Impact of generative artificial intelligence-based virtual tutors on self-directed mathematics learning: a systematic review (#2252)

Authors

Portilla Peralta, Carlos

Velarde Allazo, Edwar

Abstract

The emergence of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) has reconfigured mathematics education, shifting the boundaries of traditional instruction toward dynamic conversational tutoring models. Objective: This study aimed to analyze the impact of GenAI-based virtual tutors on self-directed learning in mathematics by identifying pedagogical trends, benefits for student autonomy, and implementation challenges. Methodology: A systematic literature review of scientific publications between 2022 and 2026 was conducted following the PRISMA guidelines. Empirical and theoretical studies integrating tools such as ChatGPT and adaptive models in higher and secondary education contexts were analyzed. Results: Evidence indicates that GenAI integration significantly improves academic performance and the understanding of abstract concepts (such as functions and calculus) while reducing math anxiety and cognitive load. These tools act as scaffolding that fosters self-regulation through immediate and personalized feedback. However, risks associated with data accuracy ("hallucinations") and potential technological dependence limiting critical thinking were identified. Conclusions: It is concluded that virtual tutors are effective catalysts for personalized mathematical learning, provided their implementation is mediated by critical digital literacy strategies and instructional design that prioritizes human validation over automation.