A Comparison of Model Selection Metrics in DataShop.
John C. Stamper, Kenneth R. Koedinger, Elizabeth A. McLaughlin · 2013
Variations of cognitive models drive many instructional decisions that intelligent tutoring systems currently make. A better knowledge component model will yield better instruction, but how do we identify better cognitive models? One answer has been to create a latent variable version of a cognitive model or a so-called knowledge component (KC) model, then compare different models by how well they predict student performance data. In this research we analyze 1,943 proposed KC models that exist in DataShop