Enhancing E-Learning Interaction through Customized LLM-Based Chatbots – A Prototype Study

Autores

  • Martin Würflein T-Systems
  • Michael Müller Ansbach University of Applied Sciences
  • Sigurd Schacht Ansbach University of Applied Sciences

Palavras-chave:

E-Learning, Large Language Models, Chatbots, Instructional Design

Resumo

 Introduction and Objective: E-learning supports both synchronous and asynchronous education and plays an important role in knowledge transfer and management. The article explores how generative AI, specifically Large Language Models (LLMs), can enhance learner interaction in E-learning environments. Based on theoretical insights into instructional and conversational design, a chatbot prototype was developed for a blended, practice-oriented course on "Container Orchestration and Coding with DevOps" targeting IT professionals.

Methodology: The chatbot integrates an open-source model (LLaMA 3.3) and employs retrieval-augmented generation (RAG) techniques within an online IDE. It offers three modes: task support, task recommendation, and Q&A, enriched with context-specific documentation. The prototype was developed iteratively, informed by theoretical research, hands-on experimentation, and the author’s long-term experience as a trainer. An initial user evaluation was conducted through an online survey. A follow-up evaluation based on an improved chatbot is planned for May 2025 and will be incorporated into the final study.

Results: Although technically functional, user engagement during the first four-week evaluation was limited. About 10% used the chatbot occasionally, while 70% did not use it at all. Preferred alternatives included traditional course materials (100%), online sources (~75%) and general-purpose chatbots like ChatGPT (~45%). Reasons for low adoption included time constraints, preference for existing online chatbots, or general lack of interest in using chatbot-based support.

Conclusions: The findings reveal that, while LLM chatbots hold promise for E-learning, effective onboarding strategies, user-centered design, and AI literacy are critical. Long-term studies are needed to evaluate their real-world impact across educational settings.

Downloads

Não há dados estatísticos.

Publicado

2025-12-05

Como Citar

Würflein, M., Müller, M. ., & Schacht, S. . (2025). Enhancing E-Learning Interaction through Customized LLM-Based Chatbots – A Prototype Study. Anais Do Congresso Internacional De Conhecimento E Inovação – Ciki, 1(1). Recuperado de https://proceeding.ciki.ufsc.br/index.php/ciki/article/view/1696