Individualized dialogic practice delivered through AI supports first graders' question-asking

Children's question-asking is an important means of information seeking and learning, yet little is known about how it changes during and following sustained individualized dialogic practice. In a two-group controlled trial, 190 first-grade children (ages 6–7) were allocated to one-to-one individualized dialogic practice delivered by an AI conversational agent or to instructor-led whole-class discussion, matched in content, materials, and session duration, across seven weekly sessions. At pretest and posttest, children completed assessments without instructional or AI-generated support. Children in the individualized-practice condition demonstrated substantially higher pretest-adjusted posttest question frequency (unadjusted posttest d = 1.23). Among children who produced at least one valid question at both assessments, those in the individualized-practice condition also demonstrated higher pretest-adjusted mean cognitive complexity at posttest (unadjusted posttest d = 0.53). An exploratory, post hoc analysis including all 190 children further indicated that children in the individualized-practice condition produced explanation-or-higher questions at a higher adjusted rate (adjusted count ratio = 3.1, 95% CI [2.5, 4.0]). Within the individualized-practice condition, question frequency showed marked phase-specific growth across practice sessions, whereas mean per-question quality increased only gradually, providing no support for a quantity-before-quality ordering. Because interaction format, per-child practice opportunities, feedback immediacy, facilitator identity, and delivery medium covaried in this design, the findings support sustained individualized dialogic practice as implemented here, rather than AI per se.