AI Institute for Adult Learning and Online Education (ALOE)

基于数十年对人类学习的研究，我们知道，人类学习的大部分是由他人介导的（由教师指导），是一个社会/协作过程，并使用一系列认知策略。儿童和成人的学习存在于认知能力的范围内。将成人学习与儿童学习区分开来的两个特点是，首先，虽然K-12的大部分学习都是关于封闭的、定义明确的问题，有明确的答案，但成人学习——尤其是在工作场所的成人学习——经常解决开放式的、定义不清的问题，这些问题可能没有明确的答案，或者可能有多种答案。其次，虽然大多数K-12学习是通用的和教师指导的，成人学习（特别是在工作场所）是特定任务和自我导向的，因此教育资源（通常是在线的）的扩散支持成人学习。乔治亚研究联盟（GRA）将建立一个名为“NSF人工智能成人学习和在线教育研究所（ALOE）”的国家人工智能研究所，其目标是通过提高可用性、提高可负担性和提高成功潜力，使教育更加公平。通过使用在线教育资源进行混合学习，提高可用性；通过低成本的虚拟助教可以提高教师的负担能力，从而扩大教师的影响范围，同时通过虚拟助教提供的认知和社会支持可以提高成功的可能性。人工智能（AI）项目旨在通过开发变革性的人工智能驱动的在线成人学习模型，将高等教育和继续教育结合起来，从根本上改善人类的学习，从而为国家利益服务。提出了一个全面而有组织的计划，同时利用人工智能来改变在线成人学习并推动人工智能的基础研究。GRA是一家拥有30年历史的私营非营利性公司，与州政府、商界和大学系统合作，推动能够产生直接经济效益的科学技术的发展。ALOE AI研究所涉及一个大型跨学科研究团队，包括两个非营利组织（Georgia research Alliance, IMS Global），三家工业公司（Boeing， IBM, Wiley）和七家教育机构（亚利桑那州立大学，德雷塞尔大学，佐治亚理工学院，佐治亚州立大学，哈佛大学，佐治亚技术学院系统，北卡罗来纳大学格林斯博罗分校）。此外，跨国咨询公司埃森哲（Accenture）正在与NSF合作，为该研究所提供资金。总体而言，该项目的目标与NSF人工智能研究所的愿景是一致的，即通过利用多个组织推进基础研究，开展使用启发式研究，并培养下一代多元化人才。在基础研究方面，预计将在四个领域做出重大协同贡献：(i)基于认知的人工智能（基于成人学习认知理论的人工智能虚拟助手，如主动学习）；（ii）大规模的基于人工智能的个性化（收集数百万成人学习者的学习数据，开发用于分析数据的新型机器学习和自然语言处理技术）；（iii）人类-人工智能协作：开发交互式可视化新技术，使教师和学习者能够建立相互的心智理论；（iv）负责任的人工智能：发现为在线成人教育设计社会技术系统的原则，其中人工智能代理合乎道德地工作以造福人类。在使用启发研究方面，将开展以人类认知和学习理论为基础的负责任的人工智能基础研究。至少有两个不同的重点：(i)开发人工智能教学助理，增强在线成人学习中的认知、教师和社会存在感，以帮助提高其效率和效果；（ii）针对成人教育大规模在线学习个性化的学习分析。所采用的方法，学习工程，是一种迭代设计方法，将工程的严谨性引入教育学科。从以人为中心的人工智能技术设计开始，人类可以在学习过程中扮演学习者、教师或不同的利益相关者，这一过程将继续部署人工智能技术，并收集和分析有关学习者和学习的大规模数据。然后，这个过程继续评估学习行为和结果，然后改进以人为本的人工智能技术。提供了一份详细的计划，以评估通过混合方法对学习和教学的影响。将使用随机对照试验来评估人工智能技术的使用如何促进和影响学习。将进行准实验研究，比较在线课程与面对面课程的学习效果和效率。项目执行过程的评估计划应由经验丰富的评估人员监督，该评估人员将采用价值参与的教育方法，寻求捕获所有利益相关者的观点、利益和价值，包括那些在评估环境中经常未被充分代表的利益相关者。国家人工智能研究所计划是一个多机构的努力，旨在建立具有人工智能长期回报潜力的研究所规模的人工智能研究。除了推进基础研究和开展使用启动型研究外，该项目还支持培养下一代人工智能人才，加强多学科人工智能研究，利用多个组织，并为人工智能研发中的协作工作提供连接点。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。

Based on decades of study of human learning, it is known that much of human learning is mediated by others (guided by teachers), is a social/collaborative process, and uses a range of cognitive strategies. Childhood and adult learning lie on a spectrum of cognitive abilities. Two characteristics that distinguish adult learning from childhood learning are first, while much of K-12 learning pertains to closed, well-defined problems with clear answers, adult learning - especially adult learning in the workplace - often addresses open-ended, ill-defined problems that may have no clear answer or that may admit multiple answers. Second, while most K-12 learning is general-purpose and teacher-guided, adult learning (especially in the workplace) is task-specific and self-directed, hence the proliferation of educational resources (often online) in support of adult learning. Georgia Research Alliance (GRA) will establish a National Artificial Intelligence Institute titled “NSF AI Institute for Adult Learning and Online Education (ALOE)”, the goal of which is to make education more equitable through enhanced availability, greater affordability, and enhanced potential for success. Enhanced availability is to be achieved through the use of online educational resources for blended learning; greater affordability is to be accomplished through low-cost virtual teaching assistants that amplify teachers’ reach, while enhanced potential for success is to be achieved through cognitive and social support provided by virtual teaching assistants. The Artificial Intelligence (AI) project aims to serve the national interest through the development of transformative AI-driven models of online adult learning that blend higher and continuing education to radically improve human learning. A comprehensive and well-organized plan is proposed that uses AI simultaneously to transform online adult learning and to drive foundational research in AI. GRA is a 30-year-old private, nonprofit corporation that collaborates with state government, business community, and university system to advance science and technology that generates direct economic benefits. The ALOE AI Institute involves a large interdisciplinary research team that includes two non-profit organizations (Georgia Research Alliance, IMS Global), three industrial companies (Boeing, IBM, Wiley) and seven educational institutions (Arizona State University, Drexel University, Georgia Institute of Technology, Georgia State University, Harvard University, Technical College System of Georgia, University of North Carolina at Greensboro). Additionally, Accenture, the multinational consulting company, is partnering with NSF to provide funding for the Institute.Overall, the goals of the project are consistent with NSF AI Institutes’ vision to advance foundational research, conduct use-inspired research, and grow the next generation of diverse talent by leveraging multiple organizations. With regard to foundational research, major synergistic contributions are anticipated in four areas:(i) cognitively-grounded AI (AI virtual assistants that are grounded in cognitive theories of adult learning such as active learning); (ii) AI-based personalization at scale (collection of learning data from millions of adult learners and development of novel machine learning and natural language processing techniques for analyzing the data); (iii) human-AI Collaboration: development of novel techniques for interactive visualization that enables teachers and learners to build a mutual theory of mind; (iv) responsible AI: discovery of principles for designing sociotechnical systems for online adult education in which AI agents work ethically to benefit humans. With regard to use-inspired research, responsible fundamental AI research grounded in theories of human cognition and learning will be conducted. At least two distinct thrusts are in place: (i) development of AI teaching and learning assistants that enhance cognitive, teacher and social presence in online adult learning to help make it efficient and effective; (ii) learning analytics for personalization of large-scale online learning for adult education. The methodology employed, learning engineering, is an iterative design approach that brings the rigor of engineering to the discipline of education. Beginning with human-centered design of AI technologies, where the human could be a learner, a teacher, or a different stakeholder in the learning process, the process continues with the deployment of AI technologies and collection and analysis of large-scale data about learners and learning. The process then continues to the assessment of learning behaviors and outcomes followed by the refinement of human-centered AI technologies. A detailed plan is provided for assessment of impact on both learning and teaching through a mixed methods approach. Randomized controlled trials will be used to evaluate how the use of AI technologies facilitates and impacts learning. Quasi-experimental studies will be carried out to compare learning effectiveness and efficiency of online versus in-person classes. A plan for evaluation of the process of project execution is to be overseen by an experienced evaluator who will employ a values-engaged, educative approach which seeks to capture the viewpoints, interests, and values of all stakeholders, including those often underrepresented in the evaluation context. The National Artificial Intelligence Institutes Program is a multi-agency effort to establish institute-scale AI research with the potential for long-term payoffs in AI. In addition to advancing foundational research and conducting use-inspired research, the program supports efforts to grow the next generation of AI talent, enhance multidisciplinary AI research, leverage multiple organizations and provide a nexus point for collaborative efforts in AI research and development.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.