This study aims to model the developmental progression of computational thinking (CT) among elementary school students using multidimensional item response theory (MIRT). Drawing on data from the Bebras Challenge 2024, we administered 12 tasks to Benjamin-level students (grades 5 and 6). Each of the 12 tasks covers one or more CT core components, including abstraction, decomposition, pattern recognition, algorithms, and evaluation. We employed a MIRT framework using a three-parameter logistic (3PL) model to estimate student CT skills across the five CT components and examine CT skill development.

By analyzing the results of the Bebras Challenge, we assessed the CT skills of students and gained insight into their learning achievements. In all, Bebras Challenge results from 55,252 participants, including 28,489 grade 5 and 26,763 grade 6 students, were compared. Utilizing an MIRT 3PL model, we obtained separate skill estimates for each student across five CT components. The research findings revealed significant differences in growth from grade 5 to grade 6 in each of the dimensions. Gender differences were also observed: boys outperformed girls in decomposition and algorithms, which may reflect stronger procedural and spatial reasoning, while girls showed higher performance in pattern recognition and evaluation, which may stem from greater attention to detail, verbal reasoning, or persistence with complex tasks. These results provide preliminary evidence supporting the construct validity of CT as a developmental competency. Instructional strategies could be tailored to support and balance these differing cognitive strengths across genders.

While this study provides evidence for the development of CT, it is limited by its cross-sectional design. In the future, research could expand to longitudinal approaches to better capture individual growth trajectories.

Keywords: Computational thinking, MIRT, elementary education, cognitive development, gender differences.