The “policy subsystem” has long been a key concept in our understanding of how policies on a given topic are produced. However, we know much less about policymaking in nascent policy subsystems. This article draws on the theories of agenda-setting and venue shopping to argue that the similarity and convergence of policy subsystems’ agendas across different institutional venues and over time are features that distinguish more nascent policy subsystems from their more established, mature counterparts. In simple terms, policy venues’ agendas converge when policy actors begin to discuss the same issues and instruments instead of talking past one another. The article illustrates this argument using textual data on Germany’s emerging Artificial Intelligence (AI) policy: print media debates, parliamentary debates, and a government consultation from the period between November 2017 and November 2019. The insights from our analysis show that actors emphasize somewhat different policy issues and instruments related to AI in different venues. Nevertheless, the longitudinal analysis suggests that the debate does seem to converge across different venues, which indicates the formation of a subsystem-specific policy agenda regarding AI.

长期以来，“政策子系统”一直是我们理解特定主题的政策如何制定的关键概念。然而，我们对新生政策子系统的政策制定知之甚少。本文利用议程设置和场地选择理论来论证，不同机构场地和时间上政策子系统议程的相似性和趋同性是区分新生政策子系统和更成熟、成熟政策子系统的特征。简单来说，当政策参与者开始讨论相同的问题和工具而不是互相谈论时，政策场所的议程就会趋同。本文使用德国新兴人工智能 (AI) 政策的文本数据阐述了这一论点：2017 年 11 月至 2019 年 11 月期间的印刷媒体辩论、议会辩论和政府磋商。我们的分析结果表明，参与者强调的观点有所不同不同场所与人工智能相关的政策问题和工具。尽管如此，纵向分析表明，争论似乎确实在不同场所趋同，这表明针对人工智能的子系统特定政策议程的形成。