The history of AI and higher education has been intertwined ever since the first AI pioneers started challenging the field of education, e.g by proposing “teaching machines” (Skinner, 1958) in the 1950s (Doroudi, 2022). The advanced “datafication of education” (Jarke & Breiter, 2019) and the continuing quantification of education is closely related to new forms of education provision, administration and governance (Williamson & Eynon, 2020). Today, these connections seem more relevant than ever when observing the advance of communicative AI technologies. Especially in the Anglo-American system these changes are influenced by the major role which is played by commercial corporations (Kerssens & van Dijck, 2021; Selwyn et al., 2023; Yu & Couldry, 2020). Communicative AI (ComAI)  are not only subject to technological advancements but are also shaped by societal expectations (Gulson et al., 2022). The prevalence of AI shapes societal discourse as overarching questions concerning trustworthiness, privacy and control emerge. Higher Education Institutions (HEIs) worldwide are put in a position where they must cope with these developments. They require new forms of conduct which are being manifested in documents, such as policy documents, ethical regulations and data protection documents. 

ComAI has been relevant in HEIs for a long time: both conversational bots, which are applied in settings of individual student learning and the automation of teaching processes, and work bots, which are used to support teachers and administrators, have been part of HEIs for many years. These technologies are integrated into what we call hybrid figurations - actor constellations made up of actors whose practices are interwoven with a media ensemble, and where the boundaries of each figuration are defined by its frames of relevance (Hepp et al., 2018; Hepp et al., 2023). We use the term ComAI as a sensitizing concept which can be applied to many different instances of AI and is therefore applicable to policy documents regarding AI even if they do not use this term specifically. 

When dealing with the challenges of implementing ComAI into their processes of teaching, learning and administration, HEIs turn to the creation of policy documents, ethical regulations and data protection documents. These are intended to guide the different HEI stakeholders in their usage of ComAI regarding educational justice, equality and ethical aspects just as much as data protection and privacy (e.g. University of California Presidential Working Group on AI, 2021). In our study we aim to examine these documents which have been published by German HEIs regarding three main aspects: How are the policies constructed and what is their underlying structure? What theoretical foundations and reference points form the basis of these policies? What are the primary focus areas and priorities addressed with these policies?  
Furthermore, we also analyze the timeline on which these documents were created and implemented in the respective institutions. This provides insight into the extent of content and directional overlap across policy documents, as well as the historical relationships between documents, including potential instances of policy transfer or textual replication (see also (mimetic) isomorphism as a method of handling organizational uncertainty (Powell & DiMaggio, 1991)). 

Institutional policy documents emerge from extended discussions and negotiations, led by actors situated in different spheres and therefore frames of relevance. Their 'becoming’ exemplifies hybrid figurations, particularly when ComAI transitions from being a discussion topic to becoming an actor itself. Policy documents first result from hybrid figurations, then become integrated into them, ultimately shaping new hybrid figurations. ComAI and HEIs become interconnected, with ComAI influencing both the present and future of higher education institutions. 

We conducted an AI-supported text analysis of policy documents, ethical regulations, and data protection documents implemented at German HEIs. Our examination focused on two main aspects: First, we analyzed the construction of the policy documents, focusing on their structure, their primary focus areas and topical priorities. Secondly, we examined the theoretical foundations and reference points mentioned in the policy documents, creating policy journeys which referenced their influences and allowed insight into the relations between policies of different institutions. To further understand these connections, we considered policy content and structure, and included specific phrases and terms in our AI-based text analysis, to identify cross-references between documents or cases of textual replication. Additionally, we integrated key developmental milestones of ComAI into the policy journeys to examine potential correlations between technological advances and institutional policy responses. This dual approach of retrospective analysis allowed us to investigate both the content evolution and temporal patterns in institutional responses to AI-related challenges. Germany has a very complex set up of over 400 Higher Education Institutions, which includes five different types, some of which can only be found in certain federal states. To narrow down our search, we only included types of Higher Education Institutions in our research which are represented in more than half of the federal states which excluded two types of HEIs (theological colleges and teacher training colleges). Moving forward, by using a randomized selection method, we chose, if applicable, one HEI of each remaining type (Universities, Universities of Applied Sciences, and Colleges of Art and Music) in each of the 16 federal states of Germany, adding up to a total of 46 HEIs in Germany. The websites of these 46 HEIs were then searched using the following keywords to identify relevant documents: (“KI” OR “Künstliche Intelligenz”) AND (“Leitfaden” OR “Handreichung” OR “Richtlinien”) [English equivalent: (“AI” OR “Artificial Intelligence”) AND (“Policy” OR “Guideline” OR “Regulation”)]. We focused specifically on documents focusing on general AI guidelines and policies, excluding such documents which only focused on the so-called declaration of authorship, term papers or theses. However, documents focusing on the ethical implications of AI were included in the analysis. We included policy documents which were uploaded as standalone PDFs as well as guidelines which were published on websites. We included both policies addressing students as well as staff.

Policy documents play a significant role in HEIs and are part of complex processes of negotiation. They need to be created for a versatile array of actors and need to be accepted and implemented on many different levels. We examined these policies specifically against a backdrop of German HEIs which can be used as a starting off point for international discussions and comparisons, on how policy making is happening across different countries and institutions. Regarding their topical focus, we could identify thematic overlaps in these policy documents. Many policy documents emphasized the significance of ComAI for academic merit, focusing on aspects such as plagiarism and being honest about the usage of ComAI in their work. Additionally, policy documents addressed exam regulations and how these are afflicted by the emergence of ComAI technologies. When it comes to the policy journeys our analysis reveals an increase in AI-specific policies in the last years. However, as shown, ComAI has been present in HEIs considerably longer than these recent policy publications suggest, indicating a potential gap between technological implementation and institutional regulation. This leads to questions regarding the general awareness of ComAI in the field of HEIs. While HEIs acknowledge the dynamic nature of ComAI and the need for regular policy updates, this awareness is not consistently reflected in their documentation. Some policy documents lack information about their currency or publication dates, raising questions about their practical applicability for institutional members. This gap may stem from various factors, including administrative processes, institutional priorities, or resource constraints. Furthermore, existing policies lack distinctions between different AI applications and their specific implications for teaching, research, and administration.

Doroudi, S. (2023). The Intertwined Histories of Artificial Intelligence and Education. International Journal of Artificial Intelligence in Education, 33(4), 885–928. https://doi.org/10.1007/s40593-022-00313-2 Hepp, A., Breiter, A., & Hasebrink, U. (Hrsg.). (2018). Communicative Figurations: Transforming Communications in Times of Deep Mediatization. Springer International Publishing. https://doi.org/10.1007/978-3-319-65584-0 Hepp, A., Loosen, W., Dreyer, S., Jarke, J., Kannengießer, S., Katzenbach, C., Malaka, R., Pfadenhauer, M. P., Puschmann, C., & Schulz, W. (2023). ChatGPT, LaMDA, and the Hype Around Communicative AI: The Automation of Communication as a Field of Research in Media and Communication Studies. Human-Machine Communication, 6, 41–63. https://doi.org/10.30658/hmc.6.4 Jarke, J., & Breiter, A. (2019). Editorial: The datafication of education. Learning, Media and Technology, 44(1), 1–6. https://doi.org/10.1080/17439884.2019.1573833 Gulson, K. N., Sellar, S., & Webb, P. T. (2022). Algorithms of Education: How datafication and artificial intelligence shape policy. The University of Minnesota Press. Kerssens, N., & van Dijck, J. (2021). The platformization of primary education in The Netherlands. Learning, Media and Technology, 1–14. https://doi.org/10.1080/17439884.2021.1876725 Powell, W. W., & DiMaggio, P. (Hrsg.). (1991). The New institutionalism in organizational analysis. University of Chicago Press. Selwyn, N., Hillman, T., Bergviken Rensfeldt, A., & Perrotta, C. (2023). Digital Technologies and the Automation of Education—Key Questions and Concerns. Postdigital Science and Education, 5(1), 15–24. https://doi.org/10.1007/s42438-021-00263-3 Skinner, B. F. (1958). Teaching machines. Science, 128(3330), 969–977. University of California Presidential Working Group on AI. (2021). Responsible Articial intelligence. Recommendations to Guide the University of California’s Artificial Intelligence Strategy. University of California. Williamson, B., & Eynon, R. (2020). Historical threads, missing links, and future directions in AI in education. Learning, Media and Technology, 45(3), 223–235. https://doi.org/10.1080/17439884.2020.1798995 Yu, J., & Couldry, N. (2020). Education as a domain of natural data extraction. Information, Communication & Society, 1–18. https://doi.org/10.1080/1369118X.2020.1764604