Integrating AI and Human Feedback to Optimize Interdisciplinary Proposal Writing in Science Graduate Students

Although the research on AI-assisted learning rapidly accelerates, there is still a notable gap in investigation into AI-generated content (AIGC) tool use in interdisciplinary academic writing. The present study addresses this gap by investigating the preference of AIGC use by science master’s students in writing interdisciplinary research proposals, and relationship between the use of AIGC tools and their writing abilities was assessed. Through a mixed-methods approach, we collected two iterations of research proposals from 33 students across diverse disciplines, along with their reflections on ChatGPT use. Faculty from both substantiation-track and teaching-track backgrounds were invited to evaluate the proposals and provide individual feedback. Quantitative topic modeling was used to identify theme patterns and areas where AI feedback contributed to improvements.The results showed a strong correlation between AIGC tool usage and higher writing scores, with students reporting that AI significantly enhanced the clarity, coherence, and technical accuracy of their proposals. However, findings also underscore that AI tools are less capable of handling deeper reasoning of niche and the methodological development required for writing in interdisciplinary contexts. Therefore, a balanced approach is necessary, where AI assists in technical aspects and human guidance focuses on advanced reasoning and structuring ideas.