CAREER: Resolving Uncertainty Visualization Reasoning Errors with Mental Model Design and Training

世界各地的人们使用数据可视化来显示对未来事件的预测，以决定如何应对迫在眉睫的危险。例如，电视新闻气象学家经常使用飓风预测路径的可视化来通知公众即将到来的风暴。不幸的是，大量研究表明，人们曲解了这些预测中最常见的可视化不确定性的方法。这个项目的目标是更多地了解为什么人们在使用预测可视化方面有困难，以及如何创建更有效的预测可视化。该项目的一个关键成果将是不确定性可视化素养理论，它将确定有效利用可视化不确定性所需的技能，并为新的可视化设计提供认知方面的知情规则。这项工作将开发更有效的方法来传达预测的不确定性，以及支持公众在应对自然灾害和公共卫生危机时做出明智决策的不确定性素养培训。研究计划包括一系列实证研究，以检验相互竞争的假设，包括一种新的理论，其核心是将用户的心理模型整合到可视化设计和培训中。新的假设预测，不确定性可视化推理错误是由于人们如何概念化预测(先验图式)和可视化呈现数据(可视化驱动的图式)之间的差异造成的。这项工作的第一阶段将创建工具来揭示先验图式和可视化驱动的图式，使用先前建立的评估图式的认知方法。这些方法包括分析参与者的绘画、眼球跟踪和记忆力测试。在第二阶段，团队将开发一种测量工具来确定两种模式之间的相对距离，使用基于人为因素开发的方法的映射协议单元，然后使用该工具对基于模式的理论进行经验性测试，并与使用不确定性可视化时推理错误的替代解释进行比较。在第三阶段，团队将使用获胜的方法来开发飓风路径预测和新冠肺炎发病率预测的可视化培训和新的可视化设计作为测试平台。这一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智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

People worldwide use data visualizations that show forecasts of future events to decide how to respond to impending hazards. For example, television news meteorologists often use visualizations of a hurricane's forecasted path to inform the public about an approaching storm. Unfortunately, a large body of research demonstrates that people misinterpret the most common methods for visualizing uncertainty in forecasts such as these. This project’s goal is to learn more about why people have difficulty using forecast visualizations and how to create more effective ones. One key outcome of the project will be a theory of uncertainty visualization literacy that will identify the skills needed to effectively use visualized uncertainty and provide cognitively informed rules for new visualization designs. This work will develop more effective methods to convey forecast uncertainty, along with uncertainty literacy training that will support the public in making informed decisions in response to natural disasters and public health crises.The research plan includes a series of empirical studies to test competing hypotheses, including a novel theory centered around integrating users' mental models into visualization design and training. The new hypothesis predicts that uncertainty visualization reasoning errors result from discrepancies between how people conceptualize a forecast (a priori schemas) and how the visualization presents the data (visualization-driven schemas). The first phase of this work will create tools to reveal a priori schemas and visualization-driven schemas using previously established cognitive methods for evaluating schemas. These methods include analysis of participants' drawings, eye tracking, and memory tests. In the second phase, the team will develop a measurement tool to determine the relative distance between the two schemas, using a mapping agreement unit based on methods developed in human factors, then use that tool to empirically test the schema-based theory compared to alternative explanations for reasoning errors when using uncertainty visualizations. In the third phase, the team will use the winning approaches to develop visualization training and new visualization designs of hurricane path forecasts and COVID-19 morbidity projections as testbeds.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.