Generative Artificial Intelligence (GenAI) tools such as ChatGPT have been transforming higher education since November 2022. Nowadays, students and staff alike use AI tools in higher education for diverse tasks such as text translation, checking grammar in assignments, grading homework, and communicating with one another (Wang et al., 2024; Abdullayeva & Musayeva, 2023; Fuchs, 2023; Rudolph et al., 2023). Studies on the uptake of AI amongst university students show relatively high adoption rates ranging from 73% (Fošner, 2024) to 92% (Freeman, 2025), varying by field of study and gender (Stöhr et al., 2024). However, this accelerated adoption poses potential risks to academic integrity and student learning. GenAI tools produce outputs difficult to differentiate from student work, complicating grading and raising questions about the extent of student learning and the value of higher education (Stöhr et al., 2024; Farazouli et al., 2024). Faculty concerns center on authorship misattribution, detection limitations, and pedagogical implications for assessment design  (Wu et al., 2024). As a result of these concerns, higher education policymakers have accelerated development and adoption of institutional policies. However, many policies are not guided by learners' experience and practical concerns, leading to inefficient responses and mismatch in resources. 

To address this, we adopt a novel computational social science approach combining policy analysis with online discourse to examine the utilisation and governance of GenAI within higher education. This methodological design enables us to assess the extent to which institutional guidelines align with the practical experiences of higher education stakeholders. First we analyse discussion surrounding GenAI in higher education, drawing on lived experiences shared online through 290 posts from academic communities on Reddit. This organic user-generated data reveals concerns including usage of GenAI, automated detection avoidance, faculty’s detection methods. Novel findings include discussions on the value of learning and increased workloads as well as unclear policies. We then conduct a comparative policy analysis of policy texts on AI usage from leading universities in the Anglosphere. Our analysis of policy texts from leading university groups shows that while such texts outline permissible GenAI use, they focus on abstract concerns including academic integrity, data privacy, and fair access. Furthermore, a comparative policy analysis identifies areas of divergence between university offered guidance and online discourse. The contrast with Reddit discussions highlights a significant gap: while policy statements remain high‑level and theoretical, the texts lack concrete guidance on practical issues that practitioners grapple with managing workload, navigating unreliable detection technologies, and addressing integrity and learning concerns, as well as students discussion on academic integrity and the value of coursework. 

The findings shed light on alignments and gaps between institutional policy and everyday experiences, emphasising the necessity of designing human-centred policies. We use these findings as a basis when presenting five holistic policy recommendations based on this study’s findings to bridge the gap between the lived experiences of those using GenAI with higher education institutions and academic policies. These recommendations complement each other to provide an agile approach to ensuring that higher education institutions manage AI use in a way that is both principled and practical. Our recommendations define acceptable AI use, introduce transparency on AI systems, redesign assessments, support faculty capacity, and embed fairness with institutional policies to encourage more responsible GenAI use in higher education. 

We adopt a novel computational social science approach, combining the collection and analysis of organic user-generated data from Reddit discussions within higher education communities across the Anglosphere alongside policy documents issued by leading higher education institutions. We employ posts from users on Reddit, a social media website with 97.2 million daily and 365.4 million weekly active users. We focus our data collection from Reddit posts on higher education communities in the Anglosphere, specifically the United Kingdom, United States, and Canada. All selected subreddits relate to higher education experiences. To identify AI-related conversations, we filtered posts in each subreddit by the search terms. We employ the PRAW API (Application Programming Interface) implemented in Python to collect data using above terms from 30 November 2022 to 12 June 2025. We employ a filter to remove irrelevant posts and comments that include the phrases and collect 312 posts in total. Our institutional analysis focuses on four leading university consortia: the Group of Eight (Australia), Russell Group (UK), Ivy League (US), and U15 (Canada). These associations share two defining characteristics justifying their selection. First, they represent English-speaking higher education systems, ensuring linguistic and cultural comparability. Second, they align geographically with our Reddit data sources, which predominantly originate from the US, Canada, UK, Australia, and Ireland through national subreddits and associated networks. Each alliance published a collective AI policy except for the Ivy League. We selected policy documents which match the criteria by Cohen et al. (2017).The final dataset therefore comprises a set of eight policy texts/guidelines that are institutionally authenticated, publicly available, and explicitly centred on AI in higher education. Our analysis proceeded in two phases: (1) inductive coding of Reddit posts in academic communities to create a bottom-up thematic scheme around the use of GenAI tools in HE context; (2) deductive coding of policy documents using the inductively derived scheme to examine convergence and gaps between “voices from the field” and official guidance on policies. Hybrid inductive-deductive thematic analysis was utilised to investigate both bottom-up patterns in public online discussion in Reddit higher education communities and their alignment with top-down institutional policies on AI. Top-down and bottom-up coding methods, which are considered as two distinct approaches to thematic analysis (Braun & Clarke, 2006; Xu & Zammit, 2020), allow us to reveal emergent concepts while also interrogating predefined, practice-relevant constructs (Fereday & Muir-Cochrane, 2006; Hsieh & Shannon, 2005).

A common finding in both Reddit discourse and institutional policy documents is the widespread concern that AI tools constitute a serious threat to academic integrity. However, while policy documents establish principled frameworks, they do not fully address the practical concerns that dominate discussions on Reddit. Policy statements from major university consortia, the Group of Eight (Go8) in Australia, the Russell Group (UK), Ivy League universities (US), U15 Canada and the Irish Universities Association (IUA), establish a clear distinction between acceptable and unacceptable uses of generative AI. All university policies evaluated emphasise clear guidelines, ongoing support for AI literacy, and modifications to teaching and evaluation methodologies, with institutions and instructors carrying primary responsibility. On the other hand, policy documents offer extensive guidance on the methods for using AI but are notably silent on AI detection methods. Furthermore, the worries about implementation in Reddit discussions, such as the increased workload of faculty and false accusations of cheating were not sufficiently addressed in the policy documents, while the theme of "data security and privacy," emphasised in the policy documents, did not figure prominently in discussions. The findings align with early post-2022 literature documenting how higher education reacted swiftly but often superficially to the GenAI surge. Improving AI policy design with grounded, user-centered evidence can bridge the gap. We make a number of recommendations that complement each other to provide an agile approach to ensuring that higher education institutions manage AI use in a way that is both principled and practical. These include clarifying acceptable use, transparency in AI, redesigning assessment to preserve the value of learning, supporting faculty capacity, and promoting fairness in higher education.

Abdullayeva, M. and Musayeva, Z.M., 2023. The impact of Chat GPT on student's writing skills: An exploration of AI-assisted writing tools. In International Conference of Education, Research and Innovation (Vol. 1, No. 4, pp. 61-66). Braun, V. and Clarke, V., 2019. Reflecting on reflexive thematic analysis. Qualitative research in sport, exercise and health, 11(4), pp.589-597. Cohen, L., Manion, L., & Morrison, K., 2007. Research methods in education (6th ed.). Routledge/Taylor & Francis Group. Farazouli, A., Cerratto-Pargman, T., Bolander-Laksov, K. and McGrath, C., 2024. Hello GPT! Goodbye home examination? An exploratory study of AI chatbots impact on university teachers’ assessment practices. Assessment & Evaluation in Higher Education, 49(3), pp.363-375. Fereday, J. and Muir-Cochrane, E., 2006. Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. International journal of qualitative methods, 5(1), pp.80-92. Fošner, A., 2024. University students’ attitudes and perceptions towards ai tools: implications for sustainable educational practices. Sustainability, 16(19), p.8668. Freeman, J., 2025. Student Generative AI Survey 2025. https://www.hepi.ac. uk/2025/02/26/student-generative-ai-survey-2025, 2025. Accessed 2 June 2025. Fuchs, K., 2023, May. Exploring the opportunities and challenges of NLP models in higher education: is Chat GPT a blessing or a curse?. In Frontiers in education (Vol. 8, p. 1166682). Frontiers Media SA. Hsieh, H.F. and Shannon, S.E., 2005. Three approaches to qualitative content analysis. Qualitative health research, 15(9), pp.1277-1288. Rudolph, J., Tan, S. and Tan, S., 2023. ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?. Journal of applied learning & teaching, 6(1), pp.342-363. Stöhr, C., Ou, A.W. and Malmström, H., 2024. Perceptions and usage of AI chatbots among students in higher education across genders, academic levels and fields of study. Computers and Education: Artificial Intelligence, 7, p.100259. Wang, S., Wang, F., Zhu, Z., Wang, J., Tran, T. and Du, Z., 2024. Artificial intelligence in education: A systematic literature review. Expert Systems with Applications, 252, p.124167. Wu, C., Wang, X., Carroll, J. and Rajtmajer, S., 2024. Reacting to generative AI: Insights from student and faculty discussions on Reddit. In Proceedings of the 16th ACM Web Science Conference (pp. 103-113). Xu, W., & Zammit, K., 2020. Applying thematic analysis to education: A hybrid approach to interpreting data in practitioner research. International journal of qualitative methods, 19, 1609406920918810.