ABSTRACT VIEW
VALIDATION OF A SELF-LEARNING METHOD FOR SECOND-LANGUAGE WRITING USING GENERATIVE ARTIFICIAL INTELLIGENCE
T. Lee1, T. Yamaguchi1, Y. Morimoto2
1 Tokyo Gakugei University, Graduate School of Education (JAPAN)
2 Tokyo Gakugei University, ICT/Information Infrastructure Center (JAPAN)
It is becoming increasingly important for many people to learn second language to communicate with people from different countries and cultures. Learners learn second-language writing by producing second-language sentences, receiving feedback on their sentences from their instructor or peers, and revising their sentences. Through this process, learners can determine parts that need to be revised in their own sentences, and learn second language vocabulary, expressions, grammar, and sentence structure by revising their sentences. When they complete their writing, they can reflect on what they have learned, track their own growth, and become more persistent in learning second-language writing.

However, when learners learn second-language writing alone, it is not easy to learn second-language writing by determining and revising the parts that need to be revised from their sentences (Problem 1). It is also difficult for learners to track their growth as they continue their self-learning (Problem 2).

The purpose of this study was to support self-learning of second-language writing. Specifically, we proposed a self-learning model for second-language writing focusing on generative artificial intelligence (AI) and self-reflection sheet. We developed and evaluated a self-learning method on the basis of the model.

We proposed a self-learning model for second-language writing that focuses on the use of generative AI and self-reflection sheets. In our proposed model, a generative AI generates indirect feedback that does not directly indicate how to improve the second-language sentences or revise the sentences but indicate parts that need to be reconsidered. And the reflection sheet is recorded by learners during self-learning of second-language writing.

To realize our proposed self-learning model, we defined the feedback to be output from the generative AI, created prompts for the AI to output the feedback for the second-language sentences input by the learner, and developed a self-reflection sheet for learners to record their learning processes of writing and revising the sentences including how they revised the sentences. We then developed a self-learning method for second-language writing using a generative AI.

Our developed self-learning method consists of the following five steps.
Step 1) The learner writes a second-language sentences on the basis of first-language sentences.
Step 2) They input the first-language sentences and second-language sentences to the generative AI.
Step 3) They receive the feedback from the generative AI.
Step 4) They revise the second-language writing sentences referring to the feedback. They repeat steps 2) to 4) until they judge that the second-language writing sentences are correct.
Step5) They reflect on what they have learned from self-reflection sheet and track their growth.

As results of evaluation of our self-learning method, we found that learners were able to identify the parts that need to be revised in their second-language writing sentences by referring to the feedback from the generative AI (Problem 1 was likely solved). We also found that they were able to persistently engage in self-learning second-language writing and track their growth (Problem 2 was likely solved).

In the future, we plan to develop a web-based system based on our developed method and conduct a detailed evaluation of the system.

Keywords: Second language, Writing, Self-learning, Reflective learning, Learning method, Generative AI.