Students using the eliciting agent showed JoL ratings correlated with system learning metrics, strongest for mastery¶

Claim · All claims
Evidence · 1 study · causal r2 · q2 · i3 large
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i3 In the elicit condition, JoL ratings correlated positively with system confidence (r = 0.287), progress (r = 0.326) and mastery (r = 0.408), all statistically significant. → Graf 2026

Evidence¶

Graf 2026¶

Graf, L., Bassner, P., Anzinger, M., Dietrich, F., Krusche, S., & Poquet, O. (2026). Interactive learning dashboards: rethinking learning visualisations as engagement tools. Education and Information Technologies. https://doi.org/10.1007/s10639-026-14082-1

q2 · i3 · causal · r2

Across 104 JoLs in the elicit condition of the five-week case study, JoL accuracy (correlation of student to system ratings) was positive for confidence and progress; the article adds "The strongest correlation was observed for mastery, with r = 0.408 (p < 0.001)".

"In the elicit condition (n = 104 JoLs), students appeared more accurate in their judgement. For confidence, the correlation was r = 0.287 (p = 0.003). For progress, the correlation was r = 0.326 (p = 0.001)."

Discussion¶

Related Claims¶

• Self Assessment Accuracy Is Low Without Training — the general claim this finding bears on, with the studies behind it
• Tell and no-agent conditions showed no significant JoL–confidence correlations, with mixed results for progress and mastery — related
• JoL–mastery accuracy trajectories over the five weeks rose in the elicit condition, declined in the none condition, and were mixed in the tell condition — a narrower finding that bears on this claim
• Most students submitted JoLs out of curiosity about the system rating, often without attending to the LA graphs, and reported behaviour change more from JoL than from the agent — related
• Students' JoL ratings became more consistent over time and evolved from under-rating to over-rating to accurate judgement — related
• In a real programming course, students with an eliciting LLM agent on a learning dashboard wrote more reflection messages than students with a telling agent — related