Ablations show every feature group contributes to skills forecasting, while recent engagement activity alone nearly suffices for minutes¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Removing any single feature group yields small MAE losses; for skills, dropping engagement activity causes the largest loss (>1%) and other groups 0.6-0.8% each, while for minutes all removals cause losses under 1%. → Eric S. Qiu 2026

Evidence¶

Eric S. Qiu 2026¶

Eric S. Qiu, Danielle R. Thomas, Boyuan Guo, Vincent Aleven, Conrad Borchers. (2026). From Heuristics to Analytics: Forecasting Effort and Progress in Online Learning. https://arxiv.org/abs/2605.12788 doi:10.48550/arxiv.2605.12788

q2 · i? · associational · r2

Ablation study (Table 5) retraining models with each feature group removed and comparing MAE to the full model with bootstrap significance tests; the article reports that for skills "every group contributes" with engagement activity removal the largest loss.

"Forskills, every group contributes: dropping engagement activity features yields the largest loss (>1%), but AFM, gaps, and prior achievement features also provide meaningful differences (0.6–0.8% each)."

Discussion¶

Related Claims¶

• Feature-based ML models reduce next-week forecasting MAE by 22-33% over heuristic baselines for weekly minutes practiced and new skills mastered — related
• Effort forecasting is driven by recent engagement activity while progress forecasting depends more on learner-centered signals such as ability, difficulty, and consistency — related