Removing teacher-in-the-loop review or the curriculum checker significantly lowers teacher satisfaction, and replacing ADDIE orchestration with unstructured prompting yields the lowest satisfaction and longest preparation time¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Removing the teacher-in-the-loop review stage causes a statistically significant drop in teacher satisfaction (p<0.05, Wilcoxon signed-rank test, n=18), despite marginally reducing preparation time. → Thang Doan Viet 2026 q2 i? Replacing full ADDIE orchestration with unstructured prompting (Generic LLM workflow) substantially increases preparation time (90–120 minutes) and achieves the lowest satisfaction (3.6±0.8, p<0.01). → Thang Doan Viet 2026 (2)

Evidence¶

Thang Doan Viet 2026¶

Thang Doan Viet, Anh Nguyen Hoang, Tinh Luong Son, Anh Hoang Thi Ngoc, Huyen Giang Thi Thu, Tai Le Quy. (2026). ConnectED: A Curriculum-Aligned AI System for Vietnamese Instructional Lesson Planning and Student Learning. arXiv. https://arxiv.org/abs/2607.28647 doi:10.48550/arxiv.2607.28647

q2 · i? · design · r2

System-level ablation (Table 4) with 18 teachers, comparing variants against the full pipeline using Wilcoxon signed-rank tests. Satisfaction fell from 4.7±0.4 to 3.9±0.6 without teacher-in-the-loop review; the authors conclude teachers "value control over AI outputs even at the cost of additional time".

"Removing the teacher-in-the-loop review stage reduces preparation time marginally but causes a statistically significant drop in teacher satisfaction (p<0.05, Wilcoxon signed-rank test [48])"

Thang Doan Viet 2026 (2)¶

Thang Doan Viet, Anh Nguyen Hoang, Tinh Luong Son, Anh Hoang Thi Ngoc, Huyen Giang Thi Thu, Tai Le Quy. (2026). ConnectED: A Curriculum-Aligned AI System for Vietnamese Instructional Lesson Planning and Student Learning. arXiv. https://arxiv.org/abs/2607.28647

q2 · i? · design · r2

System-level ablation (Table 4): the Generic LLM workflow took 90–120 minutes with satisfaction 3.6±0.8 (p<0.01 vs. full pipeline), the lowest of all variants. The authors state this demonstrates ADDIE orchestration "contributes independent value beyond model capability alone".

"Replacing the full ADDIE orchestration layer with unstructured prompting of the same VietEduQwen model (theGeneric LLM workflowrow)substantially increases preparation time and achieves the lowest satisfaction score"

Discussion¶

Related Claims¶

• Teachers and students report strong satisfaction with ConnectED, with 94% of teachers willing to reuse it — related
• ConnectED reduces 45-minute lesson preparation time to approximately 15–20 minutes in-system, versus 3–4 hours for manual preparation — related
• Implementing a student-made glossary significantly lowered first-year engineering students' satisfaction with the M-Tutor tutorial — related