I 摘要 在技术联盟中，每个成员被视为一个具有自主判断和行为能力的主体，会 根据他人的活动以及环境的变化随时调整自身的活动，这会导致联盟结构的不 稳定。我国技术联盟的发展还很薄弱，针对“技术联盟”的研究还处于起步阶 段；尤其是联盟所伴随的不稳定性，亟待理论与实践的指导。 技术联盟属于知识网络，已有的知识网络研究大都通过经济学、管理学理 论对区域网络、企业家网络等展开分析，关于知识网络结构及演化模型等方面 的研究只是零散的分布在物理、动力系统、复杂网络等领域的研究成果之中， 因此要探索借助跨学科理论系统地研究技术联盟的演化及其稳定性的影响因素； 其次，知识网络中的知识存量很难量化，很少有研究真正的去测量网络中知识 的变化；再次，虽然学者们已经对联盟的稳定性的影响因素及其作用机制进行 了探索，但是少有研究基于时间序列的纵向数据分析联盟成员的动态持续的相 互作用。本文借鉴马尔科夫链等理论，分析知识流对技术联盟稳定性的作用机 制；基于社会网络理论对技术联盟进行基于时间序列的纵向演化结构比较；同 时结合知识流的知识势差、一体化属性等相关因素对技术联盟的稳定性进行实 证研究。 本文界定了基于知识流的技术联盟稳定性的理论分析框架，论述了包括知 识流的知识势差、一体化属性、合作记忆以及组织来源等因素对技术联盟稳定 性的作用机制。具体来看，本文建立两阶段博弈模型，讨论不同知识势差的知 识流主体在技术联盟中的利益分配；在契约理论的分析框架下，判断不同一体 化属性下联盟成员的决策行为；借鉴 Boltzmann与Darwinian策略，分析成员 们的合作经历如何影响他们对于合作伙伴的搜索；依据马尔科夫链，研究如何 控制新入盟成员的组织类别来保持技术联盟的稳定。 在知识流各项因素的影响下，技术联盟的稳定性会发生演化。以AVS (Audio and Video Coding Standard) 技术联盟作为案例，本文对其稳定性的演化 进行分析，从技术持有人加入技术联盟的行为趋势着手，借助AVS的二模网 络和三个时期的网络数据，通过比较中心性、一致性、结构洞和连接性等指标 来评估AVS技术联盟稳定性的演化。中心性分析表明了技术联盟成员在每个 时期的技术交流水平；一致性分析来衡量联盟成员的贡献体现的实力差异；结 构洞分析揭示出网络的权力结构；连接性分析暴露出网络中关系的脆弱性。 在探析了AVS技术联盟稳定性的演化以后，进一步分析导致其演化的影 响因素：本文依据文献分析提出研究假设，通过实证方法对研究假设进行检验。 哈尔滨工业大学管理学博士学位论文 II 基于影响技术联盟稳定性的知识流的知识势差、一体化属性、合作记忆以及组 织来源等相关影响因素，本文构建联盟成员的关系矩阵，使用 Ucinet社会网 络分析工具对影响因素与成员们的合作走势之间的关系进行动态相关以及回归 分析，探讨各项因素对技术联盟稳定性的影响。研究发现：在技术联盟发展的 前期，成员之间的互补型、强弱结合式的合作以及成员之前较多的合作经历与 成员们积极地参与技术联盟，从而带来技术联盟的稳定运作有正向的相关性； 知识流的组织来源对成员们结盟行为的回归模型拟合度较好；关于知识流的知 识势差、一体化属性以及合作记忆等因素的回归模型也对技术联盟稳定性有显 著的解释。在技术联盟发展的后期，竞争型合作与技术联盟的稳定正相关，且 显著性好于第一阶段；势均力敌的合作有利于成员后续的结盟行为；知识流的 合作记忆与成员们的持续合作具有相关性，而且显著性比前期有所提高；关于 知识流的组织来源的回归模型的拟合度并不显著；知识流的知识势差、一体化 属性以及合作记忆等因素对技术联盟稳定性的解释程度好于第一阶段。针对 AVS技术联盟的稳定发展，本文提出了发展建议。 基于知识流，本研究进一步丰富了技术联盟稳定性的研究视角；借助社会 网络指标的变化来反映稳定性演变的结果，拓展了技术联盟稳定性的分析维度； 针对技术联盟的成员的伙伴选择展开分析，充实和完善了外部创新搜索的理论 认识。同时，本文的研究发现有助于发展稳定的技术联盟知识网络，进而提高 产业共性技术研发能力；有利于为联盟创造一个良好的内部信任环境，构筑技 术联盟的竞争优势；而各个成员组织的知识流相互促进、协调发展，可以实现 技术联盟内的知识进化过程，也会有力地促进联盟外部环境系统的知识进化过 程。 关键词：技术联盟；稳定性演化；社会网络；知识流；一致性分析；连接性分 析 Abstract III Abstract In a technology alliance, each member is regarded as a subject, which has the ability of independent judgment and behavior. Each member will modify his own behavior rules according to that of the others and environmental changes, which can lead to technology alliance instability. The development of Chinese technology alliances is still weak, and so is the research on "technology alliance". In particular, the attendant instability of technology alliances requires the guidance of theory and practice. A technology alliance is a knowledge network, and the existing research on knowledge network, mainly through economics or management theory with few references to other disciplines, analyzesd regional networks or entrepreneur networks, etc. The study on the structure and evolution model of knowledge network is only scattered in physics, dynamic system, complex network and other fields. Therefore, it is necessary to explore the factors influencing the evolution and stability of a technology alliance systematically by means of interdisciplinary theory. Next, it is difficult to quantify knowledge inventory in a knowledge network, so there were few studies that measured the change of knowledge-flow. Third, although scholars have explored the affecting factors and the relative mechanisms of action on alliance stability, few studies are based on longitudinal data of time series to analyze the dynamic continuous interaction of members. Based on Markova chain theory and others, this paper studies knowkedge-flow's action mechanisms on the stability of a technology alliance and make the longitudinal comparative analysis through time series data based on social network theory, explores factors like knowledge-flow's potential gap, integration attributes and others empirically. This study defines the theory framework and researches knowkedge-flow's mechanisms of action on technology alliance stability, including the analysis of knowledge-flow's potential gap, integration attributes, cooperation memory and organization sources. A two-stage game model is established to discuss the benefit distribution of members with different capabilities; from the perspective of contract theory, the decision behavior of members under different alignment attributes is judged; based on Boltzmann and Darwinian strategy, impact of the knowledge- flow's cooperation memory on searches of partners is analyzes; and according to Markova chain, the organization types of the members is studied to control new entrants and keep the stability of a technology alliance. Under the effect of knowledge-flow, the technology alliance stability will evolve. Then, this study uses the AVS technology alliance as a case, according to the 哈尔滨工业大学管理学博士学位论文 IV behavior trends of the organizations joining the technology alliance construct AVS's bivariate network, and with the data from the three periods evaluate the stability evolution by comparing centrality, consensus, structural hole and connectedness results. First, we use centrality analysis to compare the technology communication levels of the alliance's members during each period. Second, consensus analysis is used to reveal the magnitude of differences between the members’ contributions. Third, structural hole analysis can expose the power structure. Finally, connectedness analysis shows the vulnerability of the relationships within the network. After analyzing the evolution of the AVS technology alliance, the influencing factors of its evolution are further analyzed. Hypotheses are developed based on current research and an empirical study was carried out to testify the hypotheses. The members' relationship matrix in terms of knowledge-flow's potential gap, integration attributes, cooperation memory and organization sources were described according to the public information on the website. The relationship between influence factors and members' cooperation trends are analyzed with correlation and regression analysis by UCINET. In the former stage, the combination of the stronger and the weaker members, the mutual complementarity between members and the previous cooperation experience have positive correlation with the members' active participation, which leads to the stable operation of the technology alliance. From the influence of various factors, the regression model from