Differentiated-AI EAP versions showed limited passage-level structural separation, with automated complexity metrics failing to classify intended proficiency labels¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Across six automated complexity indicators, differentiated versions yielded η2p = 0.043 and leave-one-out classification accuracy of 11.1%, below the 33.3% balanced-task benchmark. → Gao X 2026

Evidence¶

Gao X 2026¶

Gao X (2026) From unified to differentiated materials: generative AI–supported adaptation of EAP reading materials. Front. Psychol. 17:1887565. https://doi.org/10.3389/fpsyg.2026.1887565

q2 · i? · causal · r2

Automated structural-complexity analysis treating each material version as the unit of analysis, using six z-standardized indicators and leave-one-out cross-validated linear discriminant analysis. The result shows "limited structural separation" — the differentiated versions did not form a stable passage-level structural gradient.

"The six selected indicators yieldedη2 p = 0.043, and the leave-one-out classification accuracy was 11.1%, compared with the 33.3% benchmark for the balanced three-label task."

Discussion¶

Related Claims¶

• Expert EAP reviewers rated GenAI-generated differentiated materials favorably for academic fidelity, proficiency appropriateness, and teachability — related
• The Proficiency × Material interaction on reading comprehension was significant, indicating heterogeneous material effects across proficiency strata — related
• Differentiated-AI materials exceeded unified-AI materials on reading comprehension within each proficiency stratum, most strongly for high-proficiency learners (d = 1.40) — related
• Pre-study verified that the three learning units differ in perceived complexity in the intended low, medium, and high ordering — related