多层复杂网络视角下开源软件社区集体智慧涌现机制研究

The collective intelligence of open source software (OSS) communities exhibits the powerful ability for solving complex problems. Such collective intelligence emerges from the co-evolutionary process of communication network, collaborative network and knowledge network formed by numerous community members in the absence of central regulation and control. However, most existing studies, which have been conducted from the perspective of a single network, have problems such as the coexistence of communication and collaboration relationships in the single network and the over-dependence on relationship characteristics for knowledge expression. The formation mechanism of collective intelligence in OSS communities hence needs further explored from a more comprehensive point of view. Therefore, by the viewpoint of multi-layer complex networks, in this project we mainly adopt agent-based modeling and exploratory data analytics to study the emergence mechanisms of collective intelligence in OSS communities. Especially, our study focus is placed on the co-evolutionary dynamics of member-communication network and community knowledge-network, the co-evolutionary dynamics of developer-collaboration network and community knowledge-network, and the co-evolution of developer-collaboration network and member-communication network. Based on the above studies, we further summarize and refine the emergence-theory of collective intelligence in OSS communities. Furthermore, in order to promote the formation of collective intelligence, community platform architecture and operating mechanisms have been explored. In sum, this project tries to deepen our understanding on the emergence mechanisms of collective intelligence in OSS communities, and lay a theoretical foundation for the management practice.

开源软件社区的集体智慧展现出强大的复杂问题求解能力。这种集体智慧涌现于众多社区成员自组织形成的交流网络、协作网络和社区层面知识网络相互关联的演化过程。现有研究主要从单一网络层面展开分析，并且存在网络中交流与协作关系并存、知识传播表述过分依赖个体关系特征等问题，不能有效反映开源软件社区集体智慧的形成机理。对此，本项目从多层复杂网络视角出发，采用多主体建模与探索式数据分析相结合的方法，通过对社区成员交流网络与知识网络共同演化动力学机制、开发者协作网络与知识网络共同演化动力学机制、开发者协作网络与社区成员交流网络演化一致性及人员轮换机制的深入研究，分析开源软件社区集体智慧的涌现机制。进而，在对开源软件社区集体智慧理论总结的基础上，以促进其形成为目的，进一步研究了社区平台架构及运行机制。本项目力图对开源软件社区集体智慧的涌现机制形成深入理解，为社区的平台建设与管理奠定理论基础。

开源软件本质上是由开源社区成员的集体智慧产物，其产生与发展与社区成员的自发分工协作密不可分。针对开源社区复杂自适应系统的特点，本项目从多层复杂网络视角，综合运用数据解析、多主体建模、深度学习等方法，研究了开源社区集体智慧的涌现机理。研究结果表明，宏观层面社区成员分工的明确度与地位差异度，及协作强度与协作参与度均对社区集体智慧的形成产生正向影响，同时，社区成员交互网络与软件功能具有明显的关联性，且展现出了完全自由和严格分工间权衡的协作模式。微观层面，高保有率的核心成员，以及外围成员对于流失核心成员的自发填补机制保证了开源社区交互网络在演化过程中稳定呈现经济的模块化小世界状态，在极大程度上促进了知识的流动与整合。此外，社区成员协作活动具有深刻的社会化动因（即同质相吸与差异偏好），在两种机制的共同作用下，社区协作网络由松散结构逐渐演化为有利于集体智慧涌现的多模块互连的网络状态。