On Large Language Models’ capacity to replace human teachers: a lesson from the Mahabharata

Abstract In the wake of recent developments in Large Language Model (LLM) Artificial Intelligence technologies, there has been a resurgence of scholarly conversation regarding the roles and value of human teachers in the educative process. In this article, we situate these scholarly discussions against the backdrop of an ancient tale of artificial intelligence from the Mahabharata. We begin by providing a basic explanation of how LLMs function and then highlight major trends in the literature that advocate for, or caution against, their implementation in education. Next, we explore the dangers of anthropomorphizing LLMs more generally before situating those dangers in an educational context. Finally, we outline the specific ways in which LLMs cannot fulfil the human teacher’s role. This is in large part because LLMs lack the capacity that human teachers have for empathetic understanding in relation to their students. Because LLMs function within the bounds of numerical algorithmic predictions that are based on patterns of textual data devoid of semantic meaning, we argue that LLMs as currently constituted cannot embody some of the most salient attributes of a human teacher.