=Paper=
{{Paper
|id=Vol-3938/ELEARNING_preface
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-3938/Preface.pdf
|volume=Vol-3938
|authors=Emilija Kisić,Miroslava Raspopović Milić,Mohammed Saqr,Sonsoles López-Pernas,Miguel Ángel Conde,Olga Pavlović
}}
==None==
https://ceur-ws.org/Vol-3938/Preface.pdf
Large Language Models in Higher Education -
Perspectives, Opportunities and Limitations
Emilija Kisić 1,*, Miroslava Raspopović Milić1, †, Mohammed Saqr2, †, Sonosles López-Pernas
2, †, Miguel Á. Conde3, †, Olga Pavlović1
1 Belgrade Metropolitan University, Faculty of Information Technology, Tadeuša Košćuška 63, 11000 Belgrade,
Serbia
2 University of Eastern Finland, School of Computing, Yliopistokatu 2, 80100 Joensuu, Finland
3 Universidad de Salamanca, Escuela Politécnica Superior de Zamora – Av. de Requejo, 34, 49029 Zamora, Spain
4 University of León, Escuela de Ingenierías – Campus de Vegazana S/N, León, 24071, Spain
1
1. Introduction
Nowadays many industries are changing as a result of the rapid development in the field of Artificial
Intelligence (AI), particularly Large Language Models (LLMs) [1]. LLMs have both gained attention
in different fields and education is no exception. Traditional educational models rely on teachers
passing on knowledge to students, which puts a great burden on educators. Many educators are
overwhelmed with teaching loads including preparation of teaching materials and assessing students’
activities, while leaving little room to address individual students and their unique learning needs.
Educational institutions are facing different challenges. On the one hand, students are exhibiting high
dropout rates and low student engagement [2], while on the other hand, there are many challenges
with teaching resources [3].
The new advances in AI holds promise that including AI in education may be effective in
personalizing learning experiences, automating administrative tasks, providing real-time feedback,
and helping educators identify and address individual student needs more efficiently [4]. Therefore,
researchers have rushed to experiment with LLMs and incorporated them in educational tools for
offering personalized learning [5], generating interactive simulations [6], automated grading [7],
intelligent tutoring [8], and adaptive assessments [9]. Also, LLMs have been operationalized in other
areas of education: (i) personalized learning, (ii) intelligent tutoring systems, (iii) educational resource
creation, and (iv) assessment and feedback [10]. LLMs are used to analyze students' learning patterns
and behaviors, while providing individualized resources and feedback with a goal to improve
academic performance and student engagement. In this way personalized support and tailored
resource recommendations can be enabled.
Furthermore, LLMs have been used to provide intelligent tutoring which includes real-time
problem-solving, different learning strategies, and academic guidance using interactive dialogues
with students [11]. By creating lesson plans, tests, and study aids, LLMs can reduce the workload of
teachers while preserving high-quality, standardized content. LLMs have also been used for tracking
students’ learning progress which can provide teachers with feedback on student performance [12]–
[14]. With a proper structure and input, LLMs are able to evaluate student work and provide useful
Proceedings for the 15th International Conference on e-Learning 2024, September 26-27, 2024, Belgrade, Serbia
∗ Corresponding author.
†
These authors contributed equally
emilija.kisic@metropolitan.ac.rs (E. Kisić); miroslava.raspopovic@metropolitan.ac.rs (M. R. Milić);
mohammed.saqr@uef.fi (M. Saqr); sonsoles.lopez@uef.fi (S. López-Pernas); mcong@unileon.e (M. Á. Conde);
olga.mijailovic@metropolitan.ac.rs (O. Pavlović);
0000-0003-3059-2353 (E. Kisić); 0000-0003-2158-8707 (M. Raspopović Milić); 0000-0001-5881-3109
(M. Saqr); 0000-0002-9621-1392 (S. López-Pernas); 0000-0001-5881-7775 (M. Á. Conde); 0009-0005-3033-011X (O. Pavlović)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
feedback that promotes continuous learning and growth. Moreover, LLMs can explain the outcomes
from other forms of AI in education (such as dropout prediction or plagiarism detection) in an
accessible language to make it understandable to students, teachers and other stakeholders [15].
Despite many advantages that LLMs offer in education, there are still many challenges that need
to be overcome. The integration of LLMs in education relies on technologies such as NLP, deep
learning, data mining, and multimodal learning. In order for these technologies to have a significant
impact on education the importance of preprocessing educational data, fine-tuning models for specific
tasks, and incorporating LLMs into intelligent tutoring systems and educational platforms should be
emphasized [10]. Design and deployment of educational large models must consider real-world
educational practices and the expertise of teachers to ensure that these models effectively support
actual classroom instruction [16]–[18]. Incorporating social cognitive learning remains challenging
[19]. Some of the challenges with incorporating social cognitive learning include accurately modeling
complex human interactions, adapting to different learners, and ensuring that LLMs can provide
meaningful peer collaboration rather than just give direct answers.
Ethical considerations, such as bias in training data and maintaining student engagement avoiding
to over-rely on AI, also play a very important role in successfully integrating social cognitive learning
into educational technologies. Also, data privacy, security and their ethical usage are one of the
critical concerns when protecting and processing students’ personal information. Part of the ethical
concern raises the question on interpretability and fairness of used AI models. LLMs trained on biased
data can lead to unfair conclusions, so it is necessary to build transparent models and review
approaches. While LLMs offer a lot of opportunities, technical and resource constraints as well as a
need for human-technology balance should be taken into consideration. The deployment of LLMs
requires significant computational resources, which may be a barrier in resource-limited educational
environments. While LLMs offer valuable support, the human elements of empathy, creativity, and
interpersonal interaction remain irreplaceable in education. Still, as we currently stand, several
questions remain: how can we ensure the security of educational data? How can we prevent over-
reliance on AI technologies? And how can we foster students’ abilities for independent, active
learning?
LLMs and education may be mutually reinforcing—LLMs may enhance the efficiency of
educational processes, while the wealth of data accumulated in the education sector can, in turn,
improve LLM training and performance. Greater focus should be placed on the conditions necessary
for this development. How can we create more meaningful applications for LLMs in education? The
future of LLMs holds exciting possibilities, and we look forward to its continued evolution [20].
In conclusion, LLMs may offer advanced language generation and interactive capabilities that go
beyond what traditional teaching methods can provide. This highlights the innovative venues of AI
in education, while also redefining the roles of teachers, parents, and students. To advance education,
there is a need for well-established standards in both education and AI development, along with clear
technical guidelines and data security protocols to address practical concerns [18]. These guidelines
may help us answer the pressing questions of how to use LLMs to help students learn without
negatively affecting their cognitive development in a safe environment.
References
[1] C. Wang, J. Zhao, and J. Gong, “A survey on Large Language Models from concept to
implementation,” arXiv [cs.CL], 27-Mar-2024.
[2] K. Schnitzler, D. Holzberger, and T. Seidel, “All better than being disengaged: Student
engagement patterns and their relations to academic self-concept and achievement,” Eur. J.
Psychol. Educ., vol. 36, no. 3, pp. 627–652, Sep. 2021.
[3] P. S. Smith, P. J. Trygstad, and E. R. Banilower, “Widening the gap: Unequal distribution of
resources for K–12 science instruction,” Educ. Policy Anal. Arch., vol. 24, p. 8, Jan. 2016.
[4] H. Lin, S. Wan, W. Gan, J. Chen, and H.-C. Chao, “Metaverse in Education: Vision,
Opportunities, and Challenges,” in 2022 IEEE International Conference on Big Data (Big Data),
Osaka, Japan, 2022.
[5] M. Park, S. Kim, S. Lee, S. Kwon, and K. Kim, “Empowering personalized learning through a
conversation-based tutoring system with student modeling,” in Extended Abstracts of the CHI
Conference on Human Factors in Computing Systems, Honolulu HI USA, 2024.
[6] Z. Zhang, Z. Zhou, and L. Liu, “Simulating classroom education with llm-empowered agents Z,”
J Li arXiv.
[7] C. Impey, M. Wenger, N. Garuda, S. Golchin, and S. Stamer, “Using large language models for
automated grading of student writing about science,” Int. J. Artif. Intell. Educ., Jan. 2025.
[8] J. Stamper, R. Xiao, and X. Hou, “Enhancing LLM-based feedback: Insights from intelligent
tutoring systems and the learning sciences,” in Communications in Computer and Information
Science, Cham: Springer Nature Switzerland, 2024, pp. 32–43.
[9] A. Goslen, Y. J. Kim, J. Rowe, and J. Lester, “LLM-based student plan generation for adaptive
scaffolding in game-based learning environments,” Int. J. Artif. Intell. Educ., Jul. 2024.
[10] W. Gan, Z. Qi, J. Wu, and J. C.-W. Lin, “Large language models in education: Vision and
opportunities,” arXiv [cs.AI], 22-Nov-2023.
[11] Z. Qib, J. Wua, and P. S. Yuc, Large Language Models for Education: A Survey Hanyi Xua.
Wensheng Gana.
[12] N. S. Raj and V. G. Renumol, “A systematic literature review on adaptive content recommenders
in personalized learning environments from 2015 to 2020,” J. Comput. Educ., vol. 9, no. 1, pp. 113–
148, Mar. 2022.
[13] Z. Wang, W. Yan, C. Zeng, Y. Tian, and S. Dong, “A unified inter pretable intelligent learning
diagnosis framework for learning perfor mance prediction in intelligent tutoring systems,”
International Journal of Intelligent Systems, vol. 2023, 2023.
[14] J. Rudolph, S. Tan, and S. Tan, “ChatGPT: Bullshit spewer or the end of traditional assessments
in higher education?,” Journal of Applied Learning and Teaching, vol. 6, no. 1, 2023.
[15] S. López-Pernas, Y. Song, E. Oliveira, and M. Saqr, “LLMs for Explainable Artificial Intelligence:
Automating Natural Language Explanations of Predictive Analytics Models,” in Advanced
Learning Analytics Methods: AI, Precision and Complexity, M. Saqr and S. López-Pernas, Eds.
Cham: Springer Nature Switzerland, 2025.
[16] R. Marshall, A. Pardo, D. Smith, and T. Watson, “Implementing next generation privacy and
ethics research in education technology,” Br. J. Educ. Technol., vol. 53, no. 4, pp. 737–755, Jul.
2022.
[17] R. F. Kizilcec and H. Lee, “Algorithmic fairness in education,” in The Ethics of Artificial
Intelligence in Education, New York: Routledge, 2022, pp. 174–202.
[18] W. Xu, Z. Wei, and P. Yan, “Exploring the application of Large Language Models in
Instrumentation and Control Engineering education: A questionnaire survey and examination
performance analysis,” Eur. J. Educ., vol. 60, no. 1, Mar. 2025.
[19] N. Bian et al., “Influence of external information on large language models mirrors social
cognitive patterns,” IEEE Trans. Comput. Soc. Syst., pp. 1–17, 2024.
[20] M. Y. Mustafa et al., “A systematic review of literature reviews on artificial intelligence in
education (AIED): a roadmap to a future research agenda,” Smart Learn. Environ., vol. 11, no. 1,
Dec. 2024.