Learner models with population-level parameters systematically misestimate the trial of mastery for most students whose learning rates vary¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2022 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? In a simulation, a population-parameter PFA model over- or under-predicted when mastery occurred for most simulated students, with a median absolute error of 9 trials. → Eglington 2022

Evidence¶

Eglington 2022¶

Eglington, L.G., Pavlik, P.I. (2022). How to Optimize Student Learning Using Student Models That Adapt Rapidly to Individual Differences. International Journal of Artificial Intelligence in Education. https://doi.org/10.1007/s40593-022-00296-0

q2 · i? · design · r2

Simulation of 1000 students learning one KC across 100 trials, with data generated by PFAT (student-level parameters) and predictions made by PFAP (population-level parameters). The median absolute distance between true and estimated mastery trial was "9 trials". No effect size printed.

"Figure 2b shows that this leads to substantial error with the median absolute distance between true and estimated trial of mastery being 9 trials. The median simulated student required 19 (MAD = 13.3) trials to achieve mastery."

Discussion¶

Related Claims¶

• Adding a Prior Error Valence (PEV) feature to a population-parameter model reduces systematic mastery-timing error in simulation — related
• EPFA estimates fewer trials to mastery for above-average students and more trials for below-average students relative to standard PFA — related
• Standard knowledge tracing models in use do not allow individual learning rates or individualized initial knowledge — a broader claim this one bears on