Year: 2024
Type: article
Abstract: This article explores the relationship between artificial intelligence and education, focusing on Large Language Models and the opportunities and challenges these tools present. After analysing the architectures that have enabled their success, such as the Transformer, the issues related to the transparency and reliability of the generated responses are examined, with particular attention to the phenomenon of model "hallucinations." Strategies to mitigate these challenges, such as Chain of Thought and Retrieval-Augmented Generation, are then illustrated, which improve the transparency and reliability of LLMs. The EduRag project represents a practical application of these strategies, demonstrating how AI can be effectively used in the educational field to support learning and promote the use of reliable sources. Finally, possible future scenarios are outlined in which AI will play a central role in education, with increasingly personalized, interactive, and multimodal tools capable of enhancing the educational experience through deeper integration with language, writing, and other forms of human interaction.
Source: Graphos
DOI: 10.4454/graphos.112
Authors: Dragoni, Daniele, Epure, Maria Désirée
Institutions: University of Macerata, Roma Tre University
Language: Italian
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W7111042999
