Today’s post is the second in a series on using artificial intelligence with English learners.

‘English Is About the Process’

Kelly Cray is the president of Northern New England TESOL, a graduate of the Harvard Graduate School of Education, an English-language-development teacher, and co-editor of the TESOL Differences and Disabilities newsletter:

AI has been life changing for me as a teacher, and we use it in the classroom in many ways. Before students even enter, I use it to create visually dynamic HTML-style code for my learning-management system (Canvas), so that I can embed pull-down scaffolded support without increasing visual overwhelm. It can code far better than I can, and I use it to create multilingual tabs for important documents, such as the homework policy.

Although I have found many uses for AI as a teacher, I also use it directly with students. One of my favorite activities involves students choosing a SMART (Specific, Measurable, Achievable, Relevant, and Time-Bound) goal, creating a plan to reach it, and then using Canva (free educator accounts) or Gemini to create an app to help them achieve that goal. App creation is an iterative process where students must keep communicating with the AI in English and explaining problems until the app runs smoothly.

Usually, it takes 20 to 30 “tweaks” before the app works as desired. The students learn while building it and then use it to reach their goal. For example, one student wanted to improve her spelling accuracy by 20%. She created two apps to support her: one to practice common spelling patterns and another to analyze her paragraphs and give her a percentage score based on how many words she spelled correctly.

Another way I build autonomy is by teaching students how to use AI to support their core classes. I teach them how to create bilingual study guides and how to use AI to quiz themselves before a test. I also show them how to use it to identify the most effective ways to memorize key English vocabulary and phrases, so they can clearly explain what they know in relation to test material.

We also play a lot of games to increase engagement during our longer block-scheduled class sessions. One of the students’ favorites is when I use AI to design a picture before class. I project the image, and they must use AI to try to recreate it. Whoever gets the closest wins. These activities also open the door to interesting comparison discussions.

My favorite “win” is the ability to find culturally relevant examples to illustrate concepts on the fly. When I teach literature, being able to present examples of popular hyperboles from three or four cultures creates meaningful discussion and helps us get to know each other better.

AI can also be a powerful tool for equity. Another example of helping my EL students succeed in their content classes is by inputting a math test and asking the tool to adapt the examples to be more culturally relevant for my students. It maintains the mathematical integrity while reducing unintended bias. For example, it can convert a physics problem about the Empire State Building into a similar one about the Burj Khalifa.

AI is transforming our everyday lives, including what occurs in the classroom. We have to adapt. Students need to learn to approach AI through a lens of collaboration and empowerment, not as a way to outsource their thinking. I want them to see that learning English is about the process, not just the product.

‘Preserving Desirable Difficulty’

Bill Tolley is an international educator, learning designer, and instructional leader at the International Community School of Addis Ababa:

When my multilingual grade 10 students at the International Community School of Addis Ababa begin our first unit on equality, I ask them to read excerpts from the Stanford Encyclopedia of Philosophy. They are often struck by how difficult it is. These days, their first instinct is to ask if they can simplify it with AI or ask forgiveness after they already have.

The problem is that simplifying the text immediately removes the struggle students need to build language and content knowledge. They need to encounter the difficulty and attempt to overcome it. Only then does it make sense to decide if AI support might help.

This is why I am increasingly interested in helping students become proficient users of AI while developing the judgment to decide if AI is the best tool for the learning problem they face.

Universal Design for Learning points us in the right direction here. CAST’s UDL guidelines identify learner agency as their primary goal: developing learners who are purposeful and reflective, resourceful and authentic, strategic and action-oriented. Providing different ways to access learning matters, but the larger goal is helping students understand themselves as learners and make increasingly deliberate choices about how they learn.

In my AI-informed classroom, one of the best options is a good old-fashioned physical journal.

Last year, being “on time” in my class meant being in the room, in your seat, with your laptop open to the shared lesson. This year, students grab their journals from the class basket, sit down with a pen, and leave their laptops in their bags. We end with a peer-to-peer chat and laptops away again.

That change was informed by reading on cognitive offloading and the value of preserving struggle, especially for multilingual learners. Research scientist Nataliya Kosmyna’s “Your Brain on ChatGPT” found less cognitive engagement, ownership, and retention among participants relying solely on an LLM. Language scholar Judith F. Kroll describes bilingualism as a “desirable difficulty,” with managing two languages benefiting foreign-vocabulary learning. Associate Professor Shuai Wang found that English learners who improved at switching languages also improved at filtering distractions and shifting their thinking. Researcher Jinhee Kim and colleagues found that “productive failure” can improve learning, with chatbots creating useful friction when learners use them as thinking partners in Socratic-style dialogue.

Taken together, the message was hard to miss for classroom teachers: Remove the struggle too quickly, and we may remove part of a multilingual learner’s superpower. Deliberate changes like tactile journaling at the start of every lesson and face-to-face discussion at the end bookend AI use with cognitive and social load.

In the space between the journal warm-up and discussion cool-down, students have plenty of room and reason to use AI. In the equality unit, “Professor Equitas,” an AI tutor I developed, works with students on the Stanford Encyclopedia of Philosophy article, using Socratic questioning to probe their understanding without simply supplying answers. Simone de Beauvoir helps them rehearse an argument on gender equity. John Rawls helps them test ideas of justice, and W.E.B. Du Bois offers historical perspective on racial division.

I teach students to use a protocol called START and REFINE to prompt Gemini and Gemini Notebook more effectively. START asks them to define the Situation, Task, Audience, and Requirements, then “Tell Back” to verify what the AI thinks the student is asking before proceeding. REFINE asks them to evaluate the response against their purpose and sources to shape the next prompt. For multilingual learners, crafting effective prompts also requires practice in using deliberate, precise language.

I also use an AI Co-Authorship Journal to document students’ use of AI. It includes an AI statement of use, where students state where and how they used AI, and a reflection on that use. This year, I am revising the journal to make that reflection more explicitly metacognitive, asking students to consider when they chose AI, when they chose analog methods and tools instead, and why.

These techniques and strategies support students across language backgrounds in developing the agency UDL identifies as its primary goal. Students decide when AI is the right support and when an analog method or tool will lead to better learning. Making those decisions can reduce cognitive offloading while preserving the desirable difficulty and productive failure that help learning take hold.

Thanks to Kelly and Bill for contributing their thoughts.

They responded to this question:

What is the good and the bad about using artificial intelligence with English learners?

Consider contributing a question to be answered in a future post. You can send one to me at lferlazzo@epe.org. When you send it in, let me know if I can use your real name if it’s selected or if you’d prefer remaining anonymous and have a pseudonym in mind.

You can also contact me on X at @Larryferlazzo or on Bluesky at @larryferlazzo.bsky.social

Just a reminder; you can subscribe and receive updates from this blog via email. And if you missed any of the highlights from the first 13 years of this blog, you can see a categorized list here.