Do People Learn Better with Large Language Models? Examining the Double Edged Sword and a Roadmap for Future Research

The rapid integration of Large Language Models (LLMs) into educational settings represents a transformative shift in how knowledge is accessed, constructed and mediated. This position paper examines the cognitive, pedagogical and ethical dimensions of learning with LLMs. Learners increasingly use LLMs to externalize cognitive tasks, potentially enhancing efficiency but raising concerns about critical thinking and reasoning. Emerging evidence suggests that LLMs can serve as cognitive partners and collaborative mediators, supporting distributed cognition, metacognitive scaffolding and interactive learning. The paper further discusses how learning styles and learning motivation shape usage of LLMs in educational context. Beyond cognitive and pedagogical implications, the paper also highlights ethical and societal consequences of using LLMs for learning. It concludes with a roadmap for future research, emphasizing the need to design these tools such that they benefit education and learning.