Artificial Intelligence in the Teaching of University Mathematics: Between Pedagogical Innovation and the Risk of Superficial Learning

Abstract

The integration of Artificial Intelligence (AI) in university mathematics education presents a fundamental duality. On one hand, it offers transformative opportunities such as personalized learning, immediate feedback, and democratized access to tutoring, enabling significant pedagogical innovation. On the other hand, it carries substantial risks of fostering superficial learning, characterized by cognitive dependency, erosion of reasoning skills, and an illusion of understanding. This study, based on a systematic review of specialized literature (2018-2025), identifies a critical perceptual gap: while 68% of students value AI for its immediacy and accessibility, only 25% of faculty perceive its impact positively, expressing skepticism grounded in pedagogical risks. The effectiveness of AI critically depends on contextual factors; Intelligent Tutoring Systems (ITS) demonstrate positive pedagogical impact (40% reduction in failure rates), while the use of unsupervised generic chatbots is associated with higher risks. It is concluded that the determining factor is not the technology itself, but rather the instructional design and pedagogical supervision framing its use. Successful implementation requires adopting a "cognitive augmentation" model that prioritizes the thinking process over mere answer retrieval, ensuring that AI amplifies—rather than replaces—the development of autonomous, deep, and critical mathematical thinking.