AI-supported instruction has a moderate positive effect on student achievement in STEM education (g = 0.670 under a random effects model)¶

Claim · All claims
Evidence · 1 study (2 entries) · quant-synthesis r3 · q4 · i2 medium
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q4 i2 Across 35 experimental and quasi-experimental studies, AI-supported instructional interventions produced a statistically significant positive overall effect on STEM student achievement (g = 0.670, 95% CI [0.491, 0.848]). → Doğan Yunus 2026 q4 i? The included studies showed a significant and high level of heterogeneity (Q = 199.76, I2 = 82.98% under the fixed-effect model), indicating the effect is context-sensitive rather than uniform. → Doğan Yunus 2026 (2)

Evidence¶

Doğan Yunus 2026¶

Doğan Yunus, Kılıç Zeynep, Kalınkara Yusuf, Talan Tarık. (2026). The Impact of Artificial Intelligence-Supported Instruction on Student Learning in STEM: A Systematic Review and Meta-Analysis. Journal of Intelligence. https://doi.org/10.3390/jintelligence14060109

q4 · i2 · quant-synthesis · r3

Random effects model meta-analysis of 35 experimental and quasi-experimental studies of AI-supported instruction in STEM, with Hedges' g calculated from post-test means, standard deviations, and sample sizes. The article reports "g = 0.670 (95% CI [0.491, 0.848])" with Z = 7.35, p < 0.001.

"Under this model, the effect size was found to be g = 0.670 (95% CI [0.491, 0.848]), with Z = 7.35 and p < 0.001, while the between-study variance was accounted for more appropriately under the REM (I2 = 15.75%)."

Doğan Yunus 2026 (2)¶

Doğan Yunus, Kılıç Zeynep, Kalınkara Yusuf, Talan Tarık. (2026). The Impact of Artificial Intelligence-Supported Instruction on Student Learning in STEM: A Systematic Review and Meta-Analysis. Journal of Intelligence. https://doi.org/10.3390/jintelligence14060109

q4 · i? · quant-synthesis · r3

Fixed-effect analysis of the same 35 studies quantified between-study inconsistency with the Q test and I2 statistic, showing "a significant and high level of heterogeneity", which the authors say means effect sizes do not display a homogeneous distribution.

"The meta-analytic findings indicate that there is a significant and high level of heterogeneity among the included studies (Q = 199.76, I2 = 82.98%)."

Studies bearing on this claim¶

Does building AI tools into instruction raise achievement over teaching the same content without them, and in which settings? This table gathers the studies behind this claim and the 10 narrower claims that bear on it: 9 distinct studies (3 causal, 3 design, 2 review, 1 quant-synthesis), q2–q4, 4 of 9 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Related Claims¶

• Gen-AI has a moderate positive overall effect on students' higher-order thinking (g = 0.609) across 29 experimental and quasi-experimental studies — related
• ChatGPT-enhanced instruction produced significantly higher overall post-test achievement in Grade 12 Quantum Theory than traditional teaching, with a large effect size (η² > 0.14) — a narrower finding that bears on this claim
• ITS produce significant positive effects at all grade levels and in almost all subject domains evaluated, with exceptions such as chemistry — a narrower finding that bears on this claim
• AI-supported instruction is most effective at the high school level, with between-level differences statistically significant — a narrower finding that bears on this claim
• Interventions lasting one to two months show the largest effect, and effects do not increase consistently with longer duration — a narrower finding that bears on this claim
• AI-enabled personalized STEM education yields a statistically significant overall effect of g = 0.455, which its authors class as small, compared with traditional non-AI instruction in K-12 schools — related
• Initial meta-analyses indicate the effect of AI chatbots and generative AI on learning outcomes is lower in secondary education than in higher education — related
• Access to a structured conversational AI tutor increased post-test content scores over a no-AI holdout with ambient consumer AI available — a narrower finding that bears on this claim
• Students in the AI-integrated database course showed large, statistically significant pre/post gains on seven identical database assessment questions — a narrower finding that bears on this claim
• Students' overall AI knowledge improved significantly after four integrated AI-biology lessons — a narrower finding that bears on this claim
• The authors of one course-based design study conclude, without printed measures, that structured AI integration can enhance students' digital literacy, content retention and critical thinking — a narrower finding that bears on this claim
• Prior meta-analytic evidence indicates chatbot technology has a substantial positive effect on overall learning outcomes and knowledge retention, but not on critical thinking, engagement, or motivation — related
• Learning with intelligent tutoring systems yields a moderate, statistically significant overall advantage in achievement over non-ITS modes of instruction — related
• Learning-by-design shows medium- to large-sized positive effects on student achievement in K-12 STEM education, per meta-analysis cited by the review