CHS: Small: Enhancing Data Analysis Strategies with Mixed-Initiative Visual Analytics

人们每天都在根据数据做出重要的决策，从访问城市时在哪家餐厅用餐等简单的选择，到医疗保健中关于采取哪种治疗方案的重要而复杂的决策，甚至影响国家安全和政策的决策。视觉分析系统在这些决策过程中发挥着关键作用。它们允许人们与他们的数据和分析模型进行交互，以查看数据的不同视角并获得见解。这种交互式数据分析过程由人们逐步引导分析模型生成数据的替代视图以支持其任务组成。在大多数情况下，这种人在回路中的过程会产生成功的、有见地的结果。然而，认知科学告诉我们，人们可以表现出先天的偏见行为。因此，他们的数据分析行为和策略可能会受到影响。最终，这可能会导致根据不完整的信息和如何解释数据的有限视角做出决策。有偏见的分析过程会导致有偏见的结果和错误的信息。该项目将进行基础研究，以发现如何检测此类潜在偏见并开发减轻偏见的视觉分析系统。它还将对来自STEM领域代表性不足的群体的研究生和本科生产生教育影响，部分是通过与少数群体教员举办外联讲习班，服务机构和历史悠久的黑人学院和大学，帮助他们将可视化分析和一般数据素养学习目标整合到课程中。拟议的多学科研究将开发技术，通过干预和提供必要的指导来增强混合主动视觉分析过程。为了实现这一目标，提出了三条主要的研究路线。首先，该团队将开发和评估计算指标，以从用户交互模式和系统参数中检测出不良和可能有偏见的分析策略。这些指标由概率计算模型组成，这些模型考虑了数据探索期间的数据覆盖率等指标。其次，团队将开发和研究不同的视觉分析系统设计，以指导和改进人们的分析过程。每个原型都将使用指标为人们提供指导，但通过不同的界面设计（例如，对话框、覆盖的视觉覆盖等）它们将通过出版物和开源代码进行开发、评估和提供。第三，所提出的研究将产生经验的结果和设计指南，为未来的混合倡议视觉分析systems.This奖项反映了NSF的法定使命，并已被认为是值得通过评估使用基金会的智力价值和更广泛的影响审查标准的支持。

People make important decisions based on data every day, from simple choices such as which restaurant to dine at when visiting a city, to important and complex decisions in healthcare about which course of treatment to pursue, and even decisions that impact national security and policy. Visual analytic systems play a critical role in these decision-making processes. They allow people to interact with their data and analytic models to view different perspectives of data and gain insights. This interactive data analysis process consists of people incrementally guiding analytic models to produce alternate views of the data in support of their tasks. In most cases, such human-in-the-loop processes have successful, insightful outcomes. However, the cognitive sciences tell us that people can exhibit innate biased behavior. As a result, their data analysis behaviors and strategies may suffer. Ultimately, this could lead to decisions made from incomplete information and limited perspectives on how the data can be interpreted. Biased analysis processes can lead to biased results and misinformation. This project will perform fundamental research to discover how to detect such potential bias and develop visual analytic systems that mitigate it. It will also produce educational impacts for graduate and undergraduate students from groups underrepresented in STEM fields, in part through outreach workshops with instructors from minority-serving institutions and historically black colleges and universities to help them integrate visual analytics and general data literacy learning objectives into course curricula. The proposed multi-disciplinary research will develop techniques that enhance mixed-initiative visual analytic analysis processes by intervening and providing guidance when necessary. To accomplish this goal, three primary lines of research are proposed. First, the team will develop and evaluate computational metrics to detect poor and potentially biased analysis strategies from user interaction patterns and system parameters. These metrics consist of probabilistic computational models that take into consideration metrics such as data coverage over the duration of the data exploration. Second, the team will develop and study different visual analytic system designs to guide and improve people's analysis processes. Each prototype will give people guidance using the metrics, but display information to users via different interface designs (e.g., dialog boxes, visual overlays of coverage, etc.) They will be developed, evaluated, and made available via publications and open-source code. Third, the studies proposed will generate empirical results and design guidelines for future mixed-initiative visual analytic systems.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.