Language barriers could deepen as schools adopt AI without inclusion rules

European schools risk widening language barriers for migrant and refugee students unless artificial intelligence (AI) and digital learning tools are tied to teacher training, multilingual policy and equal access, according to a new analysis that frames AI not as a quick classroom upgrade, but as part of a wider test of whether European education systems can turn multilingual inclusion from policy language into daily practice.

Titled AI-Enhanced Digital Pedagogies and Multilingualism: Policy, Technology, and Inclusion in European Education and published in AI in Education, the study examines how digital learning environments, mobile-assisted language learning, AI and immersive tools can support language acquisition, intercultural dialogue and social inclusion in European classrooms, with Greece used as an example of how migration, language policy and uneven digital readiness now collide inside public education.

Digital tools can make multilingual policy practical, but only with trained teachers

Multilingual education has become a direct measure of democratic inclusion in Europe. As migration reshapes classrooms, students need support in the host-country language without losing the value of their mother tongues. According to the analysis, this balance not only affects language scores, but also shapes academic progress, confidence, cultural identity and access to public life.

The research is based on a qualitative review of contemporary scholarship, European policy documents and institutional reports. It brings together work on multilingual education, digital pedagogy, AI, migrant inclusion and the Greek education system. The authors examine the conditions under which AI can actually help schools deliver on multilingual commitments.

European frameworks have long treated linguistic diversity as part of social cohesion and democratic citizenship. Newer digital education strategies now seek to connect that goal with online platforms, mobile tools and AI-assisted learning. The study finds that this connection can work only when digital systems are built around pedagogy, not around technology for its own sake.

Digital learning environments can support multilingual classrooms by giving students access to language resources, collaborative platforms, flexible practice and personalized feedback. Learning management systems such as Moodle, Canvas and Google Classroom can host multilingual materials and support peer interaction. Mobile apps can extend practice beyond formal lessons, while AI-supported tools can adapt tasks to learner levels and provide feedback. Virtual and augmented-reality systems can also create simulated communication settings for language practice.

For migrant students, these tools may help address gaps created by interrupted schooling, unfamiliarity with the host-country language and limited access to traditional language support. Digital platforms can provide extra practice, reduce the fear of making mistakes in front of peers and allow learners to progress at different speeds. Mobile-assisted language learning is especially important because it can move support beyond the timetable of the classroom.

The study also highlights translanguaging, an approach that recognizes how multilingual students use their full language resources rather than separating languages into rigid categories. Digital environments can support this through audio, video, interactive dialogue, multilingual materials and collaborative tasks. For students with migrant backgrounds, that means the language they bring from home can become part of learning rather than something pushed outside the school gate.

Notably, the authors warn that digital inclusion can fail if tools are poorly designed or narrowly available. A platform built mainly for dominant languages may deepen linguistic inequality and an AI system trained mostly on widely used languages may give weaker support to students who speak less-represented languages. A digital classroom without trained teachers may produce confusion rather than inclusion.

Digital multilingual education works only when it is connected to curriculum design, teacher preparation, data protection, accessibility and institutional accountability. Without those conditions, AI and digital platforms may simply give schools a more modern way to reproduce old inequalities.

Greece shows why European education goals need local capacity

Greece has faced sustained migration while also operating within European multilingual and digital education frameworks. The country has developed reception classes, intercultural schools and participation in European programs such as Erasmus+, eTwinning and CLIL-related initiatives. These efforts show that the country has formal mechanisms for inclusion, but the paper finds that implementation remains uneven.

The key problem is not the absence of policy language, but the gap between policy ambition and classroom capacity. The authors identify weaknesses in digital infrastructure, teacher training and administrative coordination. In some cases, multilingualism remains more visible as an official goal than as a consistent school practice. Migrant students cannot benefit from European-level commitments if local schools lack the tools, training and support to apply them.

One major barrier is the digital divide. Students from vulnerable households may lack reliable devices, stable internet access or suitable learning spaces. Regional differences can also affect whether schools are ready to use digital tools effectively. Without targeted investment, digital multilingual education may help students who already have better resources while leaving the most vulnerable further behind.

Teacher preparation is another obstacle. Educators need both digital competence and intercultural teaching skills. Teachers must know how to use AI tools, mobile applications and online platforms in ways that support language development and social inclusion. They also need institutional backing, because digital reform can become an added burden if it is introduced without time, training and practical support.

The paper also sheds light on the limitations of short-term or project-based innovation. European initiatives can increase intercultural awareness, improve linguistic exposure and produce useful digital materials. However, their long-term effect depends on whether they are absorbed into national curricula, teacher training systems and evaluation frameworks. A successful pilot does not become policy unless institutions are built to sustain it.

For Greece, the authors propose a more structured route. Reception Classes and Intercultural Schools could become pilot settings for AI-assisted language scaffolding, adaptive feedback and multilingual digital platforms. Such pilots would need coordination among the Ministry of Education, regional authorities and teacher-training bodies. They would also need clear indicators to track teacher participation, student access, platform use and inclusion outcomes.

The findings also have a wider relevance. European digital and multilingual policies cannot be transferred as ready-made solutions. They must be adapted to local migration patterns, school resources, language profiles and institutional capacity. The study shows that technology can support integration, but only when governments treat implementation as a public responsibility, not a school-by-school experiment.

AI in multilingual classrooms needs rules, oversight and fair language representation

The paper specifically focuses on AI, arguing that the technology should not be treated as just another classroom tool. In multilingual education, AI systems can shape access to knowledge, translation, feedback, assessment and learning pathways. This is what gives AI real educational power, but it also raises risks that ordinary software does not carry.

Generative AI and multilingual large language models can help produce explanations, translations, language exercises and adaptive learning materials. They can support differentiated instruction and give students more chances to practice. For teachers, AI may help prepare multilingual content or support learners working at different levels. Used carefully, these tools could make language support more responsive and less dependent on one-size-fits-all instruction.

The risk is that AI systems are not linguistically neutral. Their quality depends on the data used to train them. Dominant languages are often better represented in digital datasets, while minority, migrant and less widely spoken languages may be underrepresented. In a multilingual classroom, this can lead to weaker translation, uneven feedback and limited content quality for students whose languages are already marginalized.

The authors also caution about explainability and teacher oversight. If an AI system gives automated feedback without clear reasoning, students may not understand how to improve. Teachers may also struggle to judge whether the feedback is accurate, culturally sensitive or useful. Therefore, AI should support teachers, not displace their judgment.

Data governance is another major issue. AI tools used in schools may process sensitive information about students, including language ability, learning behavior and personal background. Migrant and refugee students may face added risks if data systems are not properly regulated. The study says educational AI must follow strong data protection standards and clear institutional protocols, especially when tools influence assessment, progression or access to learning opportunities.

The paper argues that AI used for multilingual inclusion should be transparent, accountable and aligned with democratic safeguards. Schools and governments must ensure that AI tools do not reproduce bias, marginalize smaller languages or weaken teacher authority.

To address these challenges, the authors propose a four-part model for multilingual digital education:

1. The political phase requires formal recognition of multilingualism and national digital platforms for language education.
2. The pedagogical phase calls for multimodal teaching, intercultural competence and virtual learning environments.
3. The technological phase focuses on accessible AI tools, open multilingual datasets and inclusive design.
4. The social phase emphasizes participation, digital learning communities, equality and linguistic tolerance.

Policy shapes classroom practice, pedagogy guides technology use, technology affects inclusion and social outcomes should inform future policy. Multilingual digital education will succeed only if governance, teaching, infrastructure and social inclusion move together, the study asserts.