SHIH, YU TING
Graduate Institute of Educational Information and Measurement, National Taichung University of Education Ph.D. Candidate
E-mail：ytshih0121@gmail.com
CHANG,DUN CHENG
Master Program of Executive Business Administration, College of Management, National Taichung University of Education Assistant professor
E-mail：dcchang@mail.ntcu.edu.tw
Abstract
As Generative Artificial Intelligence (AI) technology matures, automated grading and feedback systems have become crucial tools for reducing teacher workload and providing personalized learning guidance. This study investigated the effectiveness of a self-designed “Generative AI Automated Grading and Feedback System” implemented in a “Data Science and Problem Solving” course for university freshmen. Based on the Technology Acceptance Model (TAM), this mixed-methods study collected 70 valid responses to analyze the relationships among Perceived Ease of Use (PEOU), Perceived Usefulness (PU), and overall Satisfaction (SAT), alongside students’ specific learning pain points.
Path analysis results revealed that PU completely mediated the relationship between PEOU and SAT (𝛽 = .70, 𝑝 < .001) , demonstrating that the system’s actual ability to improve report quality is the core determinant of student satisfaction, rather than mere interface intuitiveness. Furthermore, qualitative thematic analysis indicated a strong student need for concrete guidance, knowledge structure integration, and learning scaffolding from the AI.
The study concludes that future educational AI optimization should focus on refining underlying prompts and assigning professional virtual personas to the AI, thereby enforcing the generation of logically structured recommendations to maximize the pedagogical value of Generative AI in higher education.
Keywords：Generative AI, Automated Scoring System, AI Persona