Product measures distinguish LLM-formulated text from classroom-formulated text regardless of how it was entered¶

Claim · All claims
Evidence · 1 study · design 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? Responses formulated by an LLM were harder to read and used longer words than classroom responses, regardless of entry mode. → Divyansh Chandarana 2026

Evidence¶

Divyansh Chandarana 2026¶

Divyansh Chandarana, Sandipan De, Vivek Gupta. (2026). DRAFTTRACE: A Multi-View Analytics Environment for AI-Integrated Writing. https://arxiv.org/abs/2609.36544 doi:10.48550/arxiv.2609.36544

q2 · i? · design · r2

Case study comparing 81 classroom, 81 automated, and 11 copy-typed writing sessions in a graduate NLP course. Descriptive medians show classroom responses were easier to read (Flesch 52.8) than LLM-formulated responses (38.4 automated, 39.9 copy-typed); "median word length rose from 4.67 to 5.18 and 5.40 characters." No inferential test is reported.

"Median Flesch Reading Ease was 52.8 for classroom responses, compared with 38.4 for automated and 39.9 for copy-typed responses, and me-dian word length rose from 4.67 to 5.18 and 5.40 characters."

Discussion¶

Related Claims¶

• Identifying copy-typed sessions requires combining product and process views, since neither view is sufficient alone — related
• Process measures distinguish how text was entered: classroom writers revised most, left the editor often, and wrote actively for 77% of the session — related
• In the random forest, word count was the most important readability feature while traditional readability formulas ranked near the bottom — related
• Revising texts to improve readability-formula scores often produces nonexistent or insignificant comprehension gains — related