Student Characteristics and ICT Usage as Predictors of Computational Thinking: An Explainable AI Approach.

Area of Science:

• Educational Technology
• Computer Science Education
• Learning Analytics

Background:

• Computational thinking (CT) is a crucial 21st-century skill.
• Student individual characteristics and information and communication technology (ICT) use significantly shape CT development.
• Large-scale international data is essential for understanding global CT trends.

Purpose of the Study:

• To identify and explain key predictors of CT development using a novel predictive modeling framework.
• To analyze the nonlinear effects of student-level and ICT usage variables on CT.
• To integrate machine learning and explainable AI (XAI) for educational data mining.

Main Methods:

• Analysis of 81,871 Grade 8 students' data from the 2023 International Computer and Information Literacy Study (ICILS).
• Application of a predictive modeling framework integrating two machine learning techniques.
• Utilizing SHapley Additive exPlanations (SHAP) to interpret complex nonlinear relationships.

Main Results:

• Student-level factors (e.g., educational expectations, books at home) and ICT usage show significant nonlinear predictive patterns (U-shaped, inverted U-shaped, monotonic).
• The SHAP-based approach effectively interprets complex nonlinear effects on CT development, outperforming traditional linear models.
• The study demonstrates a novel integration of educational data mining and XAI.

Conclusions:

• CT development is influenced by a complex interplay of student characteristics and ICT use, often exhibiting nonlinear trends.
• The findings provide actionable insights for designing ICT-integrated instruction and targeted interventions to foster CT.
• Future research should explore longitudinal data to understand CT developmental trajectories and causal mechanisms.