Gender Disparity in Computational Thinking Pedagogy and Assessment: A Three-Level Meta-Analysis

Abstract Gender disparity is a well-recognized issue in computational thinking (CT) education, yet few meta-analyses have examined how specific pedagogical and assessment contexts influence this gap. This study addresses that limitation by synthesizing 53 empirical studies, comprising 100 effect sizes and 15,454 participants, to explore the extent and moderators of gender differences in CT education. The analysis reveals an overall effect size of g = 0.106 (95% CI [0.024, 0.188], p < 0.05), indicating a very small but statistically significant gender disparity favoring males. Among the three groups of moderators examined, neither general study characteristics (publication type, geographical region, and educational level) nor CT assessment contexts (instrument and measured learning outcome) significantly influenced effect sizes. However, pedagogical strategies moderated gender disparities: mixed and plugged approaches, which integrate technologies, were associated with larger gaps favoring boys, while unplugged approaches tended to reduce or even reverse the disparity, benefiting girls in some cases. In terms of assessment, gender disparities were nonsignificant for CT concepts, but became significant when evaluating authentic practices (e.g., programming tasks) and identity-related perspectives (e.g., motivation, learning interest, and self-efficacy). The findings offer practical implications for advancing educational equity in CT education. Early interventions in K-12 settings, especially those targeting at CT practices and perspectives, are critical to prevent disparities from becoming entrenched. Unplugged activities can help build foundational understanding and confidence, especially for girls. Gradually introducing digital and AI tools within supportive environments, such as culturally relevant scenarios, may reduce technology-related anxiety and promote more inclusive learning experiences.