AI in Language Learning 

The rise of generative AI tools is changing the landscape of language learning. While the new tools offer exciting opportunities for language pedagogy, they also come with potential pitfalls and drawbacks. To incorporate AI into your language classroom, it is important to have basic knowledge of how generative AI functions and what it can and cannot do. 

How Generative-AI learns languages?

Every piece of human writing carries a unique and distinguishable voice. Even when we write in the same language and follow its grammatical rules and structures, we make our own aesthetic choices, reflected through the nuances of applied language. What distinguishes AI from other computer technologies is that it is quite good at not only understanding the subtleties of human language but also reproducing it at or near the human level.  

How does AI do this? AI is trained using Large Language Models (LLMs). In simple terms, AI ‘reads’ a very, very large amount of text to identify ‘patterns’ in them, which it then uses to ‘predict’ the next word in a sequence of words. From writing computer codes to translating languages, AI is performing the same core function that it is trained to do (really well!): predicting what comes next. 

How is AI so good at mimicking human language?

It has been fed unbelievably large amounts of data. For instance, it is estimated that the original GPT-3 model was trained on 570 GB of text data, which amounts to 300 billion words! For context, the complete Harry Potter series (seven books) contains just over a million words, so think of the GPT-3 model as having read 300,000 books equal to the complete Harry Potter series!

AI in the Language Classroom

Given the immense data fed into the ChatGPT, it has a super-human capability to mimic languages in a way that would be impossible for any single person by themselves. You can ask ChatGPT to write poems in Chaucer’s style or Rupi Kaur’s. You can ask it to write a speech in Barack Obama’s oration style or Donald Trump’s. You can ask it to talk like a slang-heavy, meme-infused young adult texting or a business executive presenting in a boardroom. This means there are infinite possibilities for language learners as they learn and practice using language in authentic contexts. 

Here are some tangible ways you can incorporate ChatGPT in your language classroom:

1. Setup roleplays: Ask students to roleplay with ChatGPT in a particular setting, say a marketplace in Cairo. Students can submit copies of their conversations as assignments. 

2. Speech recognition: Tools like Elsa Speak, Speechling, or Google’s Pronunciation Tool can analyze students' pronunciation and provide corrections.

3. Vocabulary building: AI tools can help in creating personalized vocabulary lists suited for the learner’s proficiency level and needs. It can also offer customized lists on particular topics and for specific contexts. 

4. Exam and Certification prep: ChatGPT can provide sample responses for different language proficiency levels and provide a score on the answers provided. 

5. Translation and paraphrasing: ChatGPT can be useful in offering translations in multiple styles so that students can observe the variances and nuances that go into translating from one language to another. Similarly, it can paraphrase to show how the same thing can be said in multiple ways, depending on stylistic preferences and situational needs. 

6. Corrections and feedback: Students can ask for corrections and feedback on their work, along with an explanation. This feature can help students experiment more with their language abilities and offer them avenues outside of the classroom to get live feedback and corrections. 

What to watch out for:

1. Less commonly used languages have a smaller digital footprint. For instance, of the 137 languages spoken in the Hawaiian archipelago, many are still not available in written form. Even if a language has an extensive digital footprint, there is no guarantee that all dialects have representation. Relying exclusively on AI could mean that the dominant dialects receive more attention, further alienating regional and less commonly used languages and dialects. 

2. Since AI is trained on existing content, it is possible that certain biases, including but not limited to gender, politics, and culture, seep into the language. Supervised learning is important to ensure that the content produced by AI does not reflect these biases. 

3. ChatGPT has been known to provide inaccurate grammatical explanations, especially with more advanced and complex grammatical structures and concepts. Avoid relying on ChatGPT’s grammatical explanations and instead, use trusted and verified sources for grammar instruction. 

4. Students should be made aware of data privacy issues in using AI. Any information provided to ChatGPT can be stored and used to train AI models. Avoid providing personal and identifiable information to ChatGPT.

5. Overreliance on AI for language learning could mean that students do not get enough exposure to using the language in social settings. Ensure that AI is seen as a supplement to language learning, rather than replacing the critical role of using the language in real-life interactions. 

Further Reading:

• Kim, S., Shim, J., & Shim, J. (2023). A Study on the Utilization of OpenAI ChatGPT as a Second Language Learning Tool. Journal of Multimedia Information System, 10(1), 79–88. https://doi.org/10.33851/JMIS.2023.10.1.79

• Pérez-Núñez, A. (2023). Exploring the Potential of Generative AI (ChatGPT) for Foreign Language Instruction: Applications and Challenges. Hispania, 106(3), 355–362. https://doi.org/10.1353/hpn.2023.a906568

• Young, J. C., & Shishido, M. (2023). Investigating OpenAI’s ChatGPT Potentials in Generating Chatbot’s Dialogue for English as a Foreign Language Learning. International Journal of Advanced Computer Science & Applications, 14(6). https://doi.org/10.14569/IJACSA.2023.0140607