The onset of artificial intelligence (AI) tools has afected various areas in business and society. AI has also begun changing the way higher education institutions carry out their work, leading educators, students, and university staf to adapt to the peculiarities of AI. While research has begun investigating the impact of AI tools on higher education, these studies largely focus on specific countries, lacking the integration of international perspectives. Since there is a need for information on available resources and cultural diferences, this integration should prove highly beneficial to help countries develop strategies to enable a safe, inclusive, and sustainable use of AI in higher education. We conducted five focus group interviews with N = 38 participants from three EU countries (Germany, Spain, and Greece), combining international experiences in the use of AI in higher education, possible benefits and challenges, and requirements for a sustainable use. Using a collection of diferent methodical approaches, we fostered an open exchange with teachers, students, and staf from higher education. Our preliminary findings ofer a cross-national perspective on the use of AI in higher education.
Even though artificial intelligence (AI) has its roots in the 1950s [
To ensure a safe, inclusive, and sustainable implementation of AI, the technology needs to be studied
in higher education institutions. Existing research has started to theorize how this implementation
might be carried out (e.g., [
To enable this integration of international perspectives and experiences, we conducted five focus group interviews with a total of N = 38 stakeholders from higher education in Germany, Spain, and Greece. Participants included higher education teachers, students, and administrative university staf. Following a semi-structured set of questions, we deployed diferent methods within the focus group interviews, generating various types of artifacts (such as recordings and digital notes). Using qualitative content analysis (QCA), we clustered answers to collect stakeholders’ experiences, needs, and perceived challenges.
Our preliminary findings contribute to the growing body of research on the use of AI in higher education by ofering insights into diferent countries’ approaches for AI implementation, as well as their experiences and needs. These findings can assist stakeholders and researchers from across diferent EU member states in developing their own comprehensive AI strategies, balancing protective measures, universal design, and environmental stewardship.
Touching briefly upon work regarding how higher education institutions should implement AI, a
common recommendation states that academic integrity might sufer unless institutions establish
comprehensive guidelines on how and for what AI may or may not be used [
Crucially, most of previous research focuses on holistic analyses, rather than involving stakeholders
directly by carrying out open exchanges. Individual publications suggest involving educators [
To explore international perspectives on AI in higher education, we conducted five focus group interviews with a total of N = 38 participants from three EU countries (Germany, Spain, and Greece). The interviews were carried out between January and March of 2025 and were conducted by diferent authors in the diferent countries. They followed a shared semi-structured questionnaire, ensuring that the same topics were covered. The interview guide covered six diferent parts, comprising the participants’ familiarity with AI in general, their interest in AI, which benefits they expect from AI for students, which benefits participants expected for their own academic work, what concerns were present regarding the implementation of AI in higher education, and whether they felt prepared for using AI in their daily work context, or whether they needed additional skills or support to do so.
Each focus group consisted of five to twelve participants and involved teachers (both lecturers and professors), research staf, administrative staf (e.g., university decision-makers), didactics experts, and students (both from bachelor and master programs). The compositions difered, with some focus groups involving no students at all, while others actively fostered exchanges between teachers and students. An overview of participants, along with their background per focus group can be found in Table 1 below.
To gather as many insights as possible, focus groups across the diferent countries were conducted
slightly diferently, yielding diferent artifacts. The artifacts that were generated during the focus groups
included interview recordings, notes summarizing the participants’ main points, and clusters of digital
notes, as well as various quantitative data. For three focus groups in Germany, participants self-assigned
superhero identities, which allowed for an anonymous connection of their recorded utterings and the
notes they left on a shared digital whiteboard. All artifacts were anonymized and transcribed into
written documents for further analysis, following common recommendations [
All focus groups followed a shared interview guide, which is included in the digital appendix provided at OSF under the following link: osf.io/p8xuv.
Artifacts were then analyzed using QCA, inductively creating a coding scheme. After the first focus group interview had been coded in its entirety, the coding scheme was shared with the other countries’ authors, who then applied and inductively extended the coding scheme to create a shared collection of codes mentioned by participants. This combined coding scheme was then used to derive and cluster participants’ responses into fitting categories, resulting in a total of 36 categories, which are laid out in more detail in the following chapter.
For n = 14 participants, we quantified their experience with AI in their daily work. These university stakeholders were using AI tools semi-regularly (M = 4.32, SD = 1.88), with the lowest value being 1 (“never using AI for work”) and the highest value being 7 (“using AI all the time for work”). A rough organization of respondents into minimal users, intermediate users, and expert users further revealed that for most findings, the frequency of AI use was connected to whether participants saw more benefits or challenges in the application of AI into their daily workflows. Additionally, we identified the tools and models that were being used already, resulting in a list of 18 diferent tools: ChatGPT (13 mentions), Deepl (6 mentions), Consensus (2 mentions), Elicit (4 mentions), Midjourney (2 mentions), Copilot (3 mentions), Gemini (2 mentions), Grammarly, Dall-E, Connected Papers, QuillBot, Copy.ai, Bard, Claude, Otter.ai, MonkeyLearn, Turnitin, and LaTeX AI (one mention each).
Regarding the possible applications of AI for both teachers and students, we identified a total of seven categories: Efectiveness, quality improvement, creativity, inclusion & accessibility, personalized learning support, assisted decision-making, and resource finding . Under efectiveness , students and teachers described that using AI helps them conduct their daily tasks more quickly. For example, one participant said: “I’m doing a lot of programming at the moment, so I have a massive increase in productivity thanks to ChatGPT”. Similarly, participants reported using especially generative AI for quality improvement, by for example increasing text readability, checking for grammar and spelling mistakes, and fixing errors in their code. Additionally, participants underlined that they use AI tools to foster creativity, for example through the generation of new ideas, images or content such as blog entries: “When you need a transition from one topic to another, it’s super helpful to actually get an output with small inputs that you can then use to create your own ideas”.
Beyond these performance benefits, participants also highlighted the boost of inclusivity and accessibility, since existing content can be checked for accessibility or lecture videos can be enriched with subtitles for deaf students, thereby improving access to course content independent from disabilities. In this context, free access to many tools was seen as reducing economic barriers, along with the possibility to use generative AI as a personal translator. Specifically students stressed receiving personalized learning support from AI tools, since lesson content can be summarized and changed to the preferences of individual students.: “The individual learning approach is strengthened and [the tool] can then react more readily to a person’s personal sensitivities”. In addition, students highlighted that AI can help them in finding resources , such as additional publications or tutorials for their course content. Finally, teachers suggested that AI can help their decision-making, for example when grading exams: “You can get advice from the AI tool to see whether you make some outlier decisions”.
Regarding potential challenges of AI use in higher education, participants mentioned a total of nine diferent aspects: Ethical problems, missing knowledge, unclear rules, loss of cognitive abilities, replacement of human labor, continuous advancement of AI, questioning the need to learn traditional skills, limits of AI, homogeneity and AI as an essential skill. The ethical problems participants identified comprised unequal opportunities caused by unequal skill levels regarding the use of AI, the reproduction of stereotypes and discrimination through biases in training data, and the unfair use of AI, for example for submitting theses or assignments that should have been written by the students themselves: “If we have students who barely use AI [...] compare their [submissions] with someone who has used a lot of AI, [we cannot know] whether he or she has really informed themselves about [the topic]”.
Additional ethical concerns surround sustainability – as AI tools need a lot of energy –, a possible loss of digital sovereignty, and increase of fake news, plagiarism, and misinformation, distinguishing between AI and human-made academic work, and possible negative efects on self-esteem (for example when comparing one’s own work against AI-generated content). The lack of knowledge suggested by participants comprises diferent sub-codes: Knowledge on biases, over-reliance, manipulation, and data protection. Even though most participants were aware that AI is often biased, they claimed to lack knowledge on the types and extent of these biases. Further, there were concerns that participants lack knowledge on the impact of over-reliance on AI, manipulation of results from foreign forces, and data protection laws and regulations.
Participants were further concerned that using AI might lead to a loss of cognitive abilities or the replacement of human labor. At the same time, especially teachers explained that the continuous advancement of AI forces them to stay up-to-date with the rapid emergence of new tools, with some participants questioning whether there is still a need to learn traditional skills: “And it is simply more dificult to convey that you should be able to do [things traditionally], this necessity is called into question”.
Interviewees also referred to the lack of true social interaction and emotional connection and technical limitations, such as limitations in processing power and more complex tasks, like for example in physics calculations. A further challenge is that many institutions still have unclear rules on how AI may be used: “I would like to use more AI in my job, but don’t know whether that’s allowed [...] or whether that could entail data protection related issues”. Beyond that, participants remarked on the homogeneity of outputs, and a lack of ideas regarding the added value that AI might ofer: “I totally forget that this tool exists. [...] And apart from that, I don’t think I really had any impulse to use it [...] because it’s not that commonplace for me yet”.
We asked participants what they required to feel prepared for broader AI use. Their responses include skills, knowledge, uniform rules, resources, and self-initiative in learning AI. The skills that most participants needed are both in regards to their AI expertise – i.e., how they can operate AI tools – and the critical reflection of outputs, since many participants stressed that they did not feel confident in always recognizing false information generated by AI.
Similarly, some staf and students lack knowledge on AI, on how it works and how it can be used, as well as knowledge on content, which includes knowing how to interpret diferent types of outputs. From their institutions, most participants agreed that they need uniform rules and guidelines on the use of AI, resources for efective AI usage – such as tutorials or workshops – and time to be able to take self-initiative for learning AI usage: “I have done some training, but I haven’t really had the time to really get into it”.
We set out to gain an integrated understanding into the experiences with and opinion on the use of AI in higher education. For this, we conducted focus group interviews in three EU countries. These focus groups included diferent stakeholders from higher education institutions, such as teachers, students, and didactic and administrative staf. Our preliminary findings show that there is a high variability in how frequently and for what purposes AI tools are being used. The participants highlight many benefits, such as more personalized, inclusive, and efective learning and teaching. At the same time, there are various concerns such as unclear legal frameworks, university guidelines and the lack of knowledge on how AI afects universities in the future. Regarding feelings of preparedness, some participants feel well-prepared, while others require time and resources to understand the way AI works and how its negative side efects can be mitigated. While we cannot claim generalizability since we conducted a qualitative study rather than quantitative analyses, the focus group interviews ofered individual and in-depth perspectives from diferent stakeholders at universities.
At this early stage of research on AI in higher education, this introduces a breadth of findings, which provides opportunities for more focused further research. Nevertheless, our work has a limitation: The generation of diferent artifacts and slight diferences in the detailed method might introduce biases that make comparisons of data problematic. However, we did not compare the countries’ results, but rather decided to pool results to gain integrated insights.
Our current steps include carrying out more focus groups in a fourth EU country (Italy) and an even deeper analysis of how the attitudes of higher education stakeholder might be quantified and measured to achieve generalizability. Additionally, we are examining how diferent countries vary in available resources and constraints (such as legal regulations). Through this, we are hoping to gain a deeper understanding of how higher education stakeholders, and especially teachers, can be enabled to confidently reap the benefits of AI while carrying out their teaching duties.
The present work resulted from the IDEAL (Integrating Data Analysis and AI in Learning experiences) project and has been funded by Erasmus+ KA220-HED - Cooperation Partnerships in Higher Education (2024-1-IT02-KA220-HED-000251425). The European Commission support for the production of this publication does not constitute an endorsement of the contents which reflects the views only of the authors, and the Commission cannot be held responsible for any use which may be made of the information contained therein.
The authors have not employed any Generative AI tools for writing this manuscript.