俄乌冲突是21世纪以来全球瞩目的军事冲突和国际事件，成为了移动互联网时代全民参与的新型社交信息战争，其中不乏社交机器人的行为。它们通过散布相关虚假信息、收集行军情报、转发固定标签、培育重点机器人账号等，设置相关议程，干扰舆论风向。本研究聚焦了俄乌冲突议题下人机用户在第四层议程设置中的互动机制。该理论的核心是媒体不仅可以决定公众是否将碎片化的信息联系起来，构建议程网络；还决定了公众如何将多种语义元素相互联系，形成具有议程内容、提及频率、指向关系、情感效价、设置主体、时滞效果等多重属性的精细议程网络。首先抓取Twitter上有关俄乌冲突议题的推文数据，识别社交机器人账号。再综合运用内容分析、社会网络分析、二次指派程序、时间序列分析等方法，确立主要议程、构建精细议程网络、检验精细议程网络的相似性、设置主体、时滞以及效果程度，得出以下结论。在精细网络议程设置的特征方面，人机用户主要关注了军事行动、局势研判、谈判会议、表态声明、国际支援、国际制裁、战争危害、难民人权、呼吁和平以及政治人物等十大议程。各个国家和国际组织被塑造为不同形象的行动者：俄罗斯是强悍的侵入者，乌克兰是无助的反抗者，联合国是能力有限的调停者，美国是卷入争端的既得利益者，中国是亲俄立场的和平呼吁者，英国和法国是附和表态的跟随者。在精细网络议程设置的互动机制方面，人机用户的精细议程网络显著相关，社交机器人在第四层议程设置中的能力更强，能够在更短的时滞内、更加快速、更大程度地影响人类议程中对不同国家和国家组织之间关系的不同情感指向。由此，第四层议程设置实现了由“频率”到“关系”再到“行动”的超越。本研究的发现不仅表明社交机器人超越了工具性介质的范畴，还进一步审视了人机传播舆论生态中存在的主体异化、中介化道德盲视和文化滞后的问题。因此，需要降低道德的工具性，提高工具的社会性，培育用户的社交机器人素养；在技术平台嵌入“计算伦理”；减少信息圈中的熵增，秉持以人为本、人机共生的价值理念。
The Russia-Ukraine conflict is one of the most complicated global military issues of the twenty-first century and escalated in 2022. Social media has become a new battlefield where users set agendas, including social bots. They spread disinformation, collect intelligence, retweet and hijack hashtags, and grow active bot accounts. In the human-bot coexistence information ecosystem, the study explores the dynamic interaction of the human agenda and social bots agenda concerning Russia-Ukraine conflict issues on Twitter. The study is based on the fourth level of agenda-setting (a more elaborated network agenda-setting, ENAS), which advances the NAS model in this hypothesis that the news media not only determine whether the public or other interest groups associate these semantic elements, but also can tell them how to associate one semantic element to another. The ways include theme, frequency, centrality, direction, valence, subject，time lag, and effect. Overall, the core concept from first- to third- and then to fourth level changes from frequency to association and then to action.The study collected social data during Russia-Ukraine war from February 24 to April 18, 2022, consisting of nearly 200 thousand Twitter posts generated by more than 80 thousand users. Using botometer to detect bots and then employed content analysis, social network analysis, quadratic assignment procedure and time series analysis to categorize agenda, visualize elaborated agenda network, examine elaborated agenda network correlations and causality between human and bots, calculate time lag and effect of elaborated agenda network between human and bots. The findings suggest that users mainly focus on ten following agendas: military act, conflict situation, negotiation and conference, statement and claim, foreign aids, international sanctions, war damage, refugee rights, peace advocation, and political figures. Involved countries are portrayed with distinct images: Russia as a tough intruder, Ukraine as a helpless resistor, the UN as a mediator with limited power, the US as a manipulator, China as a peace advocator with a pro-Russian stance, and the UK and France as followers. QAP results show a strong correlation between the network of human agenda and the network of bots agenda. Social bots predict the fourth level agenda-setting better than human with shorter time lag, faster speed and to a greater extent. Specifically, social bots exert most influence on negative links from Ukraine to Russia of human agenda. The results also demonstrate that social bots are beyond tools. Problems such as subject alienation, moral blindness of mediatization, cultural lag need to develop countermeasures. The study reminds that we should reduce the instrumentality of morality and increase the sociality of tools. To curb social bots, we should cultivate social bot literacy, embed computational ethics into social media platform, reduce entropy increase of infosphere and promote human-oriented values.
