Evaluating Learning Transitions Between Program Codes Using an AI-Based SOLO-Adapted Methodology

AI (Artificial Intelligence) has significantly influenced and transformed the way teaching, learning and assessment occur. In this era, AI has become the center of discussion and debate among educators and students in programming education. As such, current evaluation systems seek to provide immediate feedback by leveraging AI capabilities. Feedback is important because it gives both the student and the teacher a holistic view of the learning progress. However, there remains a lack of adequate solutions for evaluating and linking prior knowledge to current knowledge (learning transition) in programming education. To close the gap, we developed an AI model (based on the Google Gemini Application Programming Interface (API) and also following Design Science Research Methodology (DSRM)) that evaluates and categorises program code into SOLO (Structure of Observed Learning Outcome)-adapted stages. The aim of the AI model is to evaluate two program codes by comparing the earlier version with the current code to provide informed feedback. High-level categories of learning transitions include progression (improvement), regression (learning has deteriorated), and no learning (stagnation). The model was tested in all 25 possibilities (learning transitions) to assess its capacity on the given synthetic test cases.