Culturally responsive and equity-driven differentiated STEM pedagogy and assessment in diverse classrooms in Ghana

Persistent assessment inequities in linguistically and culturally diverse classrooms necessitate contextually grounded frameworks that advance both academic excellence and educational justice. The study developed, implemented, and evaluated a culturally responsive, equity-driven, and differentiated pedagogical-assessment framework for STEM learning in multilingual and socioeconomically diverse classrooms in Ghana. A convergent parallel mixed-methods design integrated qualitative data (interviews, observations, focus groups) with quantitative measures (pre-post-tests, subscale gains, ANCOVA, and Bayesian analyses). The study involved 210 participants 24 STEM teachers, 6 headteachers, and 180 Primary 4-6 pupils from six public basic schools in Ghana. Quantitative findings revealed statistically significant learning gains across schools and domains. Mean post-test scores increased by 13.96 points (29.12%), with a large overall effect size (Cohen’s d = 1.68; Hedges’ g = 1.89; η ² = 0.79; p < 0.001), indicating strong practical and educational significance. Subscale analyses showed consistent and convergent gains across all framework dimensions, including conceptual mastery (d = 1.72), application of skills (d = 1.67), inclusive strategy use (d = 1.62), opportunity-to-learn (d = 1.59), tiered instruction (d = 1.66), flexible grouping (d = 1.64), linguistic alignment (d = 1.70), and cultural asset integration ( d = 1.69), all demonstrating large effects (d > 1.5). ANCOVA results confirmed significant effects of the intervention across all domains while controlling for pre-test performance, including inquiry-based STEM practices (F = 26.03, p < .001, η²ₚ = 0.10), integrated STEM tasks (F = 19.38, p < .001, η²ₚ = .077), equity-driven instructional practices (η²ₚ = 0.05–0.07), differentiated pedagogical assessment (η²ₚ = 0.04–0.05), and culturally responsive assessment practices (η²ₚ = 0.05–0.06). The pre-test covariate remained significant ( F = 88.21, p < .001, η² = 0.328), indicating the importance of baseline achievement while confirming robust intervention effects Bayesian analyses indicated strong evidential support (posterior probabilities > 0.98; Bayes Factors = 28.1–36.4). Qualitative findings corroborated these results, highlighting increased teacher capacity for culturally relevant task design and improved student engagement and identity affirmation. The study offers empirically grounded guidance for decolonizing STEM assessment in multilingual and underserved contexts.