The doctoral thesis focuses on the use of artificial intelligence, and in particular Large Language Models (LLMs), in the teaching of Italian as a second language (L2) to adult migrants. The phenomenon under investigation concerns the integration of generative artificial intelligence tools into language teaching contexts, with the aim of exploring their didactic potential, limitations, and methodological implications (Sok S., Heng K., 2023). In recent years, the development of advanced language models capable of generating coherent texts, sustaining dialogic interactions, and adapting to communicative contexts has opened up new perspectives for language education. These technologies are particularly relevant in Italian L2 learning contexts for adult migrants, which are characterized by strong heterogeneity in learners’ linguistic, cultural, and biographical profiles, as well as by immediate communicative needs related to social and professional integration (Cinganotto L., 2024). The thesis therefore examines how artificial intelligence can be employed as a supportive tool for language teaching, not as a substitute for the teacher, but rather as a mediating resource capable of fostering personalization, interaction, and learner autonomy in language learning. The general objective of the research is to investigate how artificial intelligence, and Large Language Models in particular, can contribute to innovation in the teaching of Italian L2 to adult migrants, while respecting glottodidactic principles and learners’ needs (Chen Y., Jensen S., 2023; Ruggiano F., 2021). The specific objectives of the research are: • to analyze the characteristics of Large Language Models that are relevant to language teaching; • to examine the role of artificial intelligence in educational systems, with particular reference to adult language education; • to relate the theoretical principles of Italian L2 didactics to the potential offered by AI; • to design and describe examples of teaching units that integrate the use of artificial intelligence in a conscious and pedagogically grounded manner. Based on these objectives, the research is structured around the following research questions: 1. Which characteristics of Large Language Models are most relevant to the teaching of Italian as a second language? 2. How can artificial intelligence support personalized language learning processes and teaching for adult migrants? 3. What opportunities and critical issues emerge from the analysis of AI use in Italian L2 teaching contexts? 4. How can activities and teaching units be effectively designed to integrate artificial intelligence tools? The thesis is structured into four chapters: the first is devoted to Large Language Models, the second to artificial intelligence in educational systems, the third to the principles of Italian L2 didactics in relation to AI, and the fourth to the proposal of teaching units based on the use of artificial intelligence. The research adopts a predominantly qualitative approach, based on a critical analysis of existing literature and on instructional design. In the initial phase, a systematic review of scientific studies related to Large Language Models, artificial intelligence in educational systems, and Italian L2 didactics was conducted. Subsequently, the theoretical analysis was complemented by a didactic design phase aimed at creating examples of teaching units integrating the use of artificial intelligence. The activities were designed taking into account glottodidactic principles, the characteristics of adult migrant learners, and the specific functionalities of AI tools. The choice of a qualitative approach is consistent with the exploratory aims of the research, which seeks to understand the potential and limitations of AI in complex educational contexts rather than to measure quantitative outcomes. The results of the research highlight how artificial intelligence can represent a significant resource for Italian L2 teaching, provided that it is integrated in a conscious and pedagogically grounded manner. In particular, the potential of Large Language Models emerges in fostering linguistic interaction, personalization of learning pathways, and the development of learner autonomy (Cinganotto L., Montanucci G., 2024). At the same time, the analysis highlights several critical issues, including the risk of reproducing linguistic and cultural biases, the need for constant teacher mediation, and the importance of specific teacher training in the use of artificial intelligence tools (Rasul T., Nair S. et al., 2023). Overall, the thesis proposes a vision of AI as a supportive tool for teaching, capable of enriching Italian L2 instruction, but not of replacing the relational, cultural, and pedagogical dimensions that characterize language education for adult migrants. The study is situated within an interdisciplinary field that connects artificial intelligence, language education, and second language pedagogy, contributing to the expansion and deepening of a research area that is still in the process of consolidation, particularly in the Italian context. A first significant contribution concerns the application of Large Language Models to the teaching of Italian as a second language—an area that, compared to other more extensively studied languages, especially English, remains underexplored in the scientific literature. The thesis aims to contribute to research by offering a theoretical and methodological analysis specifically oriented toward the teaching of Italian L2 to adult migrants. A second contribution lies in bringing together research on Large Language Models (Brown, Mann et al., 2020; Ray, 2023) and reflections on glottodidactic principles (Balboni, 2015; Vedovelli, 2014), moving beyond a purely instrumental view of artificial intelligence. The research instead proposes a critical approach that emphasizes the role of didactic mediation and the professional competence of the teacher. Furthermore, the study contributes to the literature by proposing instructional design models and examples of teaching units that concretely demonstrate how artificial intelligence can be integrated into language learning pathways. These proposals constitute an original contribution, as they translate theoretical reflections into practical applications that are replicable and adaptable to educational contexts. Finally, the thesis enriches the scientific debate by highlighting critical issues and ethical implications related to the use of AI in language education, contributing to a more balanced and informed reflection on technological innovation in education. From a theoretical perspective, the research shows how generative language models can be used as tools of linguistic mediation capable of supporting communicative approaches, fostering interaction, and sustaining personalized learning processes. However, the analysis also highlights that the effectiveness of artificial intelligence in educational contexts strongly depends on its integration within a conscious and critically grounded glottodidactic framework (Ruggiano F., 2018). From a practical perspective, the thesis demonstrates how AI can be used to design innovative teaching activities that address the immediate linguistic needs of adult migrants, promoting autonomy, motivation, and language use. At the same time, the need for an active teacher presence clearly emerges, with teachers playing a crucial role in guiding, monitoring, and contextualizing the use of artificial intelligence tools (Jeon J., Lee S., 2023). Overall, the research supports a vision of artificial intelligence not as an automatic solution to educational challenges, but as a pedagogical resource that, when used critically and responsibly, can enrich Italian L2 teaching and contribute to innovation in educational contexts (Hernández-Sellésa N., Muñoz-Carril P. C., 2019). Despite its contributions, the research presents certain limitations. First, the study is largely based on theoretical analysis and instructional design, without including large-scale empirical experimentation. This limits the generalizability of the findings and highlights the need for further research based on empirical data collected in real teaching contexts. Second, the rapid evolution of artificial intelligence technologies means that some reflections are inevitably linked to the current state of the art, which may change in the short term. Moreover, the analysis focuses primarily on text-based Large Language Models, leaving aside the potential of multimodal AI (text, speech, images), which could be particularly relevant for adult migrant language learning. In light of these limitations, future research directions may include: • empirical studies on the effectiveness of AI use in Italian L2 classes for adult migrants; • comparative analyses between traditional teaching approaches and AImediated approaches; • further investigation into the role of multimodal AI in language education; • research on the digital and pedagogical competences required of teachers for the informed use of artificial intelligence. These future perspectives may contribute to consolidating and expanding the theoretical and applied framework outlined in the present thesis, fostering an increasingly informed and responsible use of artificial intelligence in second language teaching.
Il lavoro di tesi di dottorato ha come oggetto di studio l’impiego dell’intelligenza artificiale, e in particolare dei Large Language Models (LLM), nella didattica dell’italiano come lingua seconda (L2) rivolta ad adulti migranti. Il fenomeno analizzato riguarda l’integrazione di strumenti di intelligenza artificiale generativa nei contesti di insegnamento linguistico, con l’obiettivo di esplorarne le potenzialità didattiche, i limiti e le implicazioni metodologiche (Sok S., Heng K. 2023). Negli ultimi anni, lo sviluppo di modelli linguistici avanzati in grado di generare testi coerenti, sostenere interazioni dialogiche e adattarsi al contesto comunicativo ha aperto nuove prospettive per l’educazione linguistica. Tali tecnologie risultano particolarmente rilevanti nei contesti di apprendimento dell’italiano L2 per adulti migranti, caratterizzati da forte eterogeneità dei profili linguistici, culturali e biografici degli apprendenti, nonché da bisogni comunicativi immediati legati all’integrazione sociale e lavorativa (Cinganotto L. 2024). La tesi analizza dunque il modo in cui l’intelligenza artificiale possa essere impiegata come strumento di supporto alla didattica, non in sostituzione del docente, ma come mezzo di mediazione capace di favorire personalizzazione, interazione e autonomia nell’apprendimento linguistico. L’obiettivo generale della ricerca è indagare in che modo l’intelligenza artificiale, e in particolare i Large Language Models, possano contribuire all’innovazione della didattica dell’italiano L2 per adulti migranti, nel rispetto dei principi glottodidattici e delle esigenze degli studenti (Chen Y., Jensen S. 2023), (Ruggiano F. 2021). Gli obiettivi specifici della ricerca sono: - analizzare le caratteristiche dei Large Language Models rilevanti per l’insegnamento linguistico; - esaminare il ruolo dell’intelligenza artificiale nei sistemi educativi, con particolare riferimento all’educazione linguistica degli adulti; - mettere in relazione i principi teorici della didattica dell’italiano L2 con le potenzialità offerte dall’IA; - progettare e descrivere esempi di unità didattiche che integrino l’uso dell’intelligenza artificiale in modo consapevole e pedagogicamente fondato. A partire da tali obiettivi, la ricerca si articola attorno alle seguenti domande di ricerca: 1. Quali caratteristiche dei Large Language Models risultano maggiormente rilevanti per la didattica dell’italiano come lingua seconda? 2. In che modo l’intelligenza artificiale può supportare i processi di apprendimento linguistico personalizzati e la didattica per adulti migranti? 3. Quali opportunità e criticità emergono dall’analisi dell’uso dell’IA nei contesti di insegnamento dell’italiano L2? 4. Come è possibile progettare attività e unità didattiche che integrino efficacemente strumenti di intelligenza artificiale? La tesi è strutturata in quattro capitoli: il primo dedicato ai Large Language Models, il secondo all’intelligenza artificiale nei sistemi educativi, il terzo ai principi della didattica dell’italiano L2 in relazione all’IA ed il quarto alla proposta di unità didattiche basate sull’uso dell’intelligenza artificiale. La ricerca adotta un approccio prevalentemente qualitativo, basato sull’analisi critica della letteratura esistente e sulla progettazione didattica. In una prima fase è stata condotta una revisione sistematica di studi scientifici relativi ai Large Language Models, all’intelligenza artificiale nei sistemi educativi e alla didattica dell’italiano L2. Successivamente, l’analisi teorica è stata affiancata da una fase di progettazione didattica, finalizzata alla creazione di esempi di unità didattiche che integrano l’uso dell’intelligenza artificiale. Le attività sono state progettate tenendo conto dei principi della glottodidattica, delle caratteristiche degli apprendenti adulti migranti e delle funzionalità specifiche degli strumenti di IA. La scelta di un approccio qualitativo risulta coerente con gli obiettivi esplorativi della ricerca, che mira a comprendere potenzialità e limiti dell’IA in contesti educativi complessi piuttosto che a misurare risultati quantitativi. I risultati della ricerca evidenziano come l’intelligenza artificiale possa rappresentare una risorsa significativa per la didattica dell’italiano L2, a condizione che venga integrata in modo consapevole e pedagogicamente fondato. In particolare, emerge il potenziale dei Large Language Models nel favorire l’interazione linguistica, la personalizzazione dei percorsi di apprendimento e lo sviluppo dell’autonomia degli studenti (Cinganotto L., Montanucci G., 2024). Allo stesso tempo, l’analisi mette in luce alcune criticità, tra cui il rischio di riproduzione di bias linguistici e culturali, la necessità di una mediazione costante del docente e l’importanza di una formazione specifica degli insegnanti all’uso degli strumenti di intelligenza artificiale (Rasul T., Nair S. et al. 2023). Nel complesso, la tesi propone una visione dell’IA come strumento di supporto alla didattica, capace di arricchire l’insegnamento dell’italiano L2, ma non di sostituire la dimensione relazionale, culturale e pedagogica che caratterizza l’educazione linguistica rivolta ad adulti migranti. Il lavoro si inserisce nel filone di studi interdisciplinari che mettono in relazione intelligenza artificiale, educazione linguistica e didattica delle lingue seconde, contribuendo ad ampliare e approfondire un ambito di ricerca ancora in fase di consolidamento, in particolare nel contesto italiano. Un primo contributo rilevante riguarda l’applicazione dei Large Language Models alla didattica dell’italiano come lingua seconda, ambito che, rispetto ad altre lingue maggiormente studiate, in particolare l’inglese, risulta ancora poco esplorato nella letteratura scientifica. La tesi vuole contribuire alla ricerca offrendo un’analisi teorica e metodologica specificamente orientata all’insegnamento dell’italiano L2 ad adulti migranti. Un secondo contributo consiste nel mettere in relazione lo studio sui Large Language Models (Brown, Mann et al., 2020; Ray, 2023) e la riflessione sui principi glottodidattici (Balboni, 2015; Vedovelli, 2014), superando una visione puramente strumentale dell’intelligenza artificiale. La ricerca propone infatti un approccio critico che mette in evidenza il ruolo della mediazione didattica e della competenza professionale del docente. Inoltre, il lavoro contribuisce alla letteratura proponendo modelli di progettazione didattica ed esempi di unità didattiche che mostrano concretamente come l’intelligenza artificiale possa essere integrata nei percorsi di apprendimento linguistico. Tali proposte rappresentano un contributo originale, in quanto traducono le riflessioni teoriche in applicazioni pratiche replicabili e adattabili al contesto educativo. Infine, la tesi arricchisce il dibattito scientifico evidenziando criticità e implicazioni etiche legate all’uso dell’IA nella didattica linguistica, contribuendo a una riflessione più equilibrata e consapevole sull’innovazione tecnologica in ambito educativo. Dal punto di vista teorico, la ricerca mostra come i modelli linguistici generativi possano essere utilizzati come strumenti di mediazione linguistica capaci di supportare approcci comunicativi, favorire l’interazione e sostenere processi di apprendimento personalizzati. L’analisi evidenzia tuttavia che l’efficacia dell’intelligenza artificiale in ambito didattico dipende fortemente dal suo inserimento all’interno di un quadro glottodidattico consapevole e criticamente fondato (Ruggiano F., 2018). Dal punto di vista pratico, la tesi dimostra come l’IA possa essere utilizzata per progettare attività didattiche innovative, in grado di rispondere ai bisogni linguistici immediati degli adulti migranti, promuovendo l’autonomia, la motivazione e l’uso della lingua. Al contempo, emerge con chiarezza la necessità di una presenza attiva del docente, chiamato a guidare, monitorare e contestualizzare l’uso degli strumenti di intelligenza artificiale (Jeon J., Lee S., 2023). Nel complesso, la ricerca sostiene una visione dell’intelligenza artificiale non come soluzione automatica ai problemi educativi, ma come risorsa pedagogica che, se utilizzata in modo critico e responsabile, può arricchire l’insegnamento dell’italiano L2 e contribuire all’innovazione dei contesti educativi (HernándezSellésa N., Muñoz-Carril P. C. 2019). Nonostante i contributi offerti, la ricerca presenta alcune limitazioni. In primo luogo, il lavoro si basa prevalentemente su un’analisi teorica e progettuale, senza includere una sperimentazione empirica su larga scala. Ciò limita la possibilità di generalizzare i risultati e rende necessario un ulteriore approfondimento basato su dati empirici raccolti in contesti reali di insegnamento. In secondo luogo, la rapida evoluzione delle tecnologie di intelligenza artificiale rende alcune riflessioni inevitabilmente legate allo stato dell’arte attuale, che potrebbe mutare nel breve periodo. Inoltre, l’analisi si concentra principalmente sui Large Language Models testuali, lasciando in secondo piano le potenzialità dell’IA multimodale (testo, voce, immagini), che potrebbero risultare particolarmente rilevanti per l’apprendimento linguistico degli adulti migranti. Alla luce di tali limiti, future linee di ricerca potrebbero includere: - studi empirici sull’efficacia dell’uso dell’IA in classi di italiano L2 per adulti migranti; - analisi comparative tra approcci didattici tradizionali e approcci mediati dall’intelligenza artificiale; - approfondimenti sul ruolo dell’IA multimodale nell’educazione linguistica; - ricerche sulle competenze digitali e pedagogiche necessarie ai docenti per un uso consapevole dell’intelligenza artificiale. Queste prospettive future potranno contribuire a consolidare e ampliare il quadro teorico e applicativo delineato dalla presente tesi, favorendo uno sviluppo sempre più consapevole dell’uso dell’intelligenza artificiale nella didattica delle lingue seconde.
Insegnare la lingua italiana L2 ad adulti con l’intelligenza artificiale. Esperienze in corso e prospettive future
Morano, Raffaella
2026
Abstract
The doctoral thesis focuses on the use of artificial intelligence, and in particular Large Language Models (LLMs), in the teaching of Italian as a second language (L2) to adult migrants. The phenomenon under investigation concerns the integration of generative artificial intelligence tools into language teaching contexts, with the aim of exploring their didactic potential, limitations, and methodological implications (Sok S., Heng K., 2023). In recent years, the development of advanced language models capable of generating coherent texts, sustaining dialogic interactions, and adapting to communicative contexts has opened up new perspectives for language education. These technologies are particularly relevant in Italian L2 learning contexts for adult migrants, which are characterized by strong heterogeneity in learners’ linguistic, cultural, and biographical profiles, as well as by immediate communicative needs related to social and professional integration (Cinganotto L., 2024). The thesis therefore examines how artificial intelligence can be employed as a supportive tool for language teaching, not as a substitute for the teacher, but rather as a mediating resource capable of fostering personalization, interaction, and learner autonomy in language learning. The general objective of the research is to investigate how artificial intelligence, and Large Language Models in particular, can contribute to innovation in the teaching of Italian L2 to adult migrants, while respecting glottodidactic principles and learners’ needs (Chen Y., Jensen S., 2023; Ruggiano F., 2021). The specific objectives of the research are: • to analyze the characteristics of Large Language Models that are relevant to language teaching; • to examine the role of artificial intelligence in educational systems, with particular reference to adult language education; • to relate the theoretical principles of Italian L2 didactics to the potential offered by AI; • to design and describe examples of teaching units that integrate the use of artificial intelligence in a conscious and pedagogically grounded manner. Based on these objectives, the research is structured around the following research questions: 1. Which characteristics of Large Language Models are most relevant to the teaching of Italian as a second language? 2. How can artificial intelligence support personalized language learning processes and teaching for adult migrants? 3. What opportunities and critical issues emerge from the analysis of AI use in Italian L2 teaching contexts? 4. How can activities and teaching units be effectively designed to integrate artificial intelligence tools? The thesis is structured into four chapters: the first is devoted to Large Language Models, the second to artificial intelligence in educational systems, the third to the principles of Italian L2 didactics in relation to AI, and the fourth to the proposal of teaching units based on the use of artificial intelligence. The research adopts a predominantly qualitative approach, based on a critical analysis of existing literature and on instructional design. In the initial phase, a systematic review of scientific studies related to Large Language Models, artificial intelligence in educational systems, and Italian L2 didactics was conducted. Subsequently, the theoretical analysis was complemented by a didactic design phase aimed at creating examples of teaching units integrating the use of artificial intelligence. The activities were designed taking into account glottodidactic principles, the characteristics of adult migrant learners, and the specific functionalities of AI tools. The choice of a qualitative approach is consistent with the exploratory aims of the research, which seeks to understand the potential and limitations of AI in complex educational contexts rather than to measure quantitative outcomes. The results of the research highlight how artificial intelligence can represent a significant resource for Italian L2 teaching, provided that it is integrated in a conscious and pedagogically grounded manner. In particular, the potential of Large Language Models emerges in fostering linguistic interaction, personalization of learning pathways, and the development of learner autonomy (Cinganotto L., Montanucci G., 2024). At the same time, the analysis highlights several critical issues, including the risk of reproducing linguistic and cultural biases, the need for constant teacher mediation, and the importance of specific teacher training in the use of artificial intelligence tools (Rasul T., Nair S. et al., 2023). Overall, the thesis proposes a vision of AI as a supportive tool for teaching, capable of enriching Italian L2 instruction, but not of replacing the relational, cultural, and pedagogical dimensions that characterize language education for adult migrants. The study is situated within an interdisciplinary field that connects artificial intelligence, language education, and second language pedagogy, contributing to the expansion and deepening of a research area that is still in the process of consolidation, particularly in the Italian context. A first significant contribution concerns the application of Large Language Models to the teaching of Italian as a second language—an area that, compared to other more extensively studied languages, especially English, remains underexplored in the scientific literature. The thesis aims to contribute to research by offering a theoretical and methodological analysis specifically oriented toward the teaching of Italian L2 to adult migrants. A second contribution lies in bringing together research on Large Language Models (Brown, Mann et al., 2020; Ray, 2023) and reflections on glottodidactic principles (Balboni, 2015; Vedovelli, 2014), moving beyond a purely instrumental view of artificial intelligence. The research instead proposes a critical approach that emphasizes the role of didactic mediation and the professional competence of the teacher. Furthermore, the study contributes to the literature by proposing instructional design models and examples of teaching units that concretely demonstrate how artificial intelligence can be integrated into language learning pathways. These proposals constitute an original contribution, as they translate theoretical reflections into practical applications that are replicable and adaptable to educational contexts. Finally, the thesis enriches the scientific debate by highlighting critical issues and ethical implications related to the use of AI in language education, contributing to a more balanced and informed reflection on technological innovation in education. From a theoretical perspective, the research shows how generative language models can be used as tools of linguistic mediation capable of supporting communicative approaches, fostering interaction, and sustaining personalized learning processes. However, the analysis also highlights that the effectiveness of artificial intelligence in educational contexts strongly depends on its integration within a conscious and critically grounded glottodidactic framework (Ruggiano F., 2018). From a practical perspective, the thesis demonstrates how AI can be used to design innovative teaching activities that address the immediate linguistic needs of adult migrants, promoting autonomy, motivation, and language use. At the same time, the need for an active teacher presence clearly emerges, with teachers playing a crucial role in guiding, monitoring, and contextualizing the use of artificial intelligence tools (Jeon J., Lee S., 2023). Overall, the research supports a vision of artificial intelligence not as an automatic solution to educational challenges, but as a pedagogical resource that, when used critically and responsibly, can enrich Italian L2 teaching and contribute to innovation in educational contexts (Hernández-Sellésa N., Muñoz-Carril P. C., 2019). Despite its contributions, the research presents certain limitations. First, the study is largely based on theoretical analysis and instructional design, without including large-scale empirical experimentation. This limits the generalizability of the findings and highlights the need for further research based on empirical data collected in real teaching contexts. Second, the rapid evolution of artificial intelligence technologies means that some reflections are inevitably linked to the current state of the art, which may change in the short term. Moreover, the analysis focuses primarily on text-based Large Language Models, leaving aside the potential of multimodal AI (text, speech, images), which could be particularly relevant for adult migrant language learning. In light of these limitations, future research directions may include: • empirical studies on the effectiveness of AI use in Italian L2 classes for adult migrants; • comparative analyses between traditional teaching approaches and AImediated approaches; • further investigation into the role of multimodal AI in language education; • research on the digital and pedagogical competences required of teachers for the informed use of artificial intelligence. These future perspectives may contribute to consolidating and expanding the theoretical and applied framework outlined in the present thesis, fostering an increasingly informed and responsible use of artificial intelligence in second language teaching.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_Dott_Morano_Raffaella.pdf
accesso aperto
Licenza:
Creative Commons
Dimensione
17.26 MB
Formato
Adobe PDF
|
17.26 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/375826
URN:NBN:IT:UNISTRADA-375826