AI and Multilingual Resources Meet in New Framework for Reflective EFL Writing

Two of the most talked-about resources in language education, generative artificial intelligence and the deliberate use of multilingual repertoires, share a hidden weakness: neither one, on its own, makes anyone think. That is the starting point of a new conceptual framework published in the journal Frontiers in Education, which argues that the metacognitive value of both resources lies not in their availability but in how learners monitor, evaluate, and act on the alternatives they generate. The article, a Perspective by Mario Alvarado-Mendoza and Alexis Contreras-Falcones of Universidad Espíritu Santo in Ecuador, does not report an intervention or a trial. Instead, it lays out a theoretical architecture for English as a Foreign Language (EFL) writing instruction, with particular attention to the linguistic, technological, and institutional conditions of Latin America.

The authors call their proposal the Integrated Intentional Translanguaging–GenAI Framework for Reflective Decision-Making. Its central claim is deliberately modest: intentional translanguaging, meaning the purposeful use of a multilingual writer’s languages, varieties, registers, and other semiotic resources for an identifiable writing decision, and generative AI, which can supply ideas, explanations, feedback, and alternative formulations, are complementary sources of possibilities. But possibilities are not reflection. The framework’s core mechanism is what the authors term learner evaluative judgment: the act of appraising alternatives against relevant criteria and deciding whether to accept, modify, reject, or further examine them, all while retaining authorship and regulatory responsibility for the text.

To structure that judgment, the authors propose an author-developed heuristic that unfolds across the writing process: identify a purpose, examine alternatives, compare, judge, decide, and reflect. They are careful to describe this sequence as a recursive synthesis rather than an established metacognitive model, and it is not meant to run in a fixed order. The heuristic can recur whenever a learner consults either resource, whether during planning, monitoring, revision and evaluation, or after writing. During planning, multilingual resources may help learners retrieve prior knowledge, brainstorm, or organize ideas, while generative AI may offer structures or lexical choices. During monitoring, cross-linguistic comparison can reveal differences in meaning, vocabulary, register, or rhetorical effect. During revision, learners weigh formulations against their intended meaning, audience, and purpose. After writing, they may consider whether the strategy was useful and whether it might serve future tasks, without presuming that transfer will occur.

A crucial distinction runs through the entire framework: translanguaging and AI are resources, not scaffolds. Scaffolding, in the authors’ treatment, is external, contingent, and adjustable pedagogical support, delivered through expectations, modeling, criteria, prompts, feedback, and gradually adjusted guidance. Teacher mediation sits outside the learner’s decision-making cycle, structuring the conditions under which evaluation happens, while the monitoring, evaluation, justification, and regulation themselves remain learner responsibilities. This separation matters because, as the authors note, purposeful resource use alone is not evidence of metacognitive regulation, and reported agency or intentionality does not demonstrate observable monitoring. A student who switches between Spanish and English, or who pastes an AI revision into a draft, may be engaging deeply or merely complying; the framework insists that only converging process evidence, such as learner explanations, planning artifacts, revision traces, or stimulated recall, can support claims about what a learner was actually doing.

The way learners interact with generative AI receives particular attention. The authors argue that how a request is framed shapes what becomes available for examination: prompts that specify a purpose, request alternatives, or seek clarification offer more opportunities for comparison than requests for direct text revision. Prompting, in this view, is a strategic process rather than an isolated technical skill, one through which learners consider why they are consulting AI and what kind of response they need. AI output is treated as material for evaluation, never as evidence of reflective thought, and this distinction bears directly on documented concerns about overdependence, reduced participation in planning and problem-solving, and accountability for authorship. Post-writing reflection can ask why AI was used, whether its assistance was useful, and whether a similar strategy might fit future tasks.

The Latin American orientation of the framework is contextual and prospective rather than empirically validated, and the authors are explicit about the conditions that could shape its adaptation. Latin American EFL classrooms should not be assumed to be uniformly Spanish-speaking or pedagogically homogeneous; research documents the coexistence of English, Spanish, Indigenous languages, and other repertoires, alongside tensions produced by monolingual practices. Intentional translanguaging, the authors stress, should not be reduced to permission to use a first language. Its use depends on learners’ actual repertoires, the institutional context, and the writing purpose. For Indigenous and other minoritized learners, decisions about whether to draw on locally meaningful repertoires also involve questions of linguistic legitimacy, the maintenance of learner voice, and responsibility for authorship, particularly where English-only expectations constrain recognition of those resources.

Technological and institutional conditions add further layers of variation. Disparities in digital access and literacy persist across Latin America, and unequal infrastructure and institutional resources may constrain equitable GenAI-supported practice. Institutional responses to AI and academic integrity remain uneven across the region. The framework therefore does not assume equivalent access to generative AI, equal capacity to evaluate its output, or shared policy boundaries. Where access is limited, the technology may simply be unavailable; where AI literacy is limited, learners may struggle to evaluate outputs critically. Research on GenAI in EFL education has highlighted potential uses alongside concerns about skill development, academic integrity, and learner autonomy, and the framework treats these concerns as design constraints rather than obstacles to be waved away.

Teacher mediation emerges as the pivot on which the whole proposal turns. Pedagogical translanguaging positions teachers as planners who deliberately activate multilingual resources, while regional teacher-education research highlights tensions among professional development, institutional expectations, and local sociolinguistic realities. Research on AI-supported EFL writing similarly emphasizes guidance in evaluating AI feedback while maintaining learner authorship and decision-making. Within the framework, teachers may establish purposes and boundaries, model comparison, provide evaluative criteria, and prompt justification, with the degree and form of support varying by learner proficiency, prior experience with generative AI, available resources, institutional policy, task complexity, and classroom linguistic composition. Notably, the authors insist that neither resource is required in every task: translanguaging, GenAI, both, or neither may be appropriate depending on the writing purpose. They also warn that excessive teacher direction can constrain learner agency, making the adjustment of support as important as its provision.

The framework is candid about its own boundary conditions. It does not assume that access to alternatives produces reflection or learning. Learners may accept AI output uncritically, compare alternatives superficially, or lack the linguistic, disciplinary, or AI literacy needed for accurate judgment. Teacher-mediated scaffolding may support evaluation but can become overly directive if criteria or decisions are supplied by the teacher. The authors position their conceptual contribution not in the combination of the two resources but in foregrounding how learners monitor, evaluate, and regulate the alternatives those resources produce, with the central mechanism being the learner’s evaluative decision to accept, modify, reject, or further examine an alternative while retaining responsibility for the text.

Because the proposal is conceptual, the authors outline a research agenda for empirical examination. Future studies could use process-sensitive approaches, analyzing learner prompts, AI interaction histories, multilingual planning artifacts, successive drafts, stimulated-recall interviews, and teacher feedback to identify observable monitoring, evaluative judgment, and regulatory decisions. Comparative research across proficiency levels, educational settings, and linguistic repertoires could test how the framework should be adapted, and studies in Latin American contexts should examine how technological access, institutional AI policies, teacher preparation, and language ideologies enable or constrain it. A supplementary title- and abstract-level mapping of Scopus and Web of Science informed the authors’ positioning, suggesting limited convergence among translanguaging, generative AI, metacognitive and self-regulatory processes, and scaffolding in the retrieved literature, though the authors note this observation informs rather than demonstrates the absence of related research. For minoritized learners, the stakes extend beyond technique: the framework connects metacognitive writing decisions with questions of authorship, linguistic legitimacy, and belonging, asking not only which version of a sentence works best, but whose languages and voices are allowed to count in academic writing at all.

Subject of Research: A conceptual framework integrating intentional translanguaging and generative AI for metacognitive EFL writing with implications for Latin America

Article Title: Intentional translanguaging and generative AI for metacognitive EFL writing: reflective scaffolding and implications for Latin America

Article References: Alvarado-Mendoza, M., & Contreras-Falcones, A. (2026). Intentional translanguaging and generative AI for metacognitive EFL writing: reflective scaffolding and implications for Latin America. Frontiers in Education, 11, Article 1971754. https://doi.org/10.3389/feduc.2026.1971754

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1971754

Keywords: translanguaging, generative AI, EFL writing, metacognition, evaluative judgment, teacher mediation, scaffolding, learner agency, Latin America, language education, authorship, self-regulation