Why Language Teachers Need a Different AI Conversation
Generative AI tools are increasingly visible in educational spaces, often accompanied by excitement, uncertainty, and uneven guidance. Much of the practical guidance around AI in education emphasizes efficiency, accuracy, and productivity (Case et al., 2025). While this framing may be useful in some contexts, it does not fully reflect the goals of language teaching.
Language education is not only about correctness. It is about meaning-making, voice, identity, relationships, and context. Across ESL, bilingual, and world language classrooms, students learn to communicate with real audiences and make linguistic choices shaped by social and cultural norms. When generative AI enters these classrooms, it raises important pedagogical questions: whose language does AI treat as “correct”? Which voices and varieties are made visible or marginalized? How might AI feedback shape students’ sense of authorship, confidence, and agency as language users?
At the same time, generative AI is already embedded in many students’ everyday language practices beyond school, from drafting messages and translating texts to seeking explanations. Language classrooms therefore offer a meaningful opportunity to engage AI deliberately, helping students learn to question, evaluate, and make informed choices when interacting with AI systems.
This moment invites a different conversation about AI in language education: one that moves beyond tools or productivity claims and instead centers pedagogy, ethics, and equity. Rather than starting with what AI can do, we suggest beginning with language learning goals and the social and ethical commitments educators want students to develop. From there, teachers can consider how AI might be used to expand opportunities for language practice, voice, and critical engagement with meaning.
Our Stance: A Human-Centered Approach to Generative AI
Before discussing specific tools or classroom strategies, we begin with a shared stance: generative AI should be approached as part of a broader pedagogical and ethical ecosystem in language education (Godwin-Jones, 2025). Language learning is fundamentally social, involving interaction, negotiation of meaning, and reflection on how language is used across contexts and communities. Any engagement with AI needs to be grounded in these core purposes.
A human-centered approach begins with instructional intent rather than technological capability. Instead of focusing on what AI can produce, teachers consider what students are meant to learn and practice, such as communicating ideas, shaping messages for different audiences, making strategic language choices, and developing confidence as multilingual or emerging multilingual speakers. From this perspective, generative AI becomes one possible resource among many, and its value depends on how it is designed into learning tasks.
This stance foregrounds teacher agency and professional judgment. Language teachers routinely make decisions about scaffolding, modeling, feedback, and assessment in response to their students and instructional goals. Similar judgment is required when integrating AI, particularly when deciding when its use supports learning and when it may limit opportunities for student engagement or reflection.
Finally, a human-centered stance treats ethical and social responsibility as central to language teaching. Generative AI systems reflect particular language norms and values, shaping whose language is recognized and whose voices are less visible. Questions of equity, representation, ownership, and accountability are therefore inseparable from pedagogy.
Introducing the Position–Design–Enact (PDE) Framework
To move from stance to practice, language teachers need a way to think systematically and flexibly about how generative AI enters their classrooms. Building on a human-centered approach, we propose the position–design–enact (PDE) framework (see Figure 1) to support instructional decision-making that is pedagogically grounded and ethically informed.
The PDE framework consists of four interrelated dimensions. Position refers to how generative AI is situated within the classroom ecology: its relation to students, teachers, and language learning goals. Design focuses on intentional planning, including task structure, timing, and expectations for AI use. Enact centers on what happens in practice as teachers and students interact with AI during instruction. Running through all three dimensions is criticality, which involves sustained attention to language ideologies, representation and bias, authority and agency, and ownership and ethics. These considerations inform how AI is positioned, how tasks are designed, and how classroom interactions unfold.
The PDE framework is not a linear sequence. Decisions about positioning, design, and enactment inform one another and often need to be revisited as contexts and learners change. For this reason, the framework is adaptable across ESL, bilingual, and world language classrooms and can be taken up from different entry points depending on instructional needs.
A P-D-E Centered Critical GenAI Pedagogy for Language Teaching
Figure 1. The PDE Framework for Critical AI Integration
Unpacking the PDE Framework with a Classroom Example
Position: Locating Generative AI in the Classroom Ecology
The first dimension of the PDE framework, position, invites teachers to consider how generative AI is situated within the classroom ecology. Rather than beginning with specific tools or features, positioning focuses on relationships: the relationship between students and language, between teachers and instructional goals, and between AI and the learning environment as a whole.
Positioning involves clarifying what role generative AI plays in relation to language learning. At different points in instruction, AI may function as a source of feedback, a planning support, a model for analysis, or a partner for reflection. What matters is not assigning a fixed role but making that role visible and intentional. How AI is positioned shapes how students interpret its authority, how they engage with feedback, and how responsibility for learning is shared.
For language teachers, positioning is closely tied to instructional purpose. Tools that support noticing patterns in language use may be positioned differently from ones that generate sample texts for discussion. In language classrooms, these decisions are further shaped by students’ linguistic repertoires, proficiency levels, and prior experiences with AI. Thoughtful positioning helps ensure that AI use aligns with learning goals rather than quietly redefining them.
Positioning also involves setting boundaries. Teachers decide when the use of AI is appropriate, when it should be postponed, and when it should be set aside entirely in favor of human interaction. Making these boundaries explicit through classroom norms, task instructions, or classroom conversation helps students understand why AI is used in some contexts and not in others.
In Practice
In this example, the setting is a high school world language classroom. The learning goal is for intermediate-level students to write an informal inquiry email in response to an online roommate advertisement in a target-language-speaking context. With the goal established, the teacher determines how generative AI will function in the lesson.
In this case, AI is positioned as a writing coach. Its role is to help students notice potential issues in their drafts, explain relevant grammar patterns, and suggest alternative ways to express ideas. With this role defined, appropriate tools can be selected. Most commonly available generative AI tools (e.g., ChatGPT, Gemini, Copilot) are suitable for this task, provided they are accessible to students and permitted by school or district policy.
Design: Planning with Critical Awareness
Design focuses on finding appropriate instructional spaces for the role of AI that has already been established through positioning. Once teachers are clear about what AI is meant to do in a lesson, design addresses how and when that role is taken up within learning tasks. This includes decisions about task structure, timing, prompts, and expectations, as well as how AI use is coordinated with other forms of instruction and interaction. Design translates positioning into concrete learning opportunities that support language development.
Design decisions should always be guided by language-learning objectives and planned with critical awareness. When generative AI is introduced, teachers need to consider what beliefs about language the tool may reinforce (language ideologies); whose cultures, experiences, and knowledge are included or excluded (representation and bias); who is positioned as the expert in the learning process (authority and voice); and who owns the work that is produced (ownerships and ethics). Designing with criticality means anticipating where these issues may shape student learning and intentionally structuring tasks to address them. This can involve adjusting prompts, planning opportunities for discussion and reflection, and creating space for students to question AI output instead of accepting it directly.
In Practice
Following the positioning of AI as a writing coach in a high school world language classroom, the design of the lesson turns to the four critical domains. In this context, generative AI is likely to favor standardized and formal language, given the types of texts it draws on. It may not recognize informal registers, region-specific conventions, or culturally situated ways of writing. There is also a risk that students may treat AI feedback as authoritative, especially when it is delivered with confidence.
These considerations guide several design decisions. AI feedback is not positioned as the sole source of input, so students receive multiple forms of feedback. The lesson is structured to support students in questioning and evaluating AI suggestions, rather than accepting them automatically. Time is also built into the lesson for explicit discussion of language norms and biases, so that students can reflect on how AI treats register, appropriateness, and correctness.
To support these goals, the lesson sequence includes several stages. Students begin by drafting their emails independently and completing a self-evaluation. They then receive AI feedback. As students review this feedback, they complete a structured worksheet that prompts them to consider each suggestion, decide whether to accept it, and explain their reasoning. Peer feedback follows, allowing students to compare perspectives. A whole-class discussion takes place before revision, giving students space to compare feedback sources and reflect on differences. Students revise their writing using insights from all sources, with responsibility for final language choices remaining with the writer.
Enact: When Generative AI Meets Classroom Practice
While positioning and design shape the conditions for AI use, enactment focuses on what happens when generative AI is taken up in everyday classroom practice. It brings attention to classroom moments as they unfold, including how students interpret AI output, how they respond to it, and how teachers adjust instruction in response. Enactment is where planning meets practice and where assumptions about AI use are tested through interaction.
In language classrooms, enactment makes visible how authority and agency circulate during AI-supported activities. Even when AI use is carefully planned and scaffolded, students may still respond to AI feedback in different ways. Some may treat suggestions as authoritative, others may question them, and some may adopt them selectively based on their communicative intentions. These responses are shaped not only by how AI has been positioned and how tasks have been designed but also by students’ prior experiences, language identities, and comfort with risk-taking. Observing and responding to these dynamics is an important part of teaching with AI.
Enactment also brings ethical considerations into classroom practice in concrete ways. Questions of ownership, authorship, and responsibility often emerge as students decide which AI suggestions to accept, revise, or reject. Classroom conversations, peer interaction, and guided reflection can help make these decisions visible. Through these interactions, students articulate their reasoning and develop greater awareness of how AI influences language use and meaning-making.
A key feature of enactment is responsiveness. Teachers observe how students engage with AI and make adjustments during the lesson. This might include pausing to address misconceptions, reframing instructions, or creating additional space for discussion. These in-the-moment decisions often lead teachers to revisit earlier choices about how AI is positioned or how tasks are designed. The enact dimension therefore highlights the ongoing and adaptive nature of teaching with AI.
In Practice
Instruction begins with familiar language-teaching practices. Students examine an authentic roommate advertisement to understand the communicative context and expectations of the task. They then analyze a model response and co-construct a sample email as a class before moving into independent writing. These steps establish a shared understanding of purpose, register, and audience before any AI use occurs.
After students draft their emails and complete a self-evaluation, generative AI is introduced in its previously defined role as a writing coach. Students submit their drafts using a guided prompt and review the AI feedback with the support of a worksheet. As students work through the suggestions, classroom interaction reveals differences in how AI feedback is interpreted. Some students focus closely on surface-level corrections, while others question whether suggested changes fit the informal register of the task. These moments allow the teacher to circulate, listen, and respond by prompting students to explain their reasoning and compare AI suggestions with their original intentions.
Whole-class discussion becomes an important enactment space. Students are invited to share patterns they notice in AI feedback, including instances where suggestions conflict with the norms of informal writing. Discussion also surfaces questions about whose version of correct language AI appears to prioritize and how the types of texts AI draws on may shape its responses. Through guided conversation, students consider when AI feedback is useful for their goals and when it should be questioned or set aside.
Looking Ahead
As generative AI develops rapidly, new tools appear almost every day. Many teachers feel pressure to try these tools or uncertainty about where to begin. This sense of urgency is familiar. Past waves of educational technology have brought similar excitement and overload. There is little doubt that AI will shape teaching and learning in many domains. At the same time, these changes call for a broader human-centered view of what teaching and learning are meant to do (Lu et al., 2025). The PDE framework was developed with this context in mind. It is intended to help language educators orient themselves in a fast-changing landscape without feeling compelled to chase tools or trends. The framework does not promote any specific technology or prescribe a single way of using AI. Instead, it offers a way to approach AI use thoughtfully by grounding decisions in language-learning goals, critical awareness, and professional judgment.
In this sense, the PDE framework aligns with calls in the field to move beyond simple judgments about whether AI is “good” or “bad” (Tian and Wang, 2025). We believe that the value of AI in language education lies less in the tools themselves and more in how educators use them. The PDE framework offers one way to support this work by keeping criticality and human judgment at the center of language education in this new era.
References
Case, R., Liu, L., and Mintz, J. (2025). “Integrating AI Technology into Language Teacher Education: Challenges, potentials, and assumptions.” Computers in the Schools, 1–7. https://doi.org/10.1080/07380569.2025.2458950
Godwin-Jones, R. (2025). “Language Teacher Preparation for an AI World: A human ecological perspective.” In Z. Tian and C. Wang (Eds.), Rethinking Language Education in the Age of Generative AI (pp. 29–46). Routledge. https://doi.org/10.4324/9781003426929-4
Lu, X., Zaitseva, A., and Troyan, F. J. (2025). “AI Literacy: A core practice in world language education.” Foreign Language Annals. https://doi.org/10.1111/flan.70037
Tian, Z., and Wang, C. (Eds.) (2025). Rethinking Language Education in the Age of Generative AI. Routledge.
Zhongfeng Tian is associate professor of Bilingual Education at Rutgers University–Newark. Informed by his multilingual and transnational background, his scholarship examines translanguaging and critical AI literacy in language education. He serves as associate editor for the International Journal of Bilingual Education and Bilingualism and The Modern Language Journal.
Xinyue Lu is assistant professor of Second Language Acquisition at Howard University. Her research examines how language ideologies shape the experiences of minoritized learners and teachers in world language and TESOL classrooms. She also studies language teacher education, focusing on the identity and development of language teacher educators and on AI literacy in teacher preparation.
