The first episode of the “Navigating the Intersection of LLMs and Ethical Considerations” podcast opened an important conversation on one of the most relevant questions facing education today: how can higher education benefit from large language models while protecting ethics, integrity, and fairness?
The episode featured Andreas Pompomis from GrantXpert, an ELOQUENCE project partner, whose work connects large language models with the evolving needs and challenges of higher education. Together with the host, Andreas explored how AI is already reshaping the way students learn, complete assignments, and interact with knowledge.
One of the key themes of the conversation was bias. Large language models are trained on vast amounts of data, and if that data contains social, cultural, gender, racial, or socioeconomic biases, these can appear in the model’s outputs. While LLMs have the potential to make some educational processes more consistent, for example by supporting assessment or feedback, they may also introduce new forms of unfairness. Students with different writing styles, non-native speakers, or those from diverse educational backgrounds could be unintentionally disadvantaged if AI systems are not carefully evaluated and continuously improved.
The discussion also addressed the growing concern around academic integrity. LLMs can support learning by explaining difficult concepts, generating practice questions, or helping students structure their ideas. However, they can also be used to bypass the learning process entirely. Students may rely on AI-generated essays, coding solutions, or problem-solving tasks without fully understanding the material. This creates new challenges for educators, who must rethink where the line lies between legitimate assistance and cheating.
According to Andreas, higher education institutions need clear guidelines on the acceptable use of AI tools. At the same time, students should be supported in developing AI literacy, so they understand not only what these tools can do, but also their limitations, risks, and ethical implications. Rather than banning AI completely, the focus should be on helping students use it responsibly.
Another important point was the need to redesign traditional assessment models. Take-home essays and standard problem sets may no longer be sufficient in an AI-enabled learning environment. Instead, universities may need to place greater emphasis on oral exams, in-class exercises, project-based learning, presentations, peer review, and tasks that require critical thinking, creativity, personal insight, and real-world application.
The episode also highlighted the potential of LLMs to support personalized learning. When used responsibly, they can provide tailored explanations, additional exercises, or advanced challenges depending on each student’s needs. Still, equal access remains essential. Without reliable internet, suitable devices, and inclusive system design, AI tools could widen the gap between privileged and under-resourced students.
Ultimately, the conversation showed that LLMs are neither a simple threat nor a simple solution. Their impact on higher education will depend on how institutions, educators, developers, and students choose to integrate them. The challenge ahead is to create learning environments where AI supports understanding, strengthens fairness, and encourages students to think more deeply, not less.
Listen to the full episode here.
