智能传播时代，在看似真实的网络互动表面之下，隐藏着由算法生成的社交机器人， 它们能够生成和传播各类信息，参与对话表达观点，放大特定叙事或意识形态。目前已有大量研究关注到社交机器人行为及其对公众舆论的影响，但较少有研究系统探索社交机器人舆论操纵的机制到底是什么。社交机器人信息呈现的机制是什么?它们是如何表 达立场倾向并传递价值观?对公众议程的影响是如何随着时间变化的?本研究以俄乌 冲突、北京冬奥、新冠疫情三个议题为例，使用时间序列分析、结构主题模型分析、脉冲响应等方法深入探讨社交机器人在议程建构、叙事塑造、时间变化三个层面的作用机制。

研究发现，在俄乌冲突议程建构过程中，社交机器人表现出十分明显的选择性机制， 即通过放大对自己有利(对他人不利)的内容来改变舆论的平衡。而在国家和地区有关的讨论中，社交机器人放大了关于北约、美国和英国的讨论，将注意力转移到了其他国家。在北京冬奥的叙事塑造过程中，社交机器人在一些主题讨论中呈现出比较明显的泛 政治化倾向，这种倾向性可能在一定程度上影响了公众认知。在新冠疫情议题讨论中， 社交机器人和媒体都对公众议程产生正向影响，且随着时间的推移，媒体对公众议程的贡献度逐渐上升，而社交机器人的贡献度则呈现波动和整体下降趋势。此外，社交机器 人引起公众响应的最佳时间滞后为 1 小时，影响持续时间是 9 小时;媒体则需要更长的时间来设置公众议程，最佳时间滞后为 12 小时，影响持续时间也更长为 24 小时。

本研究通过深入探索社交机器人的舆论操纵机制，帮助我们理解智能传播时代社交机器人与公众舆论之间的动态关系，揭示了社交机器人在舆论形成过程中的选择性机制， 在公众头脑中塑造事件的认知框架，以及与公众议程之间随时间变化的动态过程。这对我们理解社交媒体中网络舆论的形成和变化具有至关重要的作用。

In the era of intelligent communication, beneath the surface of seemingly authentic online interactions, lie social bots generated by algorithms. These bots are capable of creating and disseminating various types of information, engaging in dialogues to express viewpoints, and amplifying specific narratives or ideologies. While a substantial body of research has focused on the behavior of social bots and their impact on public opinion, there has been limited systematic exploration of the mechanisms behind social bot manipulation of public discourse. What are the mechanisms behind the presentation of information by social bots? How do they express stance tendencies and convey values? How does their influence on public agendas evolve over time? This study, using the Russia-Ukraine conflict, the Beijing Winter Olympics, and the COVID-19 pandemic as case studies, employs methods such as time series analysis, structural topic modeling, and impulse response analysis to delve into the mechanisms of social bots in agenda building, narrative shaping, and temporal variation.

The research findings reveal that in the process of agenda building for the Russia-Ukraine conflict, social bots exhibited a markedly selective mechanism, namely, amplifying content favorable to themselves (and detrimental to others) to alter the balance of public opinion. In discussions related to countries and regions, social bots amplified the discourse related NATO, the U.S., and the U.K., redirecting attention to other nations. During the narrative shaping for the Beijing Winter Olympics, social bots demonstrated a distinct tendency towards politicization in some thematic discussions, which may have impacted public cognition to a certain extent. In the discourse surrounding the COVID-19 pandemic, both social bots and media exerted a positive influence on the public agenda. Over time, the media's contribution to the public agenda gradually increased, while the contribution of social bots showed fluctuations and an overall declining trend. Additionally, the optimal time lag for social bots to elicit public response was found to be 1 hour, with an influence duration of 9 hours. In contrast, the media required a longer period to set the public agenda, with an optimal time lag of 12 hours and a longer influence duration of 24 hours.

This study, by thoroughly investigating the mechanisms of opinion manipulation by social bots, aids in understanding the dynamic relationship between social bots and public opinion in the era of intelligent communication. It reveals the selective mechanisms employed by social bots in the formation of public opinion, their role in shaping cognitive frameworks of events in the minds of the public, and the evolving dynamics between these bots and public agendas over time. This understanding is crucial for comprehending the formation and transformation of online public opinion within social media contexts.