Sistema para tutoría y feedback personalizado en cursos universitarios basado en un multi agente interoperable basado en LLMs

Abstract

Abstract: Personalizing learning in higher education presents a significant challenge, particularly when providing individual feedback to hundreds or thousands of students. In this curricular integration project, we developed an intelligent tutoring system that combines multiple virtual agents and advanced artificial intelligence technology (LLMs) to address this problem. The system we designed has two complementary components: on one hand, a module that accurately answers student queries using course documents as reference, primarily based on RAG. On the other hand, we have a more autonomous module that can analyze each student's profile comprised of the teaching plan, their previous academic performance, and their individual characteristics and suggest a personalized study plan, all without requiring direct instructor intervention. To develop it, we used a step-by-step methodology, testing and improving at each stage, ultimately implementing several specialized agents that work coordinately through specific execution graphs.