Abstract

Generative artificial intelligence, represented by large language models (LLMs), has the potential to revolutionize applied linguistics, such as language teaching, language learning, and language testing. However, how to write an effective prompt (i.e., prompt engineering) remains underexplored in applied linguistics. This study aims to elaborate on the important elements of prompt engineering, including persona, audience, contexts, instruction, and output specification. Examples were used to demonstrate how these elements can form an effective prompt in applied linguistics contexts. Besides, this study also delineates several important prompting strategies to handle more complex tasks, such as iterative prompting and few-shot prompting. Most importantly, it provides some practical tips to mitigate the potential shortcomings of LLMs, including data privacy, potential bias, explainability, and hallucinations, from the perspective of prompt engineering. This study highlights the potential applications of LLMs in applied linguistics by prompt engineering and methods in prompt engineering to navigate the potential pitfalls of LLMs, fostering the application of LLMs in applied linguistics effectively and responsibly.

This work is licensed under a Creative Commons Attribution 4.0 License.