人类社会正在进入全面数字化时代。新技术的发展深刻影响着各方面，新兴技术在政治领域的颠覆性影响尤为深远。数据和算法结合，实现对人类行为进行的计算和预测，算法政治通过个性化、靶向式的思想和行为诱导，人为塑造“主流民意”，诱导大众陷入“监视资本主义”的“智能陷阱”。治理算法政治的风险迫在眉睫，研判技术对政治舆论的影响机制是数字时代公共治理的重要内容。计算宣传是最新的、普遍的、全球性的宣传范式，它是智能算法、政治机器人和社交媒体平台的集合体。较之以往的宣传方式，计算宣传通过推荐算法、趋势算法和过滤算法来影响政治和社会。借鉴注意力政治学理论，与社交媒体时代算法政治的情境结合，构建了“嵌入—分层—同步”的分析框架，刻画计算宣传对民意的影响。首先，建立原创数据库，爬取反映公共空间舆论变化的推特数据和用户的行为痕迹数据；其次，利用机器学习等多种方法对数据库的推文内容、人机对比、网络关系进行分析；再次，利用文本情感计算、LDA主题聚类分析和多智能体仿真等方法，对推荐算法、趋势算法和过滤算法影响民意的机制进行了实证探索。最后，针对计算宣传对舆论生态所产生的负面影响及治理面临前所未有的难题，从“集中治理”向“敏捷治理”模式发展，为规制算法的风险提供治理建议。科技巨头凭借场景丰富的算法能力、海量的数据和完备的基础设施，深度影响着世界的政治、经济和社会，美国是算法政治最重要的实践基地。同时，美国政府已经开始反思科技巨头和新技术带来的负面影响，开启了数字行业治理等新一轮的政策回应。因此，本文在实证中主要以美国为观察对象，以美国的政治运作现象为切入口，讨论新技术革命对政治的影响。研究美国，在一定程度上也是研究人类社会在数字化时代的当下与未来，其经验可供其他国家借鉴。研究贡献如下：一，拓展了注意力分配理论的应用场域，为计算宣传的进一步研究提供理论积累和视角补充。二，运用大数据方法分析算法政治情境，更贴合公共管理进入“数据事实”时代的发展需求，也是运用大数据方法分析公共管理领域问题的有益探索。三，积累了大量素材，为后续研究提供经验佐证和资料积累，丰富了该研究领域的语料库。四，有利于促进对数字社会民意塑造新方式的认识，深化了对新技术环境下国家治理的思考，探索了治理算法的路径。

Human society is an era of comprehensive digitization, and the increasing integration between new technology and politics is an important issue in the public sphere. On one hand, big data and artificial intelligence have dramatically enhanced our information collection and analytical capabilities, creating an unprecedented era of welfare creation based on algorithms; on the other hand, with the help of the super "computing" capabilities created by new technologies, people's "calculation" capabilities continue to rise. Algorithmic politics can artificially shape "mainstream public opinion" through personalized and targeted thinking and behavior induction, and induces the public to fall into the "intelligent trap" of "surveillance capitalism". "Algorithmic politics" brings major political security risks worldwide.Computational propaganda, defined as the use of algorithms, automation, and big data analysis to manipulate public opinion, has shaped a new, general and global propaganda paradigm. This paper uses the United States as the main case study object, and analyzes the influence of new technologies on political operation in the United States. I choose to focus on the US for three reasons: First, the United States has accumulated a strong technological innovation advantage in core fields such as artificial intelligence. Secondly, technology giants such as Twitter have profound impact on world politics, economy, and society due to their enhanced algorithmic capabilities, immense data accumulation, and outstanding infrastructure. As such, the US is the most important case study for analyzing the influence of technology on politics. Thirdly, the United States has already began to reflect on the negative impact of technology giants and new technologies on politics, and has initiated a new round of policy responses such as digital industry governance. Using the US as a case study, to a certain extent, can also be a study of the present and future of human society in the digital age.This paper takes advantage of big data tracking capabilities of social networks, combines frontier research in attention-based political theory with algorithmic politics in the social media era, and builds an interpretation framework of "embedding - layering - synchronization". As such, using big data analytics, I used Python to extract 115 days of data from social media platforms (more than 1.6 million pieces). My original database includes data from Twitter reflecting changes in public opinion in the public sphere, behavior trace data reflecting the participation of individuals in the Internet public sphere, and data reflecting sentiment expression in text using big data text sentiment classification, LDA theme cluster analysis and multi-agent simulation methods to conduct empirical exploration of the mechanisms behind computing propaganda influencing public opinion.The research contribution of this paper include: First, drawing on existing research on the politics of attention, I provided an interpretation of the influencing mechanisms of computing propaganda on public opinion through the process of "embedding-layering-synchronization", and accumulate materials for the corpus of research in the emerging field. Secondly, by using big data analytics, I was not only able to interpret the mechanisms involved in manipulating public opinion, but it also expanded the frontier of algorithmic politics research. It is also a new attempt in analyzing computing propaganda from a research method perspective, and an useful exploration in using big data analysis methods to resolve a public management issue. Thirdly, it helps to enhance our understanding of the new paradigm shaped by public opinion manipulation in the current digital society, provides a better understanding and analysis of the cases involving governance of digital industries in western countries, and provide reference for the governance of computing propaganda abuse.