Development and Evaluation of a System to Support Structural Integration in Academic Writing Processes Using Generative AI

Abstract

In this paper, we developed and evaluated a support system for undergraduate students that supports "structural integration," a reading and writing process in which integrated understanding is obtained by associating and integrating components identified from documents. The system we developed consists of a “highlighting function” that identifies the components of a document and describes them, and a “suggestion function” that provides hints for students to review and revise their own decompositions made with the highlighting function. The suggestion function was implemented by using the GPT API to generate questions that require students to answer the difference between the decompositions performed by the student and the assignment designer who created the assignment. In an evaluation experiment, undergraduate students wrote summaries based on the content of the documents by reviewing the structure of the documents they had found themselves in an interactive process with the suggestions generated by the system. This suggested that the system was effective in eliciting structural integration from the undergraduate students.