Assessing the usefulness of large language models in geotechnical education: EduHackathon at the third workshop on the future of machine learning in geotechnics (3FOMLIG)

This article reports on the process and outcomes of the EduHackathon “Assessing the usefulness of Large Language Models in geotechnical education” held in Florence, Italy, as a session of the third Workshop on the Future of Machine Learning in Geotechnics (3FOMLIG) on October 16, 2025. Recognizing the increasing implementation of Large Language Models (LLMs) and Generative AI (Gen AI) for domain-specific geotechnical workflows, the event aimed to explore their potential utility in geotechnical engineering education. The EduHackathon was the culmination of a collaborative effort by a core group of geotechnical educators who presented and discussed with conference attendees a series of studies conducted, prior to the meeting, to evaluate LLM and Gen AI tools from both student and instructor perspectives. The article details aims and main outcomes of these studies, as well as the main conclusions of the discussions carried out in person, which were also driven by the results of a survey previously undertaken by 3FOMLIG participants. The findings suggest that while LLMs offer significant potential for enhancing interactive and reflective learning, they should be viewed as flexible pedagogical tools rather than autonomous instructors. It can be emphasized that the educational value of these tools depends heavily on careful integration with sound teaching principles and human supervision to mitigate potential risks, such as cognitive debt, inaccuracies, and the loss of critical information. Ultimately, the article proposes to keep exploring the potential benefits and the main limitations associated with the use of these tools within educational modules in soil mechanics and geotechnical engineering, where LLMs may serve as collaborative interlocutors to both geotechnical educators and students, in a sort of epistemic co-production with human expertise.