中国层级政府注意力网络的时空耦合与协同演变OACHSSCD
Spatiotemporal Coupling and Collaborative Evolution of Chinese Hierarchical Government Attention Networks
[目的/意义]探索国家顶层设计在"多领域、跨时空、跨层级"体系中的协同机制,是理解中国特色治理模式的关键.[方法/过程]针对传统文本分析难以捕捉政策议题复杂关联结构的局限性,本文引入网络拓扑与时空耦合视角,定量刻画中国各级政府注意力的时空共振效应与跨层级响应状态.首先,基于大语言模型提取历年各级政府文本关键词并构建政府注意力网络;其次,识别政府核心议程,并计算网络间的时空耦合度,揭示政府注意力在纵向层级落实过程中的结构对齐特征;最后,结合定量指标识别近二十年重大事件下的政策治理模式,并开展跨层级协同的影响机制分析.[结果/结论]研究发现,中央治理范式在2013-2014年发生清晰的结构性断裂;政府注意力对齐呈现层级差异,区域治理经历从"同质化"到"差异化协同"的演变;地理距离和官员变动是影响央地议程对齐度的最显著因素.本研究为理解中国特色的协同治理模式提供新的分析框架与经验证据.
[Purpose/Significance]Exploring how national top-level design is collaboratively implemented across a complex system spanning multiple domains,hierarchical levels,and spatiotemporal dimensions is critical for profoundly understanding the governance model with distinct Chinese characteristics.Traditional qualitative policy analyses and basic text mining approaches often struggle to capture the complex,dynamic,and associative structures of policy issues over long periods.To bridge this gap,this study introduces a novel analytical framework combining network topology and spatiotem-poral coupling.We aim to quantitatively delineate the spatiotemporal resonance effects and cross-hierarchical response states of the Chinese government's governance philosophies,thereby providing new empirical evidence and tools for evalu-ating policy synergy and structural evolution.[Method/Process]This study collected and analyzed 6,353 annual Govern-ment Work Reports from the central,provincial,and municipal levels spanning two decades(2005-2024).To overcome the limitations of traditional unsupervised keyword extraction,we employed a large language model(Qwen2.5-7b),fine-tuned via Low-Rank Adaptation(LoRA),to accurately extract policy-relevant keywords at the sentence level.Following semantic alignment,we constructed multidimensional,time-stamped government attention networks based on keyword co-occurrences.Subsequently,we analyzed static network topologies and community structures to identify core policy agen-das.We then calculated the spatiotemporal coupling degrees among multi-level networks,utilizing Node and edge cou-pling strength in the network,to measure structural similarities during policy transmission.Finally,we utilized Spearman correlation analysis,incorporating macro-level provincial indicators,to quantitatively assess the driving factors behind the alignment of central and local agendas.[Result/Conclusion]The study reveals several key findings.First,the central go-vernance paradigm exhibits a distinct structural shift from prioritizing"high-speed growth"to focusing on"high-quality development"around 2013-2014,while consistently maintaining economic construction and livelihood improvement as its core pillars.Second,the transmission of top-level design demonstrates clear hierarchical differentiation.Provincial go-vernments serve as a pivotal bridging hub,translating macroeconomic strategies into actionable regional plans.Conse-quently,regional governance has evolved from early-stage"homogenization"into a mature paradigm of"differentiated colla-boration".Third,geographical distance(spatial decay effect)and official turnover(political stability)emerge as the most significant factors influencing the alignment degree between central and provincial agendas,substantially outweighing economic or cultural variables.While this research offers a robust quantitative framework,current co-word networks are limited in capturing deep causal semantics.Future research should integrate advanced entity relations and link policy net-work structures with actual socioeconomic output data to further evaluate governance efficacy.
巴志超;竺乐祺;刘祖军;孟凯
南京大学数据管理创新研究中心,江苏 苏州 215163||南京大学信息管理学院,江苏 南京 210023南京大学数据管理创新研究中心,江苏 苏州 215163南京大学数据管理创新研究中心,江苏 苏州 215163||智慧足迹数据科技有限公司,北京 100033南京大学数据管理创新研究中心,江苏 苏州 215163
社会科学
政府工作报告顶层设计跨层级传递政府注意力网络协同治理
government work reporttop-level designcross-level transmissiongovernment attention networkcolla-borative governance
《现代情报》 2026 (7)
83-99,17
国家自然科学基金面上项目(项目编号:72374098)南京大学文科人工智能交叉研究计划(AI for HASS)课题(项目编号:2025300106).
评论