RII Track-1: Data Analytics that are Robust and Trusted (DART): From Smart Curation to Socially Aware Decision Making

DART研究计划将创建一个阿肯色州研究人员联盟，以协同、综合的方式关注数据分析研究的卓越。教育和劳动力发展计划的愿景是创建一个全州范围的数据科学和分析教育生态系统，在这里，学习者可以接受设计好的、一致的、有序的和模块化的数据科学教育，并在他们的学术道路上的适当位置提供工作或进一步教育的机会。这些努力，再加上密集的行业合作，将为提高阿肯色州的研究能力和竞争力提供所需的支柱支持。DART将开发：1)提高数据管理和标记的速度和效率的方法；2)保护隐私和识别公正内容的技术；3)利用机器学习的预测能力，同时增加预测背后过程的可解释性的方法；以及4)更具包容性的数据科学课程，并为学生以数据为中心的未来做好更好的准备。通过将来自不同但互补的研究领域的一大批有才华的科学家聚集在一个研究项目中，这些进展将成为可能。该项目将支持数学、统计学、数据科学和计算机科学的基础研究，这些研究将通过可视化、更好的数据挖掘、隐私和安全保护、机器学习等实现数据驱动的发现。该项目将为研究人员和学生建立一个开放的计算基础设施，并开发创新的教育途径，以培养下一代数据科学家。DART将包括为本科生设立的数据科学暑期学院，以及为中学教师提供的广泛课程支持。设计和开发数据科学和分析学位课程的一个关键机会将是利用DART研究领域和主题作为课程的现实范例，并将这些整合到课程中。DART将把具有不同但互补的研究兴趣、背景和技能的数据科学研究人员聚集在一起，以刺激创新。DART的科学目标有助于国家科学基金会(NSF)在数据科学的基础、算法和系统方面利用数据革命(HDR)的大想法，并进一步开发协调的全州范围的数据网络基础设施。该项目将研究更好地进行大数据分析的关键障碍，并开发改进的算法和方法，以提供：1)更自动地精选异类、非结构化和结构不良数据的手段；2)通过增强手动方法来更快、更稳健地进行模型训练；3)通过保护贡献者的隐私来更安全地保护数据；4)改进数据质量的衡量标准；5)新的无偏见的模型预测和决策支持系统；以及6)在复杂机器学习模型的预测能力和统计模型提供的可解释性之间取得更好的平衡。这些研究成果中的每一项都将创建一个更好的框架，以平衡新数据分析技术的风险和收益。随着该州将其投资与行业优势更好地结合起来，将会有更多的机会来改善阿肯色州的生活质量，并稳步提高教育程度和工资。DART将包括为本科生设立数据科学暑期学院、暑期实习和研究经验、增加数据科学教育机会、为全州中学教师提供综合支持，以及修改课程以纳入相关数据科学主题和顶峰项目。数据网络基础设施的发展将增加教育机构、研究机构和行业之间的信息共享。该奖项反映了NSF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。

The DART research program will create a consortium of Arkansas researchers with a synergistic, integrated focus on excellence in data analytics research. The vision of the education and workforce development program is to create a statewide Data Science and Analytics educational ecosystem, where learners receive a designed, consistent, sequenced, and modular education in data science with job or further educational opportunities available at appropriate points in their academic path. These efforts, combined with intensive industry collaboration, will provide the pillars of support needed to improve research capability and competitiveness in Arkansas. DART will develop: 1) the means to increase the speed and efficiency of data curation and labeling; 2) techniques to protect privacy and identify impartial content; 3) methods for harnessing the predictive power of machine learning while increasing the interpretability of the processes behind the predictions; and 4) data science curricula that are more inclusive and better prepare students for a data-centric future. These advances will be made possible by bringing together in one research project a large group of talented scientists from diverse, but complementary, research areas. The project will support basic research in math, statistics, data science, and computer science that will enable data-driven discovery through visualization, better data mining, privacy and security protections, machine learning and more. The project will build an open computational infrastructure for researchers and students and develop innovative educational pathways to train the next generation of data scientists. DART will include a data science summer institute for undergraduates and extensive curriculum support for middle-school teachers. A key opportunity in the design and development of the Data Science and Analytics degree program will be to leverage DART research areas and topics as real-life examples for the courses and to integrate these into the curriculum. DART will bring together data science researchers with diverse, but complementary, research interests, backgrounds, and skills to stimulate innovation. DART scientific objectives contribute to the National Science Foundation's (NSF) Harnessing the Data Revolution (HDR) Big Idea in foundations, algorithms, and systems in data science and further develop a coordinated state-wide data cyberinfrastructure. The project will study key barriers to better big data analytics and develop improved algorithms and methods to provide: 1) the means to more automatically curate heterogeneous, unstructured, and poorly-structured data; 2) faster and more robust model training by augmenting manual methods; 3) more secure data by protecting the privacy of contributors; 4) improvements in metrics of data quality; 5) novel unbiased model predictions and decision support systems; and 6) a better balance between the predictive power of complex machine learning models and the interpretability provided by statistical models. Each of these research outcomes will create a better framework for balancing the risks and benefits of new data analytics technologies. As the state better aligns its investments with industry strengths, more opportunities to improve the quality of life in Arkansas and to steadily increase educational attainment and wages will develop. DART will include a data science summer institute for undergraduates, summer internships and research experiences, increased data science educational opportunities, integrated support for middle school teachers across the state, and revamped curricula to include relevant data science topics and capstone projects. Developments in data cyberinfrastructure will increase sharing of information among educational institutions, research institutions, and industry.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.