Scaffold-based visual analytics tools for complex activities

根据加拿大国家研究理事会的说法，为了支持加拿大的数字经济，研究和开发的一个高度优先领域是软件解决方案，这些解决方案可以处理数据的爆炸性增长以及对“使用计算机进行决策，综合信息和发现新知识的革命性方法”的不断升级的需求。作为来自不同领域的专业人士（例如，健康、法律、金融、保险、能源和教育）越来越多地使用大型和异构数据集来进行研究、创新、解决问题、制定决策和制定政策，因此需要先进的软件工具来支持这些领域中的日常复杂活动。我的研究涉及设计，实施和评估以人为本的可视化分析工具（VAT），这些工具支持执行复杂的活动，例如研究疾病的原因，解决科学问题和分析健康数据。增值税通过结合自动化机器学习模型、可视化和用户交互来促进这些复杂活动的性能。增值税利用人类的视觉，推理和决策能力以及计算机强大的数据发现和分析能力。复杂的活动不能完全由增值税执行;需要人类的判断和不断的参与来引导活动过程。这意味着必须通过高度的人力-VAT协作来执行任务。人类必须（a）确定输入并解释所有中间任务的输出，必须（B）决定如何交互地引导机器学习模型和可视化，并且必须（c）选择将哪些任务委托给工具以及哪些任务自己执行。将人类认知和计算结合起来并支持复杂活动的VAT的设计带来了一系列挑战;这种设计必须考虑一系列相互关联的问题：数据集和任务的协调和分布，机器学习模型，可视化，用户交互和人类认知。我的研究项目旨在开发技术，方法，原则和框架，用于设计以人为本的可视化分析工具，以支持用户执行复杂的活动。我将建立和研究工具，使人类和他们的数据之间的密切合作，用于分析目的。我将使用基于脚手架的技术来支持人类如何处理复杂的数据和机器学习模型。随着人类活动变得越来越复杂，这类研究是必不可少的。设计良好的增值税将使所有类型的用户都能进行涉及大数据的知识工作，通过显著提高工作效率来帮助简化他们的日常任务。这对于增加加拿大在不断增长的信息和通信技术市场中的全球份额至关重要。

According to the National Research Council of Canada, in support of Canada's digital economy, a high-priority area of research and development is software solutions that deal with the explosive growth of data and escalating needs for “revolutionary ways to use computers to make decisions, synthesize information, and discover new knowledge.” As professionals from a diverse set of fields (e.g., health, law, finance, insurance, energy, and education) increasingly use large and heterogeneous datasets to do research, innovate, solve problem, make decisions, and set policies, there is need for sophisticated software tools that support the daily complex activities in these fields. My research involves designing, implementing, and evaluating human-centred visual analytics tools (VATs) that support the execution of complex activities, such as studying causes of diseases, solving scientific problems, and analyzing health data. VATs facilitate the performance of these complex activities by combining automated machine learning models, visualizations, and user interactions. VATs take advantage of the visual, reasoning, and decision-making abilities of humans and the powerful data discovery and analysis strengths of computers. Complex activities cannot be performed entirely by VATs; human judgment and constant involvement are required to steer the activity process. This means that tasks must be carried out through a high degree of human-VAT collaboration. The human must (a) determine the input and interpret the output of all intermediate tasks, must (b) decide how to interactively steer machine learning models and visualizations, and must (c) choose which tasks to delegate to the tool and which to perform themselves. The design of VATs that couple human cognition and computation and support complex activities poses a series of challenges; such design must consider a host of interrelated issues: coordination and distribution of datasets and tasks, machine learning models, visualizations, user interactions, and human cognition. My research program aims at developing techniques, methods, principles, and frameworks for the design of human-centred visual analytics tools that support users in the execution of complex activities. I will build and study tools that enable a close collaboration' between humans and their data, for analytic purposes. I will use scaffold-based techniques to support how humans work with complex data and machine learning models. As human activities become more and more complex, this type of research is indispensable. Well-designed VATs will enable all types of users to carry out knowledge work involving large data, helping to streamline their daily tasks by providing significant improvement in work efficiency. This is important for increasing Canada's global share of the growing information and communications technology market.