HUMAN-VERIFIED GENERATIVE AI IN FYP PROPOSAL DEVELOPMENT

About the Project

Developing a defensible Final Year Project (FYP) proposal requires Computing students to identify research gaps, justify problems, and align objectives with feasible scope. Although generative AI can assist drafting, uncritical reliance on its outputs raises concerns about students’ analytical engagement and the reliability of academic work (Zhai et al., 2024). To preserve human agency in AI-supported education (Miao & Holmes, 2023), this study developed and evaluated FYP Proposal Content Generator, a web-based application integrating proposal generation with mandatory source verification. Students upload three relevant journal articles, examine AI-extracted research gaps, confirm or correct each gap, and record supporting evidence locations. Proposals remain inaccessible while any gap is marked unsupported. The workflow makes source verification integral to proposal development, keeping responsibility for evidence assessment and academic decisions with students. Four experts assessed 30 proposals, producing a mean quality score of 4.00/5 (SD = 0.29). Twenty-six proposals (86.7%) received recommendations for acceptance or acceptance with minor revision from at least three of the four experts. Among 119 students, overall usability perceptions were moderately positive (M = 3.56, SD = 0.76); learnability scored highest (M = 3.77), whereas satisfaction scored lowest (M = 3.39). The findings provide preliminary support for the application as a scaffold for evidence-informed proposal development while retaining student responsibility and academic judgement. Because the pilot lacked a comparison group, the findings do not establish learning gains or improvements in final proposal quality. Further refinement and controlled evaluation are required to examine educational effectiveness and the quality of completed proposals.