CyberTraining: Implementation: Small: Building Future Research Workforce in Trustworthy Artificial Intelligence (AI)

该项目的目标是通过开发教学材料，让学生接触到值得信赖的人工智能系统的各种挑战，培训当前和未来的研究人员在值得信赖的人工智能(AI)方面。该项目的重点是国家对训练有素、知识渊博的值得信赖的人工智能研究人员的迫切需求，他们能够解决复杂人工智能系统中的现实世界问题，并帮助安全地采用人工智能系统。该项目将通过培训研究人员来应对值得信赖的人工智能挑战，从而对公共和私营部门产生直接和长期的影响。佐治亚州立大学是一个为少数族裔服务的机构(MSI)，该项目将形成一个协调网络，由研究型大学、四年制学院、历史上的黑人学院和大学(HBCU)、为拉美裔服务的机构(HSI)以及亚特兰大大都会和更广泛地区的女子学院组成。合作将显著增加项目的集体影响，使许多来自代表性不足群体的学生受益，并有助于增加研究队伍的多样性。该项目团队将开发互动教学材料，包括一系列动手实验室，这些实验室采用最先进的值得信赖的人工智能技术来应对人工智能系统的各种挑战。将开发的教学材料包括对抗性机器学习、逃避攻击和防御、数据中毒攻击和防御、隐私攻击和防御、测试和验证以及公平、问责、透明度和道德(命运)等模块。该项目采用学习科学原则，特别是主动学习和基于探究的学习策略，以加深学生的理解，并向教师提供形成性反馈。这些教学材料基于真实世界的系统，旨在系统地涵盖值得信赖的人工智能和实践技能的基本原则。该项目还将提供指导方针，帮助教师将模块整合到他们的课程中。动手实验室将仅基于开源软件和工具构建，这些软件和工具可免费用于教育目的，并将通过免费云平台分发。项目评价包括形成性评价和总结性评价，采用定量和定性两种方法，将由一名有经验的外部评价人员在项目小组的帮助下进行。该项目将通过教育和研究社区的培训研讨会传播开发的材料。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 goal of the project is to train current and future research workforce members in trustworthy artificial intelligence (AI) by developing instructional materials that expose students to various challenges of trustworthy AI systems. The project is focused on the vital national need for well-trained and highly knowledgeable researchers in trustworthy AI who are capable of solving real world problems in complex AI systems and help enable secure and safe adoption of AI systems. The project will have direct and long-term impact in both the public and private sectors by training the research workforce to address trustworthy AI challenges. Georgia State University is a minority-serving institution (MSI), and the project will form a coordination network consisting of research universities, 4-year colleges, historically black colleges and universities (HBCUs), Hispanic-serving institutions (HSIs), and women’s colleges in Metro Atlanta and the broader region. The collaboration will significantly increase the collective impact of the project, benefit numerous students from underrepresented groups and help increase the diversity of the research workforce. The project team will develop interactive instructional materials including a set of hands-on labs that employ state-of-the-art trustworthy AI techniques to address the various challenges of AI systems. The instructional materials to be developed include modules on adversarial machine learning, evasion attacks and defenses, data poisoning attacks and defenses, privacy attacks and defenses, testing and verification, and fairness, accountability, transparency, and ethics (FATE). The project employs learning science principles, specifically the active learning and inquiry-based learning strategies that result in deeper understanding by students and provide formative feedback to instructors. The instructional materials are based on real-world systems and are designed to systematically cover fundamental principles in trustworthy AI and practical skills. The project also will provide guidelines to help instructors integrate the modules into their curriculum. The hands-on labs will be built based on only open-source software and tools that are free to use for educational purposes and will be distributed via free cloud platforms. The project evaluation includes formative and summative evaluations that use both quantitative and qualitative approaches and will be conducted by an experienced external evaluator with help from the project team. The project will disseminate the developed materials through training workshops for the educational and research communities.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.