Scaffolded role-play for cross-text integrative writing in elementary Chinese: a teacher-vetted AI-drafted materials...

The Chinese Language Curriculum Standards for Compulsory Education (2022 Edition) emphasize multi-text learning through learning task clusters, yet classroom practice often continues to treat linked texts as separate reading lessons. This study examined an integrated instructional package for linked-text reading and cross-text writing in Grade 3 Chinese.

A posttest-only classroom quasi-experimental design was adopted. Participants were two intact Grade 3 classes from a public primary school in Tianjin, China, with a valid analytic sample of 54 students. Both classes studied the same four thematic units with identical texts, instructional time, teacher, learning objectives, and posttest writing task. The intervention class received an instructional package combining teacher-vetted AI-drafted materials, scaffolded role-play, probing-question chains, evidence checkpoints, structured turn-taking, and cross-text writing. The comparison class received conventional reading instruction. Outcomes included a cross-text integrative writing task and a post-intervention learning-experience questionnaire.

The intervention class achieved a higher total writing score than the comparison class, Welch's t(51.8) = 2.98, p = 0.004, d = 0.81. Supplementary ANCOVA controlling for midterm Chinese achievement showed a similar between-class difference, F(1, 51) = 8.72, p = 0.005, partial η² = 0.146. Differences were concentrated in thematic integration, textual evidence, and comparison and transfer, whereas expressive structure showed no reliable difference. Students in the intervention class also reported higher engagement, evidence awareness, perceived integration and transfer, strategic confidence, and acceptability, together with lower perceived task difficulty and pressure.

The findings provide preliminary classroom evidence that an integrated instructional package may support evidence-aware cross-text writing and improve students' learning experiences. Further multi-class studies incorporating pretests, delayed assessments, and classroom-process data are needed to examine the robustness and persistence of these effects.