The Accuracy, Linguistic Appropriateness, and Prompting as Properties of Use in ChatGPT Relevant to Korean EFL Learners

Abstract

The primary aim of this research study is to investigate the potential use of AI chatbots as a useful source of language input for L2 learners in terms of accuracy, linguistic appropriateness, and prompt-based interaction through a case study methodology. The researcher, who is a PhD student, acted in dual roles as both the researcher and participant, having seven 20-minute chat sessions with the voice-based ChatGPT for social and academic purposes. To measure the grammaticality and appropriateness of ChatGPT’s language for different levels of EFL learners/users, the researcher first investigated the output of ChatGPT in unprompted mode. Then, the researcher attempted to prompt certain types of language use regarding linguistic features. The results showed that the language from ChatGPT is grammatically accurate, typically indicating over 90% accuracy in form. However, the analysis also showed that the language produced in unprompted mode is more suitable for higher-level users due to its lexical and syntactic complexity, as well as idiomatic aspects. The findings underscore the necessity of crafting effective prompts to align ChatGPT’s language output with learners’ different proficiency levels.

Ähnliche Arbeiten