Detalle de: ENGRAMMER: New Solutions to Leverage AI in the Classroom

Artificial intelligence (AI), and in particular large language models (LLMs), is having a significant impact on education, and not always in a positive way: the use of systems such as ChatGPT by pupils from a young age can reduce their intellectual effort and encourage excessive dependence, including psychological dependence. Furthermore, in certain situations, these systems may provide inappropriate answers, introduce bias and raise privacy concerns, which are particularly sensitive when used by minors. In this article, we present the ideas we are promoting through a platform (ENGRAMMER) to encourage better use of AI in education. To this end, we offer basic training in the subject both for teachers, through a specific micro-credential, and for pupils, through a series of informative talks in schools. These talks propose the idea that pupils should actively use AI assistants in their learning, following the so-called Feynman method. In parallel, using RAG (Retrieval Augmented Generation) technique on a graph database, and an open-source LLM model, we have developed an AI assistant “APRENDE” (“LEARN”) that allows teachers to define their content, and guarantees student privacy by using only local or institutional resources. This AI assistant can be accessed via mobile devices (tablets, smartphones), and its usefulness in educational activities has been tested both in the classroom and during museum visits. We are currently integrating this system with the Moodle platform using MCP and OpenClaw, so that content management and activity monitoring can be carried out directly by the teacher. We hope to make this platform available to secondary school teachers in collaboration with the education authorities in our region, starting in the next academic year.