Collective agency has been investigated in many empirical studies on collaborative learning. However, there is limited research investigating how participants develop collective agency via online collaborative learning, and the correlation between group interaction mode and collective agency. This mixed methods study investigates a class of 40 third-year college students' online learning underlying a multiple-agent interaction design where collective cognition and collective agency are prominent design features. Process and perception data were collected from multiple sources, including surveys, log records, content analysis of postings, and semi-structured interviews. The results of the study indicated the effectiveness of the multiple-agent interaction design, by analyzing students' collective agency and the collective knowledge construction via social network analysis and content analysis. The findings drawn from linear regression analysis further suggested that interactive network patterns within a group were generally not associated with students' perception of collective agency, but the network clustering coefficient was negatively associated with students' perception of collective agency. Moreover, inter-group interaction and collective knowledge construction process were generally not associated with students’ perception of collective agency. The study extends the literature on collective agency as it explores overt indicators, such as social network parameters, which may help to reflect the level of collective agency. Results also provide implications for online collaborative learning design and further research on collective agency.

许多关于协作学习的实证研究都对集体代理进行了调查。然而，关于参与者如何通过在线协作学习发展集体能动性以及群体互动模式与集体能动性之间的相关性的研究有限。这项混合方法研究调查了 40 名三年级大学生的在线学习，该学习基于多代理交互设计，其中集体认知和集体代理是突出的设计特征。从多个来源收集过程和感知数据，包括调查、日志记录、帖子内容分析和半结构化访谈。研究结果表明多智能体交互设计的有效性，通过分析学生的 通过社会网络分析和内容分析进行集体代理和集体知识构建。线性回归分析的结果进一步表明，群体内的互动网络模式通常与学生对集体代理的感知无关，但网络聚类系数与学生对集体代理的感知呈负相关。此外，群体间互动和集体知识构建过程通常与学生对集体代理的感知无关。该研究扩展了有关集体代理的文献，因为它探索了可能有助于反映集体代理水平的社会网络参数等公开指标。