From text to symphony : a longitudinal study of GenAI integration in a group project assessment

Abstract

This longitudinal mixed-methods study examined the evolution of student engagement with generative artificial intelligence (GenAI) in a health science general education course at The Hong Kong Polytechnic University across seven cohorts (2022–2025, n = 398). Drawing on 74 GenAI usage logs and associated reflections, course-end surveys (n=185), and focus groups, it tracked a shift from text-based exploration (e.g., ChatGPT for brainstorming) to multimodal orchestration (e.g., Suno AI, Perplexity for music, verification, and video). Surveys revealed consistently high perceptions of GenAI fostering critical thinking (M=4.6–4.8/5) and tool proficiency enhancement (Mean = 4.5–4.8/5). Qualitative data highlighted persistent challenges including hallucinations requiring human validation, assessment anxiety over rubrics, preferences for human versus AI voices, and a “paywall barrier” limiting premium multimodal access, exacerbating equity divides. These findings advocate redefining AI literacy as workflow management, urging institutions to provide premium AI tools and ungraded practice spaces to ensure equitable assessments for all students.