CAREER: Prioritizing the Development of Team Cognition in Human-AI Teams to Engender the Advancement and Acceptance of AI Teammates

团队合作是现代社会的长期基础，用于完成许多重要和有意义的社会贡献。随着人工智能（AI）的不断进步，它将被用于与人类合作者组成团队。有效的团队合作需要分享知识和信息，这样团队成员才能分享如何完成目标的理解。目前，人工智能代理优先完成简单的任务，几乎没有概念，什么是团队合作和团队认知，或者什么是一个好的队友需要。为了确保人类和人工智能能够安全地合作，需要进行研究，以了解人类对人工智能队友的需求。此外，还需要设计和开发人工智能团队成员，让他们有意识地、积极地促进人类与人工智能团队的认知。该项目全面探索了如何将AI设计、创建和实施到人类-AI团队中，以推进团队认知。该项目将产生能够与人有效合作的人工智能系统。通过提升团队认知而获得的关于人类与人工智能有效合作的知识，将提高人类对人工智能队友的接受程度，提高人类与人工智能合作的热情。本项目的技术目标分为三个相关的目标。首先，对现实世界的员工进行访谈，然后进行大规模的调查实验，以确定人类希望人工智能为团队认知做出贡献和受益的方式。第二，采用参与式设计，确保AI队友积极促进团队认知。在人工智能团队中。然后在混合方法实验（收集和分析定量和定性数据）中对这些设计进行测试、验证和改进。最后，通过一个额外的混合方法实验，将AI队友设计、人类-AI团队认知形成和人类对AI队友的接受联系起来，更好地理解人类对AI队友的接受。然后，通过与学术和工业机构的合作，扩展和应用这项工作。这项研究的结果将不断融入教育机会，以促进下一代人工智能从业者和研究人员以人为本的观点。本研究解决的三个目标为在人类-人工智能团队中创建团队认知，以及设计更有效地为团队过程和结果做出贡献的团队成员奠定了以人为本的基础。这个基础包括确定人类希望人工智能队友如何有利于团队认知，人工智能队友可以采取哪些行动来创造所述利益，以及实现的利益如何创造高性能的人类-人工智能团队和高度接受的人工智能队友。研究成果将为人类-人工智能团队、团队认知和以人为本的人工智能提供概念和设计原则。本项目由CISE-IIS和既定计划刺激竞争研究（EPSCoR）共同资助。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。

Teamwork is a long-standing foundation of modern society, used to accomplish many important and meaningful societal contributions. As artificial intelligence (AI) continues to progress, it will be used to form teams with human collaborators. Effective teamwork requires sharing knowledge and information so teammates share understanding of how to accomplish goals. Currently, AI agents prioritize simple task completion and have little to no conceptualization of what teamwork and team cognition is or what being a good teammate entails. To ensure that humans and AI are able to work together safely, research that seeks to understand what humans want and need from an AI teammate is needed. Furthermore, work is needed to design and develop AI teammates that intentionally and positively contribute to human-AI team cognition. This project provides a comprehensive exploration of how AI can be designed, created, and implemented into human-AI teams to advance team cognition. The project will result in AI systems that enable effective teaming with people. The knowledge gained about effective human-AI teaming through advancing team cognition will improve human acceptance of AI teammates and increase human enthusiasm for working with AI.The technical goals of this project are divided into three related aims. First, interviews with real-world workers, followed by a large-scale survey experiment, are used to identify ways that humans want AI to contribute to and benefit team cognition. Second, participatory design is used to ensure that AI teammates actively promote team cognition. during human-AI team. These designs are then tested, validated, and refined in a mixed methods experiment (collecting and analyzing quantitative and qualitative data). Finally, an additional mixed methods experiment links AI teammate design, human-AI team cognition formation, and human acceptance of AI teammates to better understand humans' acceptance of AI teammates. This work is then extended and applied through collaborations with academic and industry institutions. The results of this research will be continually integrated into educational opportunities to promote human-centered perspectives in next-generation AI practitioners and researchers. The three aims tackled by this research create a human-centered foundation for creating team cognition in human-AI teams, and for designing teammates that more effectively contribute to teaming processes and outcomes. This foundation includes the identification of how humans want AI teammates to benefit team cognition, the actions AI teammates can take to create said benefit, and how that realized benefit creates high-performance human-AI teams and highly accepted AI teammates. The outcomes will contribute concepts and design principles for human-AI teaming, team cognition, and human-centered AI.This project is jointly funded by CISE-IIS, and the Established Program to Stimulate Competitive Research (EPSCoR).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.