Meta-analysis finds AI gives elementary students a small maths boost

The review also tested eight moderating variables. Mathematics learning topic and grade level significantly changed how much AI helped, while research type, research design, sample size, intervention duration, AI type and organization did not show significant moderating effects.

The findings suggest AI tools may modestly support maths learning in primary schools, but the size of the benefit depends on what is taught and to which grade.

The study pooled evidence from 21 empirical studies with 30 independent samples published between January 2000 and June 2022, using a random-effects meta-analysis to estimate AI's effect on elementary students' mathematics achievement.

The pooled effect was 0.351, which the authors classed as small. They then examined three research characteristic variables (research type, research design and sample size) and five opportunity-to-learn variables (maths topic, intervention duration, AI type, grade level and organization).

Only maths learning topic and grade level significantly moderated the effect, so the benefit differed across topics and grades. The authors drew practical and research implications from these results.