在数据智能迅速发展的背景下，在线社群中的知识协作与群体认知形成呈现出高度社会化与人机协同化特征，但其信息行为机制及技术支持方式仍缺乏系统认识。本项目围绕在线知识协作中的个体、团队与群体信息行为，系统探究认知形成机制及人工智能介入下的协作模式，旨在为人智协同知识建构提供理论基础与技术路径。项目采用跨阶段、多方法的研究设计。在理论层面，综合信息行为、认知科学与知识建构理论，系统梳理并建构协同知识建构的核心概念体系与阶段模型，明确知识共同体、社会互动与共享认知在群体认知形成中的关键作用。在实证层面，围绕健康信息、教育学习、科研协作与在线讨论等典型场景，综合运用眼动追踪、脑电与生理信号采集、用户实验、访谈分析及在线社群数据挖掘等方法，揭示个体选择性信息行为、负性偏向、态度转变及其认知机制，阐明团队成员态度差异、协作倾向与沟通模式对知识构建过程的影响。
项目构建了包含眼动、心电、皮电与脑电指标的多模态实验数据集，以及基于 Reddit CMV 社区的在线讨论数据集，为研究在线说服、知识协商与群体认知提供了关键数据支撑。在此基础上，系统分析了生成式人工智能在团队协作与众包场景中的角色定位、信任形成及潜在风险，深化了对人智协同认知过程的理解。在技术实现方面，项目设计并开发了 AI 参与的协同知识建构平台 Seminar.AI，引入对话式 AI 智能体支持小组讨论、信息整合与知识沉淀，并通过可用性测试与教学应用验证了其在促进信息共享与意义建构方面的有效性。本项目从多层级视角系统揭示了在线知识协作与群体认知形成机制，拓展了信息行为与人机协作研究的理论边界，为教育实践、平台治理及 AI 支持的协作系统设计提供了重要的科学依据与应用参考。
With the development of rapid digitalization and advances in artificial intelligence (AI), knowledge collaboration and group cognition formation in online communities have become increasingly socialized and human–AI collaborative in nature. However, the underlying mechanisms of information behavior and effective modes of technological support remain insufficiently understood. This project focuses on individual-, team-, and group-level information behaviors in online knowledge collaboration, systematically investigating cognitive formation mechanisms and patterns of collaboration involving artificial intelligence, with the aim of providing theoretical foundations and technical pathways for human–AI collaborative knowledge co-creation. The project adopts a multi-stage, mixed-method research design. At the theoretical level, it integrates information behavior theory, cognitive science, and knowledge construction theory to develop a coherent conceptual system and stage-based model of collaborative knowledge construction, clarifying the roles of knowledge communities, social interaction, and shared cognition in group cognition formation. At the empirical level, the project examines representative contexts—including health information, education and learning, research collaboration, and online discussions—using a combination of eye-tracking, electroencephalography and physiological signal collection, user experiments, interviews, and large-scale online community data analysis. The findings reveal mechanisms underlying selective information behavior, negativity bias, and attitude change at the individual level, and elucidate how team members’ attitudinal diversity, collaboration tendencies, and communication patterns shape the knowledge construction process.
The project has constructed key research datasets, including a multimodal experimental dataset comprising eye-tracking, electrocardiogram, electrodermal activity, and EEG indicators, as well as an online discussion dataset derived from the Reddit CMV community. These datasets provide critical empirical support for studying online persuasion, knowledge negotiation, and group cognition. Building on this foundation, the project further investigates the roles, trust formation processes, and potential risks of generative AI in team collaboration and crowdsourcing contexts, deepening understanding of human–AI collaborative cognition. On the technical side, the project designed and developed Seminar.AI, an AI-supported collaborative knowledge construction platform that incorporates conversational AI agents to facilitate group discussion, information integration, and knowledge consolidation. Usability testing and instructional deployment demonstrate the platform’s effectiveness in supporting information sharing and sensemaking. Overall, this project systematically advances understanding of online knowledge collaboration and group cognition formation from a multi-level human–AI collaboration perspective, extending the theoretical boundaries of information behavior and human–computer collaboration research, and providing scientific evidence and practical implications for educational practice, platform governance, and the design of AI-supported collaborative systems.