由于人工智能的重要性日益凸显，世界各国政府从科技政策层面推动人工智能的发展在全球已成为普遍现象和必然趋势。中国人工智能发展水平已进入国际前列，政府及其颁布的科技政策已成为国内人工智能发展最重要的动力。北京、上海、深圳、杭州人工智能综合竞争力位于全国前列，其产业发展和政策模式不仅具有很强的代表性和示范效应，而且具有较高的研究价值和研究意义。这四个城市都非常重视利用科技政策创造良好发展环境，同时运用各种政策工具推动和拉动人工智能的发展，成为全国乃至世界的人工智能产业重镇。鉴于人工智能对人类的正反两方面作用、人工智能与科技政策之间的双向互动关系、全球尚未形成公认的成熟人工智能政策模式以及国内人工智能发展的“极化”现象，为了确保人工智能政策的科学性、合理性、效率和效力，并使其发挥积极的作用，对人工智能发展先进地区的政策研究就变得至关重要和不可或缺。

本研究以1997年作为时间起点，收集截至2020年6月京、沪、深、杭四地政府公开颁布的287份人工智能相关政策文本，建立政策文本数据库，进行政策文本编码，通过文献计量法、社会网络分析法、内容分析法和比较研究法对政策文本进行全面分析。一方面利用Excel、NVivo、UCINET和NetDraw等软件，对政策数量、政策主题、政策效力级别、政策主体及其府际关系等政策文本特征进行全方位量化分析；另一方面构建政策目标——政策工具——创新价值链三维分析框架，采用定性与定量相结合的研究方法，对政策文本进行全面的内容分析，量化呈现和比较四地人工智能政策特征，洞察政策目标、政策工具和创新价值链三者之间的关系，挖掘政策文本背后蕴藏的深层信息和规律，对科技政策的质性研究进行有益补充。

经过分析研究，发现京、沪、深、杭四地人工智能政策存在政策主体间府际关系松散、政策力度波动显著、政策内容趋同性明显、政策工具使用结构性失衡等问题，针对存在的问题给出政策优化和改进建议，并以此为基础提出对其他城市乃至全国人工智能发展的借鉴意义。

As Artificial Intelligence (AI) has become increasingly important, it has become a common phenomenon and inevitable trend for governments around the world to promote the development of AI through science and technology (S&T) policies. China's AI development has entered the international forefront, and the government and its science and technology policies which it has promulgated have become the most important driving force for the development of AI. The comprehensive competitiveness of artificial intelligence in Beijing, Shanghai, Shenzhen and Hangzhou is among the best in China, and their industrial development and policy models are not only highly representative and exemplary, but also of high research value and significance. These four cities devote much attention to creating a favorable environment for the development of AI through science and technology policies, as well as using various policy instruments to promote and stimulate the development of AI, making them important cities for the AI industry in the country and the world. Given the positive and negative effects of AI on human beings, the two-way interaction between AI and science and technology policies, and the fact that the world has yet to develop a recognized and mature AI policy model, as well as the "polarization" of domestic AI development, research on policies in regions with advanced AI development becomes essential and indispensable in order to ensure the scientific validity, rationality, efficiency and effectiveness of AI policies and to make them play a positive role.

Taking 1997 as the starting point, this study collects 287 policy texts related to AI which were promulgated publicly by the governments of Beijing, Shanghai, Shenzhen and Hangzhou up to June 2020, builds a database of policy texts, codes the content of policy texts, and makes a comprehensive analysis of the policy texts using the bibliometrics, social network analysis, content analysis, and comparative research methods. On the one hand, this research makes an omni-directional quantitative analysis on characteristics of policy texts such as the number of policy texts, policy effectiveness level, themes of policy texts, policy subjects and their intergovernmental relationship by using multiple softwares such as Excel, NVivo, UCINET and NetDraw; on the other hand, the research builds a three-dimensional analytical framework of policy objectives-policy instruments-innovation value chains, adopts the combination of qualitative and quantitative research methods, makes a comprehensive content analysis of policy texts, and quantitatively presents and compares the characteristics of AI policies in the four cities, gain insight into the relationship among policy objectives, policy instruments and innovation value chains, and digs out the deep information and the regular patterns hidden behind policy texts, making it an useful addition to qualitative research on science and technology policies.

After the analysis and research, it is found that the AI policies in Beijing, Shanghai, Shenzhen and Hangzhou are characterized by loose inter-governmental relations, significant fluctuations in policy intensity, obvious convergence in policy content, and structural imbalance in the use of policy instruments, etc. Optimization and improvement suggestions about AI policies are given to address the existing problems, and on this basis, reference significance is proposed for the development of AI in other cities and even the whole country.