NIG-head predictive uncertainty is dominated by epistemic uncertainty and shows negligible correlation with the absolute engagement score (Pearson r 0.016 aleatoric, 0.025 epistemic)¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · no dated study recorded · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Across engagement-level bins, total uncertainty from the NIG regression head is heavily dominated by epistemic uncertainty, and raw engagement scores correlate negligibly with both aleatoric (r=0.016) and epistemic (r=0.025) uncertainty. → uncert-1

Evidence¶

uncert-1¶

Alperen Kantarcı, Visvanathan Ramesh, and Gemma Roig. 2026. Mind the Student: Behavioral and Contextual Cues for Automated Engagement Prediction in Online Learning. ICMI '26, Napoli, Italy. https://doi.org/10.1145/3776574.3832485

q2 · i? · design · r2

Uncertainty decomposition analysis (Table 4) grouping test samples into five equal-width bins on the 1–5 engagement scale. The printed correlations are r 0.016 (aleatoric) and r 0.025 (epistemic).

"Pearson correlation coefficients show negligible linear relationships between raw engagement scores and both aleatoric and epistemic uncertainties, indicating that model confidence is driven by factors other than the absolute magnitude of the engagement score itself."

Discussion¶

Related Claims¶

• The hybrid confidence measure with aleatoric uncertainty yields the best calibration (Brier 0.138, ECE 0.044, MCE 0.100) among compared methods — related
• Latent (token-probability) confidence is the weakest single-source confidence signal for LLM-based ASAG, with the lowest AUROC and worst calibration — related
• A hybrid confidence measure combining model-based signals with dataset-derived aleatoric uncertainty achieves the highest AUROC and strongest accuracy gains in selective grading of LLM-graded short answers — related