The Correcting Default: How AI’s Convenience May Narrow the Multilingual Educator’s Skill

Nurdan Kavaklı Ulutaş, Izmir Demokrasi University, Türkiye

A brief moment

A teacher pastes a bilingual student’s paragraph into a writing tool and asks for feedback. The tool returns a more polished version. Yet, the rhythm of the home language has diminished. The teacher is tired and behind schedule. So, that more polished version is approved. This brief moment is echoed in a term that is elaborated below.

The correcting default

Generative AI has entered multilingual classrooms as a supportive tool. It generates content, assesses writing, develops materials, and also assigns grades. Teachers value this, but the concern remains beyond the typical discussions about academic dishonesty or accuracy alone. The most used models are trained on standard, dominant-language data, leading them to adjust instinctively towards one prestigious variety. Call it “the correcting default”. When a teacher acknowledges the default output without any further adjustment, the classroom tends to lean towards a monolingual standard, even in an environment intended for multiple languages.

Why it matters for multilingual education

Multilingual teaching relies on a teacher’s capacity to understand a learner’s whole repertoire. A skilled bilingual teacher recognizes a home-language influence and views it as a foundation for growth. The same feature, processed through a tool, returns flagged as an error to eliminate. On the other hand, human reading is a professional skill. It develops over time, and it defies classification within a framework. If a teacher’s daily work changes from making these readings to approving machine recommendations, the skill that validates learners’ repertoires diminishes. The classroom carries its multilingual label but loses its multilingual practice.

Why now

Lately, generative tools have become inexpensive and readily accessible. Institutions are currently beginning to support them to alleviate workload, and financial constraints compel teachers to implement whatever saves time. In this regard, the easiest option is to rely on the default output. No one chooses to narrow the classroom, but the narrowing accumulates through many small acceptances, one paragraph at a time.

Perceiving this as deskilling

Braverman (1974) traced how work is degraded once conception is split from execution. Managers are responsible for planning while workers keep the narrow tasks. A version of this division can settle into teaching. The tool holds the working model of ‘good’ language. The teacher performs review and delivery. Over a few terms, the teacher may lose fluency in the very judgement the tool now performs. This is the process of deskilling, functioning at the systemic level. A careful teacher using a convenient tool can be deskilled without ever noticing the loss.

Recent guidance in the field warns teachers about over-reliance and provides ways to maintain skill proficiency (Kohnke & Moorhouse, 2026). That work evaluates the risk within the realm of cognition and competence. Adding a linguistic axis to this, it is arguable what diminishes is the teacher’s capacity to consider a learner’s full repertoire.

The tension with humanizing pedagogy

Humanizing pedagogy encourages teachers to view learners as the authors of their own meaning (Bartolomé, 1994; Salazar, 2013). Learners’ languages carry that authorship. A tool that rewrites a multilingual voice into a standard monolingual one pulls against this aim, however helpful it feels at the time. The problem is narrow. It lives in the default the tool carries, and it can be worked with once teachers see it.

Implications

The cost is borne on more than one front. Learners may find their heritage and home varieties treated as blemishes to be smoothed away. Teachers may find their diagnostic skill and their professional standing quietly diminishing. A program can then run a widening gap between what it says about multilingualism and what its daily practice rewards. None of this is fixed in advance. It follows from letting the default stand.

What’s left to teachers and researchers

Teachers can keep the reading in human hands. They can use AI for lower-stakes drafting while holding on to their own judgement about a learner’s repertoire, and they can instruct a tool to preserve a learner’s variety instead of correcting it. These changes are small, and we still do not know how effectively they withstand actual workload pressure.

The research questions stem from that gap. Which teaching judgements are most highlighted when AI enters the feedback loop in multilingual environments? In what ways do teachers describe the shift from author to approver? Which design tools assist teachers in maintaining their repertoire-sustaining skills? How can institutions formulate AI guidelines that protect multilingual practices? Since the default is evident, it is essential to unravel these questions.

A final remark

The solution is not refusing the use of these tools. But the challenge is to recognise what their convenience directs us to surrender. In multilingual education, it is the teacher’s attentive ear for a learner’s full voice, and it is what we can least afford to lose.

References

Bartolomé, L. I. (1994). Beyond the methods fetish: Toward a humanizing pedagogy. Harvard Educational Review, 64(2), 173-194.

Braverman, H. (1974). Labor and monopoly capital: The degradation of work in the twentieth century. Monthly Review Press.

Kavaklı Ulutaş, N., & Kikutadze, V. (2026). Rethinking language education after crisis: AI-driven learning analytics as a site of social science inquiry. European Scientific Journal (ESJ), 20(22), 98-118. https://doi.org/10.19044/esj.2026.v22n20p98

Kohnke, L., & Moorhouse, B. L. (2026). Avoiding de-skilling and dependency: A practical guide to using generative AI in TESOL teacher education. TESOL Quarterly. https://doi.org/10.1002/tesq.70203

Salazar, M. del C. (2013). A humanizing pedagogy: Reinventing the principles and practice of education as a journey toward liberation. Review of Research in Education, 37(1), 121-148.

Nurdan Kavaklı Ulutaş is an Assistant Professor of English Language Teaching at Izmir Demokrasi University, Türkiye. Her work centers on language teacher identity, assessment, and, more recently, developing a critical look at AI in education. She is co-founder and Editor-in-Chief of the Journal Futuristic Implementations of Research in Education (FIRE).