AI-Enhanced transformative pedagogy in secondary mathematics: A quasi-experimental investigation of problem-solving, stem readiness, and academic achievement

Source abstract

Evidence on artificial intelligence (AI) in transformative mathematics teaching remains limited, particularly regarding higher-order problem solving and preparation for Science, Technology, Engineering, and Mathematics (STEM) study and careers. This study examined the association of AI-enhanced transformative pedagogy with secondary students’ mathematical problem-solving, STEM readiness, mathematics achievement, and six-week retention, while exploring students’ learning experiences. Participants were 142 Grade 10 students from four secondary schools in Türkiye. Two intact classrooms (n = 72) received the AI-enhanced intervention, while two (n = 70) continued traditional instruction. A non-equivalent control-group, pretest–posttest–delayed-posttest design was implemented over 14 weeks, comprising a 12-week intervention and six-week maintenance interval. Quantitative analyses used multilevel models with cluster-robust standard errors, Bayesian sensitivity analyses, and randomization inference. Qualitative evidence comprised semi-structured interviews with 24 experimental-group students and 72 reflection-journal entries. The experimental group demonstrated higher problem-solving (d = 0.87, 95% CI [0.53, 1.21]), STEM readiness (d = 0.79, 95% CI [0.46, 1.12]), mathematics achievement (d = 0.94, 95% CI [0.60, 1.28]), and six-week retention (d = 0.68, 95% CI [0.36, 1.00]). Qualitative findings indicated contingent support, cognitive challenge, individualized pacing, and stronger mathematics–STEM connections. Findings are preliminary and hypothesis-generating rather than causal. Limitations include small classroom numbers, non-random assignment, bundled intervention components, possible teacher effects, and self-reported STEM-readiness measures. Larger randomized factorial studies are needed to isolate AI’s contribution.