Exploring the Educational Implications and Improvement Processes of an AI-Based Process-Centered Assessment System: An Action Research Approach

This study explored the educational implications and improvement processes involved in applying an AI-based process-centered assessment system at G Science High School for the Gifted. Drawing on action research conducted by four teacher-researchers, the study collected and analyzed self-report journals, semi-structured interviews, and student and teacher surveys. Four key findings emerged. First, the AI-based assessment system demonstrated meaningful educational potential by providing personalized feedback and reducing repetitive scoring workload, thereby enabling teachers to devote greater attention to individual student interaction. Second, notable limitations were identified, including inconsistent AI scoring, difficulties in recognizing handwritten responses, and inadequate evaluation of higher-order thinking and affective domains. Third, through reflective cycles across two implementation phases, participants inductively identified three conditions for effective field application: selectively determining the scope of AI use based on subject and task characteristics, designing precise rubrics, and establishing teacher-led verification systems for AI-generated results. Fourth, teachers reconstructed their professional identities by positioning AI as an auxiliary assessment tool, expanding their roles as rubric designers and critical reviewers of AI outputs, and reaffirming the value of the teacher-student educational relationship. These findings suggest that AI-based assessment systems can meaningfully contribute to realizing the aims of process-centered assessment, provided that teachers’ active engagement and professional judgment remain central to the evaluation process.