Primary mathematics teachers rate post-generation control significantly higher than pre- and mid-generation control on perceived numerical correctness of generated visuals¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Teachers' ratings of whether the image matched the number of objects differed reliably across paradigms, with post-generation control rated significantly higher than both pre-generation and mid-generation control. → Zhengxu Li 2026

Evidence¶

Zhengxu Li 2026¶

Zhengxu Li, Junling Wang, and April Yi Wang. (2026). When Should Teachers Control AI Generation for Mathematics Visuals? arXiv preprint. https://arxiv.org/abs/2605.10672 doi:10.48550/arxiv.2605.10672

q3 · i? · causal · r2

Within-subject questionnaire results (Q6, 7-point Likert) from a study with 24 primary mathematics teachers who used all three paradigms. Friedman test showed reliable differences (Z = 24.83, p < .01); post-hoc tests showed "post-gen was rated significantly higher than pre-gen" and mid-gen. No effect size printed.

"Post-hoc Wilcoxon signed-rank tests revealed a clear ordering: post-gen was rated significantly higher than pre-gen ( 𝑍= 0.0, 𝑝<. 01) and mid-gen (𝑍= 3.5, 𝑝<. 01), indicating that teachers perceived post-generation control as most effective for preserving numerical correctness"

Discussion¶

Related Claims¶

• Manually verified numerical accuracy of generated visuals is lowest under pre-generation control (12.2%), intermediate under mid-generation control (59.83%), and highest under post-generation control (97.85%) — related
• No single control-timing paradigm dominates across subjective measures: eight of ten questionnaire items show no significant differences between paradigms — related
• Teachers' tool preferences are stage-dependent, with pre-gen valued for ideation, mid-gen for structuring, post-gen for finalization, motivating hybrid workflows — related