Text reproducers copied nearly all of their final text from ChatGPT with long verbatim sequences, while text integrators showed fragment-level micro-uptake¶

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

Subclaims¶

q3 i? Cluster 1 (8 participants) showed the highest proportion taken over (0.997) and long n-grams, whereas Cluster 2 (72 participants) showed the lowest proportion (0.55) with only short overlaps. → Teresa Luther · Kate Belcher · Joachim Kimmerle · Detmar Meurers · Ulrike Cress 2026

Evidence¶

Teresa Luther · Kate Belcher · Joachim Kimmerle · Detmar Meurers · Ulrike Cress 2026¶

Teresa Luther · Kate Belcher · Joachim Kimmerle · Detmar Meurers · Ulrike Cress (2026). Collaborative knowledge construction with generative AI: Exploring argumentative co-writing processes through n-gram and cluster analysis. Intern. J. Comput.-Support. Collab. Learn. https://doi.org/10.1007/s11412-026-09470-1

q3 · i? · associational · r2

Cluster characterization from feature means in the n=131 dataset. Text reproducers reused very long verbatim sequences (456_781_grams mean 0.89), while text integrators had all long bins zero and the lowest takeover proportion, "indicating that nearly all of their final text was composed of copied content" for reproducers only.

"Participants in this cluster also showed the highest total_tokens_taken_over (0.77) and proportion_taken_over (0.997), indicating that nearly all of their final text was composed of copied content."

Discussion¶

Related Claims¶

• AI-supported argumentative writing yields three distinct uptake clusters (text reproducers, text integrators, text reconstructors) based on recurring n-gram overlap — a broader claim this one bears on
• Metacognitive awareness shapes whether learners use ChatGPT reflectively or copy shallowly, and metacognitive support sustains self-regulated learning — related
• Engage-to-Unlock participants incorporated less verbatim AI-response text into final essays than Standard Chatbot participants — related
• Affinity for technology interaction differed across clusters, with text reproducers highest and integrators scoring higher than reconstructors (d = 0.59) — related
• Current ChatGPT users were overrepresented among text reconstructors and underrepresented among text integrators (Cramér's V = 0.26) — related
• Clusters did not differ significantly in the number of prompts sent to ChatGPT — related
• Cluster membership showed a strong multivariate effect on standard text-similarity metrics (Pillai's trace = 0.85, F(12, 248) = 15.31, p < 0.001) — related