Computational thinking across the learning continuum: a systematic review of integration strategies, pedagogical practices, and outcomes in primary, secondary, and higher education

Abstract

Computational thinking (CT) has emerged as a foundational competency for the twenty-first century, yet evidence on how CT integration strategies, pedagogical practices, and learning outcomes compare across educational levels remains fragmented. This systematic review aimed to (1) map patterns of CT integration across early childhood, K-12, and higher education, (2) compare pedagogical approaches and associated learning outcomes, and (3) identify gaps in the existing literature. A systematic literature review was conducted using six databases (ERIC, IEEE Xplore, Scopus, Web of Science, Education Source, and Google Scholar), selecting 86 studies published between 2010 and 2024 that satisfied defined inclusion criteria. The review reveals that CT integration is best characterized as an “adolescent discipline”—expanding in scope but lacking theoretical and methodological consolidation. Key findings include: (a) substantial heterogeneity in CT operationalization across studies, with 14 different assessment instruments identified; (b) a critical distinction between subjective CT self-efficacy (which often declines following instruction) and objective CT performance (which consistently improves), suggesting a Dunning–Kruger-like effect; (c) non-linear developmental trajectories, with gender differences fluctuating across grade levels rather than following stable patterns; and (d) persistent discontinuities in CT skill transfer across educational transitions, as only 23% of reviewed studies addressed vertical curriculum alignment. Early childhood environments are dominated by “unplugged first” approaches that emphasize play-based learning, with a moderate effect size advantage over plugged-first approaches for concept acquisition (g = 0.48, Chen et al. (Chen et al. in Int J STEM Educ 10:34, 2023)). Secondary education benefits from cross-disciplinary modules, though effect sizes vary substantially (η2 = 0.31–0.93) depending on intervention modality. At the higher education level, research-based CT laboratories have shown positive outcomes, but 61% of studies in this category lacked control groups, limiting causal inference. The synthesis underscores the need for curriculum designers, educators, policymakers, and researchers to adopt a cohesive strategy for CT integration that supports inclusivity, addresses transition discontinuities, and mitigates educational disparities.