Creating an Effective, Simple Language Simulation with AI for Underprivileged Students in a Small Village School

Creating an effective, simple AI language simulation can be challenging, especially with students who have a low level of language proficiency. This task becomes even more complex during a hot spell, in a sunny and stuffy classroom in a small village school in rural Hungary. Difficult terrain. Moreover, these children seem to understand and appreciate artificial intelligence less, showing little interest as achieving tangible results with it can be ‘difficult’. This article discusses how I managed to save my class using ChatGPT.

The situation was not simple, and yet I still had to try to teach English, albeit with little success initially. This continued until the lesson covered basic vocabulary, grammar (asking questions, giving simple one-sentence answers, as the children were real beginners), describing daily routines, and learning about different professions — for instance, asking about a typical day in the life of a firefighter. Naturally, this was rather dull, and there was no chance they would engage with it.

Giving laptops to students often results in ‘mayhem’ since as soon as they turn them on, they find their favorite online car racing games, and my lesson plan will fall apart. Therefore, I created more controlled conditions, telling them that they would be talking to a ‘person’ in the USA. Although many recognized the ChatGPT interface quickly, they started to enjoy the interaction. I explained that they were interacting with a specific person loaded with information, which they could extract by asking questions (this led to a request to switch to Hungarian to make communication easier, which I denied, but they didn’t become discouraged). My plan was for everyone to ask questions to ChatGPT, with everyone able to see the answers. They were to take notes from these interactions, which would keep everyone busy, and eventually, they would compose an essay. This last part did not happen in our case, but I think it’s a great method to keep students engaged throughout, even if they are not asking questions at the moment.

How did I do all this? I simply asked ChatGPT to respond to the children using a maximum of 500 words of vocabulary, sticking to simple, basic entences. There were some inside jokes involved too (e.g., I knew a question about BMW or Mercedes would come up, and I taught the chatbot to react strongly against Mercedes — it was a big hit :))

Here are some details I provided based on the lesson material:

Joe: firefighter, American, 42 years old, married, 2 children (Anna 10, Frank 6). Lives and works in New York. Daily routine: early rise, works 12-hour shifts, sometimes night shifts. It’s a dangerous job with a lot of stress. Hobbies include playing football (soccer, not American football). He has a few good friends, Adam and Chris are his best friends. He likes casual clothes, not into fancy stuff. He drives a 10-year-old BMW (you can improvise on that, you hate Mercedes and you make sure people know it!). His favorite football team is Arsenal from England.

If you have time, you can hide more inside jokes and references, even relating to specific students (e.g., someone’s favorite left-back in football might be XY from the group). Based on the task sheet they received, they had to ask questions, and they were increasingly amazed but also enthusiastic to discover that Joe, the firefighter, seemed like someone they knew. It was also crucial to set the language level, the length of the answers, and the vocabulary width (500 words, very beginner). Occasionally, ChatGPT attempted more complex sentences — interestingly, at times when a more skilled student asked a tougher question, as if it sensed that it could loosen up :)

Using AI in the classroom provided an exciting turn and practiced authentic language use for the students. There was immediate feedback, and they genuinely looked forward to what the AI would respond to their questions. There wasn’t a lot of preparation involved, and it generated a significant amount of language output.