Generative AI in Intelligent Tutoring Systems: Paradigm Shift and Hybrid Control

Abstract

The emergence of Generative AI (GenAI) is radically transforming the field of Intelligent Tutoring Systems (ITS), marking the transition from rigid symbolic architectures to dynamic content generation. This article analyzes this paradigm shift by comparing the limitations of traditional ITS with the opportunities offered by Large Language Models (LLMs). While these technologies overcome the constraints of content creation, they introduce major challenges such as hallucinations and the risk of social isolation. In response, we introduce a hybrid perspective through the IMTS architecture, which proposes to frame the power of GenAI with human supervision (“Human-in-the-loop”) to ensure learning that is fluid, ethical, and reliable.