【摘要】 开源创新作为人工智能生态中的关键组织模式,对技术发展与高质量创新具有战略意义。随着“深度求索”“千问”等开源大模型全球应用,中国开源组织模式备受关注。然而,学界对其在国家政策引导下的政策效果感知机制仍缺乏系统研究。本文基于全国性问卷调查,分析不同组织模式下开源创新行为与政策感知的关系。研究发现:在普遍性政策环境下,层级化程度较低的开源组织政策效果感知显著更强;该关系通过创新成本中介实现:行为激励成本与产出确权成本具有正向中介效应,表明适度治理投入与清晰成果归属可增强政策感知;规则协调成本则呈负向中介效应,反映制度摩擦会削弱感知;产出确权成本对资金配置感知影响尤为突出。结果揭示了不同类型创新成本在政策感知中的异质性作用,表明治理性与协调性成本影响方向相反。研究基于中国开源组织实践的经验证据,表明开源治理并非单纯社区自治,而是一种制度嵌入型的混合治理模式,从而为优化政策设计、组织治理及推动人工智能创新生态发展提供理论与实践参考。更多还原

【Abstract】 Open-source innovation, as a pivotal organizational model in the artificial intelligence(AI) ecosystem, holds strategic significance for technological advancement and high-quality innovation. With the global adoption of open-source large AI models such as Deep Seek and Qwen, China’s open-source organizational models have drawn increasing international attention. Yet, scholarly understanding of how these organizations perceive the effectiveness of national open-source innovation policies-and the underlying mechanisms-remains underexplored. Drawing on a nationwide survey of Chinese open-source organizations, this study examines the relationship between organizational modes of open-source innovation and policy effectiveness perceptions. Findings reveal that, under broadly implemented national policies,organizations with lower levels of hierarchical structure report significantly stronger policy effectiveness perceptions. This relationship is mediated by innovation-related costs: behavioral incentive costs and output appropriation costs exhibit positive mediating effects, indicating that moderate governance investment and clear attribution of outputs enhance policy perception; in contrast, regulatory coordination costs show a negative mediating effect, reflecting how institutional friction and coordination burdens diminish perceived policy efficacy. Notably, output appropriation costs exert a particularly pronounced influence on perceptions of funding allocation. These results uncover the heterogeneous roles of distinct innovation costs in shaping policy perceptions, highlighting opposing directional effects of governance-oriented versus coordination-oriented costs. Empirically grounded in China’s open-source practices, the study further demonstrates that open-source governance is not merely community self-governance but rather an institutionally embedded hybrid governance model. The findings offer valuable theoretical and practical insights for refining policy design, optimizing organizational governance, and fostering the development of AI innovation ecosystems.更多还原