ISSN 1671-3710
CN 11-4766/R
主办：中国科学院心理研究所
出版：科学出版社

• •    

引入AI“新队友”：人-AI协同自主性对人-AI协作的影响机理

樊子立, 程垦, 朱梦霆   

  1. 湖北工业大学经济与管理学院, 湖北 430068 中国
    浙江工业大学管理学院, 浙江 310023 中国
    上海交通大学安泰经济与管理学院, 上海 200030 中国
  • 收稿日期:2026-02-19 修回日期:2026-07-24 接受日期:2026-08-10
  • 基金资助:
    国家自然科学基金(72402062); 湖北省教育厅哲学社会科学项目(22Q054); 大学生创新训练计划项目(S202610500118)

When AI becomes a teammate: The impact of collaborative autonomy on human-AI team dynamics

FAN Zili, CHENG Ken, ZHU Mengting   

  1. , 430068, China
    , 310023, China
    , 200030, China
  • Received:2026-02-19 Revised:2026-07-24 Accepted:2026-08-10
  • Supported by:
    The National Natural Science Foundation of China(72402062); Philosophy and Social Sciences Project of Hubei Provincial Department of Education(22Q054); College Student Innovation Training Program Project(S202610500118)

PDF (PC)

83

摘要： 随着“人工智能+”的深入发展，越来越多的工作需要人与AI协同进行。然而，以往研究大多以工具视角探索AI技术系统对人-AI协作的影响，却忽略了随着技术的“自主性”日益精进，人与AI的关系也发生了转变（工具→队友）。面对关系模式的变化给团队带来的冲击，如何保障人-AI协作的有效性成为亟待解决的问题。本文跳出了以往孤立地从AI技术系统单方面地看待与人类互动，转而把人也纳入到分析系统中，聚焦于人-AI协同自主性的四大维度：适应性学习、不确定性、不透明性、开放性，基于角色理论，分别进行了3项研究。研究1指出不确定性和不透明性维度导致的角色冲突在人-AI协作过程中起到了一定的阻碍作用；研究2指出开放性维度会通过AI的角色领悟来让AI系统进一步理解和把握与人类协作过程中所须具备的技术特质和角色要求，从而增强AI的心智韧性，以便AI更好地理解人类意图并展开协作；研究3指出适应性学习维度会进一步通过角色学习增强自身的拟人化程度，让人类团队能更好地理解和接受自己，最终达到促进人-AI协作的目的。3项研究相互呼应，构建了一个“人-AI协作模型”，旨在探索人-AI协同自主性会如何保障人与AI的有效协作。本文为组织成功引入AI、促进人机共创、打造“超级个体”提供了理论依据和实践指南。

关键词: 人-AI协同自主性, 人-AI协作, 角色冲突, 角色领悟, 角色学习

Abstract: With the deepening development of AI+, more and more work requires collaboration between humans and AI. However, previous studies have mostly explored the impact of AI technology systems on human-AI collaboration from a tool perspective, neglecting the shift in the relationship between humans and AI (tools → teammates) as technological autonomy improves. Faced with the impact of changes in relationship patterns on the team, how to ensure the effectiveness of human-AI collaboration has become an urgent issue to address. This paper focuses on the four dimensions of technological autonomy: adaptive learning, uncertainty, opacity, and openness. From the perspective of the coupled teammates of humans and AI, based on role theory, three studies are conducted respectively. Study 1 points out that the uncertainty and opacity dimensions of human-AI collaborative autonomy lead to role conflicts that hinder human-AI collaboration; Study 2 points out that the openness dimension is conducive to promoting AI to actively adapt to human roles through role taking; Study 3 points out that the adaptive learning dimension is conducive to promoting AI anthropomorphism and enhancing their acceptance of AI. The three studies echo each other, constructing a "human-AI collaboration model" aimed at exploring how human-AI collaborative autonomy can ensure effective collaboration between humans and AI. This paper provides theoretical basis and practical guidance for organizations to successfully introduce AI, promote human-AI co-creativity, and create "super individuals".