The Complex Dynamics of Aggregate Learning Curves.
Tristan M. Nixon, Stephen E. Fancsali, STEVE RITTER · Educational Data Mining · 2013
Mastery learning in intelligent tutoring systems produces a differential attrition of students over time, based on their levels of knowledge and ability. This results in a systematic bias when student data are aggregated to produce learning curves. We outline a formal framework, based on Bayesian Knowledge Tracing, to evaluate the impact of differential student attrition in mastery learning systems, and use simulations to investigate the impact of this effect in both homogeneous and mixed populations of learners.