The advent of Generative Artificial Intelligence (AI) in education presents an opportunity to rethink the way we learn and teach. This involves moving beyond simply making information more accessible and instead positioning students as active participants in a continuous dialogue with tools capable of generating knowledge dynamically and in a personalized way. Drawing on the theoretical foundations of Bakhtin's dialogism, Vygotsky's social constructivism, and Piaget's genetic epistemology, this paper explores the potential of Large Language Models (LLMs) to catalyze a pedagogical r-evolution.

Bakhtin's concept of dialogism emphasizes the social and interactive nature of language, where words acquire meaning through dialogue and personal appropriation. Vygotsky's Zone of Proximal Development (ZPD) highlights the importance of social interaction and scaffolding in learning, with LLMs serving as dynamic mediators within the ZPD. Piaget's genetic epistemology posits that understanding comes from actively reconstructing knowledge, a process LLMs can facilitate, particularly in scientific and mathematical disciplines.

This paper proposes two pedagogical models for leveraging the use of LLMs tested in primary school and college:
The Historical Genesis of Concepts: LLMs support orchestrating activities and stories that guide students to reconstruct the historical paths of scientific/mathematical discoveries, deepening abstract concept comprehension.

Dialogical Texts:
LLMs, such as through NotebookLM, transform passive text consumption into dynamic dialogues, enabling interaction with historical figures, texts, and diverse perspectives.

These models illustrate how LLMs can mediate and catalyze the internalization of social dialogue (Vygotsky), the reinvention of knowledge (Piaget), and the personal appropriation of meaning through dialogue (Bakhtin). The teacher's role evolves into that of an architect of dialogical experiences, guiding students to formulate deeper questions, co-create knowledge and promote critical thinking.

The integration of LLMs opens up new possibilities for personalized learning, catering to diverse learning styles and cognitive needs. LLMs can provide tailored support, acting as dynamic scaffolding within each student's ZPD, while facilitating the active reconstruction of knowledge. Examples are provided coming from exploratory studies carried out both at the elementary school and at the University College.

The paper also addresses the ethical and pedagogical challenges associated with LLMs, emphasizing the importance of critical reflection and meta-cognitive awareness. Drawing on examples from school districts like LAUSD, it highlights the role of reflection journals and discussions in fostering responsible and effective use of AI in education.

In conclusion, the integration of LLMs in education presents an opportunity for a profound transformation of pedagogical practices. By embracing a dialogical, mediated, and reconstructive approach, we can leverage LLMs to cultivate critical thinking, creativity, and meta-cognition, empowering students to become active co-creators of their knowledge and lifelong learners.

Keywords: Dialogical Thinking, Learning by invention, NotebookLM, Concepts genesis.