How to harness AI in language education

Artificial intelligence is rapidly transforming how languages are taught and learned. As an applied linguist and teacher educator, Penn GSE Professor of Practice Anne Pomerantz sees AI as a tool with the potential to enhance language education, helping teachers refine their practice and providing students with new ways to achieve their learning goals. 

Since ChatGPT’s debut in 2022, she has experimented with generative AI through a linguist’s lens, investigating how chatbots interact with language learners and considering the applications of AI tools to language education.

Pomerantz says AI can be useful for instruction and practice but is not a substitute for a well-designed language class or real-life language use. It can complement—but not replace—human instruction and interaction. Blending AI with human instruction can give educators powerful new tools and help learners expand their skills.

As educators return to the classroom, Pomerantz offers these tips for navigating the quickly evolving technology.

When AI is a helpful tool

1. Creating custom materials

One of the big challenges for language teachers is finding ways to tailor activities and materials to individual students’ levels. In one class, there may be five students whose language proficiency is quite advanced, another five who are really struggling, and fifteen who are somewhere in the middle.

AI can help teachers to create learning environments that provide each learner with the right amount of challenge and support. Whether it is asking AI to generate different versions of a reading passage or using chatbots to create speaking tasks that push each learner just a little bit outside their comfort zone, teachers now have a tool for customizing instruction.

“Differentiation,” Pomerantz explains, “is a necessary, but time-consuming task for language educators. Having a tool that can quickly produce multiple versions of the same text or task can help language teachers to support all learners more effectively.”

2. Developing speaking skills

Ask reluctant language learners what they find most daunting or uncomfortable about learning an additional language, and they will likely say “speaking.” Language educators have long struggled to find ways to develop learners’ oral language skills.

Chatbots may encourage shy or apprehensive students to practice speaking, since they create a more private space for working through pronunciation or rehearsing an interactional routine. As Pomerantz notes, “Chatbots have infinite time and patience. They will never refuse to practice again or mock an anxious learner’s pronunciation.” Chatbots’ ability to engage in extended conversations also makes them a great resource for students to develop the skills necessary to participate in longer, more complex interactions—something that is often hard to simulate in the kinds of peer-to-peer conversations that characterize language classrooms.

Pro-tip: Chatbots are willing conversational partners, but they need extensive guidance to be effective. Educators must carefully engineer prompts that direct how the bot engages with learners.

3. Assessing students’ speaking and writing

Spontaneous, non-automated assessments like in-class writing tasks or on-the-spot oral exams have long figured into language classes. Language educators want to know what their students can do independently in the language of instruction. But giving meaningful feedback on students’ written and oral performances is, Pomerantz acknowledges, “a bear.”

“Language teachers provide feedback in principled and targeted ways. They don’t correct every grammatical error or comment on every pronunciation issue,” Pomerantz says. “Knowing what to flag and how to provide feedback can be a very labor-intensive process.” AI systems can be prompted to systematically highlight certain errors and to provide specific forms of corrective feedback. In addition, they can be trained to document patterns across students’ language use over the course of a school year. This longitudinal view of students’ development, Pomerantz says, is something busy teachers often struggle to capture and one that can really be useful for helping students to achieve their language learning goals.

When human instruction is preferable

1. Handling the interpersonal dimensions of interaction

Today’s chatbots sound impressively natural, but do they interact like humans? While it is difficult not to be wowed by recent advances in chatbot technology, Pomerantz notes that human interaction is rarely so neat and orderly. Hedges, self-corrections, interruptions, abrupt topic changes—these are all part of normal, everyday conversation. And, they serve important functions.

Take a word like “um” in English, Pomerantz says, “it signals to your interlocutor that you are uncertain, or that you have something to say, but are not quite ready to say it yet.” These filler words can help language learners to buy some time in a conversation without losing their turn to talk. Chatbots, at least for now, tend to be very orderly and unflappable conversation partners. This is what makes those customer service bots so infuriating. “The flip side of their precision and patience,” Pomerantz observes, “is that they don’t provide learners with much practice navigating the emergent and contingent nature of human interaction. Language learners need opportunities to use the language they are studying with real people; otherwise, they will develop competencies that are oriented primarily to interacting with robots.”

Language educators can help learners to notice how human interaction works and to develop the strategies they need to navigate conversational situations.

2. Developing intercultural skills

Part of learning a new language involves engaging with the pragmatic dimensions of language use or what is meant by what is said, written, or signed. “A classic example in Philadelphia English,” Pomerantz offers, “involves the expression, ‘Hi (or sometimes hey), how’s it going,’ a phrase used at the start of an interaction. Is this a conversational opening, a sincere inquiry, or simply a way to acknowledge someone’s presence? Many a newcomer has wondered why locals walk away before hearing the answer to a question that figures into most English language textbooks.

Educators offer human insights, guiding students toward contextually situated, culturally relevant language use in ways AI’s machine learning can’t replicate. And, Pomerantz adds, language learners who aspire to use their language skills in real life may need some help to develop their pragmatic competence. “It’s not about directing learners to memorize a set of relationships between phrases and their meanings but rather constructing activities that push learners to notice how language is used in real life and to consider how we engage in the process of meaning making.”

This kind of reflexivity, Pomerantz observes, is very human and something skilled language educators excel at cultivating. “Sometimes it takes the act of learning a new language to reflect on how we make meaning in our comfortable, familiar languages.”

3. Facilitating critical thinking

AI tools gather information from across the internet with few filters, and the materials they produce can contain misinformation, stereotypes, objectionable content, or bias. Worse yet, AI systems can be prompted to create texts, images, audio recordings, and videos that are explicitly designed to be manipulative or harmful. In contrast, language teachers understand cultural nuances and can facilitate respectful, culturally and socially conscious discussions and materials.

Likewise, language educators can help language learners to develop the critical thinking skills they need to engage with and evaluate machine-generated content. Pomerantz observes that immigrant communities in the U.S. are often targeted by scammers, as are students studying abroad. Along with reflexivity, Pomerantz encourages language educators to emphasize activities that foster inquiry, fact-checking, and indeed even skepticism.

“Many of us have had language learning experiences that were simply exercises in memorization or repetition,” she says, “but the language classroom is fertile ground for activities that require critical thinking.”

The bottom line

Educators should explore AI tools but remember that language learning is a very human activity.

“AI can be very helpful,” Pomerantz says, “and interacting with it can help educators determine if and when they want to use it ethically, productively, and competently. But it is important to remember that ultimately language learning is about connecting with other people.”