Feedback type shifts engagement by question type: educator feedback yields more MCQ submissions while AI feedback trends toward more open-ended revisions, with no overall difference¶

Claim · All claims
Evidence · 1 study (3 entries) · causal r2 · q3 · i0 negligible
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Overall active-learning actions did not differ significantly between conditions (U = 4075.5, p = 0.0738). → Chloe Qianhui Zhao 2026 q3 i? For multiple-choice items, baseline participants submitted more attempts than AI-condition participants (U = 5660.5, p = 0.02); for open-ended items the AI condition showed a trend toward more submissions (U = 4060.5, p = 0.066). → Chloe Qianhui Zhao 2026 (2) q3 i0 Total learning time did not differ significantly between groups (Cohen's d = −0.062). → Chloe Qianhui Zhao 2026 (3)

Evidence¶

Chloe Qianhui Zhao 2026¶

Chloe Qianhui Zhao, Jie Cao, Jionghao Lin, and Kenneth R. Koedinger. (2026). LLM-based Multimodal Feedback Produces Equivalent Learning and Better Student Perceptions than Educator Feedback. LAK 2026. https://arxiv.org/abs/2601.15280 doi:10.48550/arxiv.2601.15280

q3 · i? · causal · r2

Log-data analysis of submit/update actions during the learning-by-doing stage of the randomized experiment (RQ2). Overall means were 21.39 (SD 9.05) vs. 21.94 (SD 5.78) actions; no effect size printed.

"A Mann–Whitney U test was conducted (Table 1) and showed no significant overall difference between Original Educator Feedback and AI Multimodal Feedback (𝑈= 4075.5,𝑝= 0.0738), indicating comparable aggregate levels of active-learning behavior across conditions."

Chloe Qianhui Zhao 2026 (2)¶

Chloe Qianhui Zhao, Jie Cao, Jionghao Lin, and Kenneth R. Koedinger. (2026). LLM-based Multimodal Feedback Produces Equivalent Learning and Better Student Perceptions than Educator Feedback. LAK 2026. https://arxiv.org/abs/2601.15280

q3 · i? · causal · r2

Question-type breakdown of submission counts in RQ2 (Table 1: MCQ means 13.78 vs. 11.92; OEQ 7.61 vs. 8.44). The OEQ difference was a non-significant trend; equivalence was not tested.

"for multiple-choice items, partici- pants in the baseline condition submitted more attempts than those in the AI condition (𝑈= 5660.5,𝑝= 0.02*), consistent with greater trial-and-error under human feedback; for open-ended items, the AI condition showed a trend toward more submissions (𝑈= 4060.5, p = 0.066)."

Chloe Qianhui Zhao 2026 (3)¶

Chloe Qianhui Zhao, Jie Cao, Jionghao Lin, and Kenneth R. Koedinger. (2026). LLM-based Multimodal Feedback Produces Equivalent Learning and Better Student Perceptions than Educator Feedback. LAK 2026. https://arxiv.org/abs/2601.15280

q3 · i0 · causal · r2

Time-on-task comparison in RQ2: median 21.13 minutes (baseline) vs. 25.44 minutes (AI). The printed effect size d = −0.062 is negligible; the difference was not significant.

"An independent-samples t-test indicated no significant difference in learning time between groups (𝑡=−0.413,𝑝=0.680,𝐶𝑜ℎ𝑒𝑛′𝑠𝑑=−0.062)."

Discussion¶

Related Claims¶

• AI multimodal feedback produces post-test learning gains equivalent to educator feedback after controlling for pre-test performance — related
• The AI feedback system returns responses in under a second for MCQs and about six seconds for open-ended questions, and its reusable prompt is roughly 2.77 times more time-efficient to prepare than equivalent educator text — related
• Feedback improves learning by a medium amount on average, but its effect varies widely with the information it carries, and a substantial share of feedback effects are negative — a broader claim this one bears on
• Layered AI feedback led to significantly more tasks requiring three or more submissions, while total submissions, learning time and edit distance did not differ — related
• Immediate post-feedback improvement is near zero on average and does not differ significantly across feedback conditions — related