CHS: Small: Developing a Probabilistic Grammar of Graphics for Flexible Uncertainty Visualization

本项目旨在更系统地理解不确定信息的数据可视化，并使数据可视化设计者更容易构建这种可视化。在报道从选举到自然灾害等高风险和不确定的话题时，记者经常使用不确定性可视化来帮助公众回答重要问题，例如“谁将赢得下一届总统选举？”或“面对可能发生的洪水，我应该撤离吗？”例如，记者使用“不确定性锥体”来说明飓风路径预测，尽管有证据表明不确定性锥体可视化很难解释，而且存在更好的不确定性可视化。在实践中采用更有效的不确定性可视化滞后于研究，部分原因是构建更复杂和有效的不确定性可视化比构建常见但效率较低的不确定性可视化（如置信区间）更难，而开发原型以探索不确定性可视化的设计空间是昂贵的。该项目将创建各种不确定性可视化的正式描述，一个使用此描述使创建此类数据可视化更容易的工具包，以及评估工具包是否使设计人员更容易创建不确定数据的可视化。将数据可视化构造形式化为代数或语法，试图降低用户探索一系列正确有效的可视化所需的技术复杂性。虽然存在几个流行的例子，但这些语法都没有形式化不确定性或概率的概念，要求用户重新组合统计转换和几何（或者更糟的是，在可视化框架之外转换输入数据），以便创建有效的不确定性可视化，包括现代频率帧不确定性可视化，如假设的结果图。这项工作提出了一种图形的概率语法，它直接结合了概率分布，使探索不确定性可视化的设计空间变得更容易。这项工作将包括系统地对现有的不确定性可视化进行编目，开发一致的语法来描述它们，然后在现有框架中实现该语法，并评估其潜力，使有经验的可视化设计师和可视化设计专业的学生更容易构建高质量的不确定性可视化。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。

This project aims to build a more systematic understanding of data visualizations of uncertain information and to make it easier for data visualization designers to construct such visualizations. When reporting on high-stakes and uncertain topics ranging from elections to natural disasters, journalists routinely employ uncertainty visualizations in an attempt to help the public answer important questions, such as "Who will win the next presidential election?" or "Should I evacuate in the face of potential flooding?". For example, journalists use "cones of uncertainty" to illustrate hurricane path predictions, despite evidence that uncertainty cone visualizations are hard to interpret and better uncertainty visualizations exist. The adoption of more effective uncertainty visualizations in practice lags behind research, in part because the construction of more sophisticated and effective uncertainty visualizations is harder than the construction of common but less effective uncertainty visualizations like confidence intervals, while developing prototypes to explore the design space of uncertainty visualizations is costly. This project will create a formal description of a wide variety of uncertainty visualizations, a toolkit that uses this description to make the creation of such data visualizations easier, and evaluations of whether the toolkit makes it easier for designers to create visualizations of uncertain data.Formalizations of data visualization construction into algebras or grammars have sought to decrease the technical sophistication users need to explore a range of correct and effective visualizations. While several popular examples exist, none of these grammars formalize the notions of uncertainty or probability, requiring users to recombine statistical transformations and geometries (or worse, to transform input data outside of the visualization framework) in order to create effective uncertainty visualizations, including modern frequency-framing uncertainty visualizations like hypothetical outcome plots. This work proposes a probabilistic grammar of graphics that directly incorporates probability distributions to make it easier to explore the design space of uncertainty visualizations. The work will involve systematically cataloging existing uncertainty visualizations, developing a consistent grammar to describe them, then implementing that grammar in existing frameworks and evaluating its potential to make it easier for both experienced visualization designers and visualization design students to construct high quality uncertainty visualizations.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.