Musica generativa e intelligenza artificiale

Errore e memoria nella classe di italiano L2 per adulti migranti

Abstract

This paper describes a teaching experience conducted at a CPIA (Provincial Centre for Adult Education) with adult migrant learners of Italian as a second language, starting from a structural challenge typical of these contexts: managing mixed-ability classes where learners bring highly heterogeneous linguistic, biographical, and educational backgrounds — including participants with low or no L1 literacy. In response to this complexity, the paper proposes the use of generative artificial intelligence to transform systematic classroom errors into purpose-built didactic songs, constructed from situations authentically experienced by the group. The theoretical framework integrates a reflection on the limitations of traditional corrective feedback for fossilized interlanguage structures (Selinker 1972; Krashen 1982; Monami 2021), neurocognitive research on the relationship between music, rhythm, and memory retention (Sun et al. 2024; Caon 2011), and an ethnodidactic approach that treats real learner errors as generative resources (Gaeta, Ianni 2025). The paper focuses on one lesson that served as the generative moment of the entire approach, showing how an authentic classroom episode can become the text of a song with the potential to lower the affective filter, support memorization, and simultaneously address different levels of competence. As a descriptive, exploratory account based on classroom observation rather than on systematic data collection, the paper discusses the potential and limitations of this approach, including the didacticization of AI-generated songs and differentiated exercises.