Individual trajectories suggest heavy LLM use focused on support other than full answer generation can coexist with improving scores, whereas reliance on LLM-generated summaries was associated with declining scores¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r1 · q1
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q1 i? S61, who used all categories except Generating Answers, showed steadily increasing midterm scores (69% to 72% to 79%). → Minju Park 2026 q1 i? S2, who used only Summarizing, showed consistently declining midterm scores (65% to 48% to 38%). → Minju Park 2026 (2)

Evidence¶

Minju Park 2026¶

Minju Park, Ivan Orozco Vasquez, and Cristina Conati. (2026). Characterizing Students' LLM Usage Behaviors and Their Association with Learning in Critical Thinking Tasks. https://arxiv.org/abs/2605.04534 doi:10.48550/arxiv.2605.04534

q1 · i? · associational · r1

Single-student case within the High-Reliance group of the two-term course dataset. S61 used all usage categories except Generating Answers, and the article reports the scores "69% to 72% to 79%" as illustrating that extensive non-generation LLM use can coexist with improved performance.

"Notably, S61's midterm scores steadily increased across the term (69% to 72% to 79%), illustrating that extensive LLM use—when focused on forms of support other than full answer generation—can coexist with, and may even accompany, improved performance."

Minju Park 2026 (2)¶

Minju Park, Ivan Orozco Vasquez, and Cristina Conati. (2026). Characterizing Students' LLM Usage Behaviors and Their Association with Learning in Critical Thinking Tasks. https://arxiv.org/abs/2605.04534

q1 · i? · associational · r1

Single-student case within the High-Reliance group. S2 used only the Summarizing category, and midterm scores "declined consistently—from 65% to 48% to 38%"; the authors suggest S2 increasingly relied on LLM-generated summaries rather than engaging with the full readings.

"S2 used only Summarizing(S), and their performance provides a clear illustration of the potential consequences of this pattern: their midterm scores declined consistently—from 65% to 48% to 38%."

Discussion¶

Related Claims¶

• Among LLM users, Low-Reliance students (use in under half of submissions) consistently outperformed High-Reliance students by roughly 7-8% across all three midterms — related
• Students whose LLM use was mostly student-driven support scored higher on Midterm 1 than LLM-driven users (about 10%), but the two groups converged on later midterms — related
• Positive LLM tutor perceptions do not necessarily translate to learning gains, creating an over-reliance risk — related
• Students who used LLMs on assignments scored significantly lower than non-users on the first midterm (large effect size), but this gap narrowed and was non-significant on the later midterms — related