Large Language Model
The Rise of Large Language Models in CALL: Opportunities and Challenges
Introduction
In the past two years, the field of computer-assisted language learning (CALL) has undergone a major transformation due to the emergence of large language models (LLMs) such as ChatGPT. Frankly speaking, the technical means discussed in the previous blog are all subsumed by large language models. This blog post explores how LLMs are used in CALL, their impact on motivation and learning, and the delicate balance between human teachers and AI tools.
LLM as a language practice tool
One of the most promising applications of LLMs is conversational chatbots as language practice. A recent user study involving 160 English learners showed that interactive sessions with an LLM chatbot significantly improved learners' confidence and performance in real-world conversations. The study found that learners of different proficiency levels benefited differently from these interactions, with lower-level learners showing more significant improvements. In addition, the study proposed a method for detecting language complexity by asking learners to click on unfamiliar words to get definitions. The number of clicks correlates with the complexity of the language, providing valuable insights into learners’ comprehension challenges.
Error Detection and Writing Improvement
In the area of writing, LLMs have shown remarkable potential in identifying surface errors such as spelling and grammar. A study published in the Arab World Journal of English found that ChatGPT could effectively detect most surface errors in EFL learners’ writing. However, it struggled with deeper structural and pragmatic errors, which are more easily identified by human teachers. This highlights the complementary nature of LLM and human teachers, where AI can handle routine error detection while teachers focus on more subtle aspects of writing.
Opportunities and Challenges in Educational Settings
Another study examined the potential of ChatGPT as an educational tool in communication, business writing, and composition courses. The results were both encouraging and cautionary. On the one hand, ChatGPT was able to generate accurate and reliable responses, making it a valuable resource for students seeking answers to theoretical questions and generating ideas for applied tasks. On the other hand, the study highlighted significant challenges, including the risk of unethical use by students, which could lead to a decline in critical thinking skills. Educators were also concerned about the difficulty in distinguishing between human-generated and AI-generated work, which complicates the assessment process.
Balancing Human Teachers and AI
The integration of LLM into education raises questions about the role of human teachers. A study focusing on the complementary relationship between ChatGPT and teachers identified several roles for AI, including interlocutor, content provider, and teaching assistant. On the other hand, teachers play a key role in coordinating resources, promoting active learning, and raising awareness of AI ethics. This highlights the need for a balanced approach, where AI is a tool that augments, rather than replaces, human teaching.
Conclusion
The past two years have seen significant progress in the use of LLM in CALL. From improving conversational skills to enhancing writing and motivation, these models offer many opportunities. However, challenges such as ethical use and the need for teachers to collaborate with AI must be addressed. As we move forward, it is critical to harness the power of LLM while retaining the irreplaceable value of human teachers.
