How Do the Process and Attitudes of Writing Japanese Task Texts Differ Depending on Whether Web Search and Language-generating AI is Used?

Considering the recent spread of language-generating AI tools, this study conducted a survey to clarify how writers with a certain level of Japanese academic writing skills engage the process of writing tasks with and without the use of web search and AI, and to explore differences in the writing process and attitudes among the same Japanese language learners. The subjects were three Chinese native-speaking postgraduate students with advanced Japanese proficiency, who were recorded on the screen of writing task texts (within 600 characters) (task 1) and task 2 with Web/AI and interviewed in 2023, respectively. The six texts were then evaluated by three Japanese language teachers in addition to self-evaluation. In task 1, all participants wrote with a strong awareness of the keywords required by the task instructions. In task 2, although all the participants started with a theme search, they used various methods of use, such as speed-reading and citing English literature using the English-Chinese translation function, then importing AI output and combining it with his own sentences, selecting Japanese articles from the search results and AI output results and connecting them with her own sentences, and using the Japanese-Chinese translation function to speed-read Japanese web articles and connecting them with her own sentences. The methods of use showed individual characteristics. The time required for task 2 was longer for all participants, and only one participant gave a high self-evaluation of task 2, and all mentioned the difficulty of deriving logical sentences from multiple information sources within the word limit. All were aware of the danger of plagiarism. The results of the teacher evaluation showed that only one student had a higher score for task 2 than for task 1, and the same order of evaluation trend was observed for tasks 1 and 2. These results confirm that the use of Web/AI does not necessarily make the task easier for advanced students, at least in writing tasks with a high constraint of around 600 characters.