermarket, you arrive home with the tomatoes, but not the basil. The goal of this proposal is to understand the conditions that produce incidental finding errors in socially-important search tasks and to use that understanding to generate interventions that can improve the efficiency of those searches and reduce failure rates. We will develop a laboratory model of incidental findings built on aspects of search behavior studied extensively in our lab in the prior grant period: First, we have studied “hybrid search” tasks in which observers search for multiple types of target at the same time, combining visual search and memory search. Second, we pioneered the study of human “foraging” search - search for an unknown numbers of instances of one target type (e.g. picking berries). Combining those two paradigms, we have studied “hybrid foraging” in which observers search for an unknown number of instances of multiple targets (analogous to some medical situations). Finally, we have extensively studied `prevalence effects'; the finding that rare items are found less successfully in search, simply as a function of their low prevalence. We have documented these prevalence effects in the lab, in baggage screening and in cancer screening. Building on this prior work, there are three specific aims: 1) We will combine hybrid search, foraging, and low prevalence methods into a model system to test the hypothesis that the structure of some clinical search tasks makes it more likely that incidental findings will be missed. We will compare our model system to radiologists' behavior and we will assess interventions to reduce these false negative errors. 2) We will use a second model system to study the impact of vigilance/time-on-task and interruption. Here we use signal detection methods to distinguish changes in detectability from criterion shifts. 3) Our Guided Search (GS) model has been a leading account of search performance in classic search tasks. The third aim is to have a GS model that can predict conditions that will produce high rates of incidental finding errors. More importantly, the model will be a tool that can be used to identify testable conditions that might minimize those errors. Our experiments will provide insights into fundamental processes of visual search and can lead to interventions that can be used to reduce a dangerous type of medical error.