【作者】 曾进群；

【导师】 杨建梅；

【作者基本信息】 华南理工大学 ， 管理科学与工程， 2013， 博士

【摘要】 随着网络技术的普及和发展及全球化进程的加快，一些不以层级和控制为基础的社区、协作和自发组织应运而生。有学者指出这种新的组织方式最终将替代传统的公司组织成为国家财富创造的主要动力。开源社区是这种新型组织的典型代表之一。Crowston（2012）指出开源社区研究还处于起步阶段，其一，一手数据难以获取，大部分学者采用的是二手数据，这样可能会造成结果偏差；其二，目前开源社区的研究主要是基于社会交换理论、软件开发理论等，存在一定的局限性。因此，开源社区亟需用新的理论与新的模型进行多视角深入研究。本文以开源社区为研究对象，着力于研究社区结构与行为及其特点，试图回答以下几个问题：开源社区内生产合作关系、沟通关系结构特征如何表征？随着时间的推移这些关系的演变情况又如何？不同参与人的行为特征怎样？开源社区内简单个体或业余个体组成的集体力量有时候比专业公司还大，体现了集体的智慧，开源社区内集体智慧该如何评价？从集体智慧视角，开源社区结构与行为又有哪些深层特征？集体智慧产生又有哪些制度特点？这些特点能给知识型企业管理带来哪些启示？依循上述问题，本文选择具有代表性的微软codeplex开源社区为对象，先利用数据挖掘技术，获取该社区内c#小社区中2006年6月-2012年6月共计6年的全部生产记录、沟通记录以及bug讨论记录共计近40万条；然后综合利用复杂网络及其零模型和人类行为动力学方法对开源社区结构与行为特点进行分析；再利用DEA方法和相关分析对开源社区集体智慧进行评价，并从集体智慧视角对开源社区深层特征进行剖析；最后，利用归纳方法对开源社区集体智慧发生的制度特点进行探讨，利用类比分析对知识型企业的管理提出建议。研究结论显示：第一，基于建立的项目-生产二分加权网络模型、生产者合作加权网络模型、项目生产二分网时间序列模型、沟通关系时间序列模型、bug讨论关系有向加权网络模型及时间序列模型，发现加权复杂网络模型更能体现生产合作关系、沟通关系结构及其演变特征；开源社区开发模式为洞穴式和网络式共存，随着时间的推移这种共存模式基本维持不变；参与者沟通结构为网络式，具有异配性、富人俱乐部特征，沟通版块不仅具有沟通功能，还具有“非直接通信”特征，随着时间的推移，论坛式沟通网络规模先增加而后趋于稳定，而问答式沟通网络规模逐步增大。第二，基于人力行为动力学方法，发现活跃度、时间间隔分布、阵发性和长程相关性能很好揭示参与者的时间行为特性；参与者持续时间呈广延或幂律分布，即大部分参与者仅浏览一下，而少部分参与者持续时间非常长；开源社区中人类行为不仅没有“标准类”，而且时间间隔分布存在多样性，有些呈指数分布，有些呈幂律分布，这些成果对开源社区管理和仿真研究有着重要意义。第三，通过DEA绩效评价、布鲁克斯法则及开源社区治理机制的探讨，发现开源社区体现出的是“群智”而不是“群愚”，项目参与人越多，绩效越大，bug数越少；从集体智慧视角来看，开源社区参与者角色、背景、结构、行为、观点都存在多样性，这些多样性是集体智慧发挥的源泉；开源社区各种角色拥有不同的权利，这些角色个体间可以根据能力和贡献量等互相转换，但角色间信息完全公开，这种分权制度和信息公开原则为决策独立性提供了制度支持；开源社区各项目以bug得票数进行集体决策，项目绩效与开发者人数弱相关，而与得票数强正相关。第四，开源社区规则、产权制度、社区自发式治理、项目内部治理为开源社区集体智慧发生提供了制度保障。基于开源社区与知识型企业特点的类似性，该研究还对知识型企业管理提出了建议。更多还原

【Abstract】 With the development and popularization of network technology and globalizationspeeding up, the community, collaboration and spontaneous organizations emerge as thetimes require, these organizations is not based on hierarchy and control. Some scholars havepointed out that this kind of organization eventually will replace the traditional organizationand become the main motive force of national wealth creation. The open source community isa typical representative of the new organization. Crowston (2012) pointed out that theresearch on the open source community is still in the initial stage, the first hand data, it isdifficult to obtain, most of the scholars is the use of secondary data, which may causedeviations; secondly, researchers need to draw on theoretical foundations that have beenutilized in prior research on social interaction, software development, as well as othertheoretical bases that are relevant to FLOSS phenomenon, to develop a more theoreticallygrounded understanding of FLOSS development from multiple angle with a new model andnew theory.In this paper, the open source community as the research object, focused on the researchcommunity structure and behavior and its characteristics, and try to answer the followingquestions: How to characterize the participants in the open source community partnershipsand communicative relationship structural features? The evolution of these relationships overtime and how is? What is the behavioral characteristics of the different players? Simpleindividual or collective power of amateur individuals sometimes in the open sourcecommunity than the professional company, reflects the collective intelligence. What depthfeatures from the collective intelligence perspective in open source community structure andbehavior? The collective intelligence how to evaluate? What institutional features of thecollective intelligence? These features give the knowledge-based enterprise management whatare the implications?Follow the above problems, Microsoft CodePlex open source community representativefor the object. First of all, the use of data mining technology, access to the community in asmall c#community in June2006-2012year in June a total of6years of all productionrecord, communication records as well as the bug discussion records a total of nearly400000,to get the first hand data. Then, the comprehensive utilization of complex network and its nullmodel and human behavior dynamics method to the open source community structure andbehavior characteristics for analysis; the DEA method and correlation analysis to evaluate the collective wisdom of the open source community; From the perspective of the collectivewisdom of the open source community structure and behavior depth features to analyze;Finally, Inductive method to explore the institutional features of the collective wisdom of theopen source community; Analogy analysis to make recommendations to the management ofknowledge-based enterprises.Conclusion of the study: Firstly, based on the establishment of the project-theproduction of binary weighted network model, producer cooperation weighting networkmodel, project production time series model, communication network time series model, bugdiscussed weighted network model and the time series model, weighted complex networkmodel to better reflect the cooperative relations of production, communication relationshipstructure and evolution characteristics; coexistence of the open source communitydevelopment model the cave and network type, this mode of coexistence essentiallyunchanged over time; The participant communication structure for the network type;communication area having a communication function, but also has the characteristic of "Stimergy", first increased and then stabilized over time, Forum formula communicate networkscale, however, a question-and-answer communication network size indeed graduallyincreased; Secondly, based on the dynamics of human behavior and found activity, timedistribution, burst and Long-Term Correlation reveals good time behavior of the participants;participants duration was extensive or power-law distribution, that is, most of the participantsonly browse, but a small number of participants very long duration; human behavior in theopen source community is not only no "standard class", and the time interval distribution ofdiversity, some of the exponential distribution, some power-law distribution, these results onthe open source community management and simulation research of great significance.Thirdly, the DEA performance evaluation, Brooks’s Law and the open source communitygovernance mechanisms discussed the open source community reflects the collective wisdom;project participants, the more people, the greater the performance, the less number of bugs;From the perspective of the collective intelligence, the role of a participant in the open sourcecommunity, background, structure, behavior, view there is diversity, diversity is a source ofcollective wisdom of the play; various roles of the open source community has differentrights,these roles among individuals according to the capabilities and contributions amount toconvert each other, but between the role of information in an open, decentralized system andinformation disclosure principles of the independence of the decision-making system; opensource community project based bug votes collective decision-making, project performanceand the number of developers weakly correlated, while the strong positive correlation with the number of votes. Fourthly, open source community rules, property rights system,Spontaneous governance, project internal governance system to protect the collective wisdomof the open source community. Based on similar characteristics of the open source communityand knowledge-based enterprises, the knowledge-based enterprise management is proposed.更多还原