Design and Performance Validation of a Computational Thinking Gamified Intelligent Tutoring System Focusing on Thinking Process

Computational thinking (CT) is recognized as a crucial problem-solving skill in the age of artificial intelligence. However, there remains a significant gap in the availability of intelligent tools that effectively foster computational thinking in line with contemporary educational needs. To address this, the present study proposes a gamified intelligent tutoring system that supports learners' CT development through a process-oriented assessment methodology and targeted guidance. The assessment framework leverages machine learning algorithms to construct predictive models based on the frequency of six behavioral elements observed during gameplay, including a process-oriented failure risk assessment model for formative evaluation and a summative computational thinking level assessment model. The system identifies the critical process element contributing most to the prediction outcome to provide targeted guidance for learners at risk of failure or exhibiting low computational performance. Using a dataset of 754 real learner cases, five machine learning algorithms were compared, and multi-metric evaluation confirmed strong predictive performance, with accuracy exceeding 80%. Logistic regression was selected as the final algorithm due to its simplicity, efficiency, and interpretability, ensuring practical applicability. The tool will enable the scaling and personalization of computational thinking development.