Summer Developments in the Field of AI in Education

At first glance, the summer’s shifting opinions on the use of AI in education seem chaotic. Some call for a strict ban on its use, while others advocate for its full integration into the classroom. But it is precisely in this “tension” that we see an opportunity for open dialogue. At Oxford, too, they believe that with improved guidelines, ongoing training, and the involvement of students in shaping the rules for AI use, we can achieve more than with bans that are often counterproductive. In practice, this means that educational institutions must rethink their policies, provide clearly defined examples of acceptable and unacceptable AI use, and establish support resources for the ethical use of these tools. All of this takes time, but it also presents an opportunity to strengthen trust between the faculty and the student community (more information).

And it is precisely among doctoral students that there has recently been the greatest uncertainty due to the lack of clear rules regarding the use of tools such as ChatGPT. Students are wondering how to document the role of AI in their work, how to cite the assistance provided by AI tools, and how to avoid covert misuse. Advisors and educational institutions must take an active role by organizing workshops, defining expectations in research ethics guidelines, and fostering open dialogue between advisors and students. Transparency should remain key to maintaining the integrity of research (more information).

The following news item is intended to encourage educators to consider, even in light of institutional guidelines regarding the use of AI, their own practices or rules for students at the course level. Five educators have compiled their examples, which represent a wide spectrum of approaches, ranging from stricter restrictions on use to the promotion of open and structured integration of AI tools. Policies and practices regarding the use of AI should be diverse, pragmatic, educational, and developed through dialogue between educators and students (more information).

The summer also brought red flags to the process of publishing scientific articles. It was discovered that AI tools can generate articles that appear academically credible while simultaneously attempting to bypass standard detection mechanisms. What’s more, AI-generated contributions—or even just references that lead nowhere—have been published in the works of individuals who list real, credible authors as co-authors (without those authors’ knowledge). Consequently, publishers and editors are recognizing the need for more robust procedures—from the stage of verifying authors’ identities to new protocols for verifying source data and documentation. Indeed, the consequences of a lack of response can be severe—trust in the scientific communication of knowledge can wither away in this way, harming all of us who base our work on open and verifiable knowledge. An initiative has emerged among various publishers to provide additional training for staff in the field of open access and to develop shared tools for exchanging information about “suspicious” cases of submissions (more information). Have you ever come across a paper where your name appeared in the references alongside a paper that doesn’t exist? Some faculty members at our university have already had this experience.

Various scammers are also exploiting the apparent online legitimacy of universities. Recent research shows that a network of fake university websites is emerging, featuring sophisticated copies of content, generated images, and chat rooms designed to target vulnerable users. In this area as well, we must be proactive, regularly monitor our online presence, and educate students on how to verify the authenticity of sources. As educators, we can contribute to prevention by raising students’ awareness and incorporating the development of source-verification skills into the introductory courses students take (more information).

Educational institutions are increasingly opting for customized training on the use of AI tools. One such example comes from Indiana University in Bloomington, Indiana. They have developed a modular university-wide genAI course that is available to students, faculty, and administrative staff. The course is designed not only to explain how to use these tools but also to teach ethics, fact-checking, and responsible use. Through short modules, self-paced learning, practical examples, and the opportunity to earn certificates, they have sought to encourage participants to complete the course and thereby raise the basic digital literacy of the entire community. Educators can adapt the course content to their own needs, incorporate their own use cases, provide mentorship, and bridge the gap between technological capabilities and curricular goals (more information). The UK Department for Education has also launched free training for educators and leaders of educational institutions on how to use AI in an effective and responsible manner. Participants have the opportunity to take various modules for educators: Understanding AI in Education, Interacting with Generative AI, Safe Use of Generative AI, and Use Cases for Generative AI; as well as a module for educational institution leaders: Using AI in Education—Support for Leadership (more information).

In at least our last two AI newsletters, we’ve written extensively about the new situation regarding student assessment at the end of the semester—many instructors have decided to switch back to traditional methods of assessment, namely in-person, classroom-based written exams on paper. The authors believe that even though returning to handwritten exams may seem like a simple solution to the challenge of assessment in the age of AI, it is in fact a return to a symptomatic solution. The real challenge lies not only in AI tools but in the very method of assessment itself. Educators who focus on the learning process—such as drafting, reflection, portfolios, and presentations—reduce the potential for cheating while simultaneously increasing the authenticity of the learning process. Such an approach teaches students various skills and enables AI tools to become “collaborators” in learning, rather than a means of cheating. In practice, this means the need to redesign assignments and introduce assessment elements that require personal contribution and demonstrate understanding (more information). In connection with the use of AI as a “collaborator” in learning, we would like to highlight another dilemma that has been emerging for some time and is challenging diversity within the academic community. If we rely too heavily on written expression using various large language models, this could lead to uniform writing styles and ideas. If we all start using the same writing styles, how innovative will we actually be able to be, and how diverse will our knowledge and approaches to solving challenges be? This challenge is further evidence that we must promote tasks in education that require originality, interdisciplinary connections, and personal interpretation. In other words, tasks that are difficult for algorithmic replication to reproduce (more information).

Recently, we’ve been seeing another phenomenon or trend in the news—one that isn’t particularly technical, but is quite significant for society. Students are increasingly turning to chat rooms as a quick source of support when facing personal and/or emotional distress. This raises important questions for counseling services and educational practices, where we must ask ourselves how to ensure that students receive professional help when they need it and, at the same time, how to educate them about the limitations of AI in the role of a “virtual conversation partner” (more information).

Before we wrap up today’s newsletter, we can’t overlook the summer news about OpenAI’s new model, ChatGPT 5, which differs significantly from previous models. With this model, they’ve introduced a unified system that uses specialized models for more in-depth answers to complex questions and highly responsive models for answers to more everyday questions. Based on our prompt, ChatGPT 5 automatically identifies which model is needed to provide the best answer (more information). Have you had a chance to try out the model?

To conclude, a few words about the even more intense debates over whether AI will replace educators. Here, we must not forget that, at least for now, someone still has to design, monitor, and evaluate AI systems. And this is precisely where the academic sphere plays a crucial role and has an opportunity to develop models of collaboration in which expertise sets the boundaries, while algorithms enable scalability. Furthermore, universities remain the driving force in fostering a critical stance toward various commercial forces seeking to privatize educational processes. Interdisciplinary collaboration and the promotion of the public interest are the cornerstones of responsible implementation (more information).

Today’s news roundup is not merely a chronicle of recent events; it is a call to action. We invite you to initiate discussions within your departments about clear practices for using AI and reforming assessment and teaching methods.

Authors: Sanja Jedrinović, Mateja Bevčič, Eva Kern Nanut, Eva Škraba, University of Ljubljana Center for the use of ICT in pedagogical process