LLM AGENTS: ADVANCING TRULY PERSONALIZED ONLINE EDUCATION

Since the advent of ChatGPT, the development and release of Large Language Models (LLMs) have accelerated at an unprecedented pace.The most advanced LLMs, such as Deepseek R1 and ChatGPT-4o, exhibit sophisticated reasoning capabilities with significant potential for transforming education.These models can enable personalized learning by dynamically adapting to individual preferences and needs.However, despite their growing adoption, the real-world impact of LLMs in education remains limited, with few implementations demonstrating measurable effectiveness.In this paper, we investigate the potential of LLMs to transform education and propose a framework centered on LLM agents in online learning environments.First, we advocate for a new generation of online learning platforms where LLM agents are not merely add-ons but integral components, shifting from traditional Learning Management Systems (LMS) that function as static content repositories to dynamic, interactive ecosystems.In these environments, agents actively "intervene" to support and guide learners.Second, we emphasize the development of multi-agent systems that facilitate content delivery, foster interaction, promote critical thinking, enhance engagement, and provide personalized feedback.Finally, we highlight the necessity of rigorous evaluation methods to assess the actual impact of these technologies on learning outcomes.This paper presents the initial findings of our approach, which focuses on automating content generation in online learning environments while addressing predominant learning styles: Visual, Auditory, Read/Write, and Kinaesthetic.We aim to convert traditional text-based content into multimodal formats tailored to diverse learning preferences.While this transformation presents significant challenges, it is greatly facilitated by LLM agents leveraging advanced generative AI capabilities, including text-to-text, text-to-image, text-to-audio, and text-to-video generation.