Evaluating a simulated student using real students
data for training and testing
Matsuda, N., Cohen, W. W., Sewall, J., Lacerda, G., & Koedinger, K. R.
(2007). Evaluating a simulated student using real students data for training
and testing. In C. Conati, K. McCoy & G. Paliouras (Eds.), Proceedings of
the international conference on User Modeling (LNAI 4511) (pp. 107-116).
Berlin, Heidelberg: Springer
Abstract: The Simulated Students are machine-learning agents that
learn cognitive skills by demonstration. They were originally developed as a
building block for the Cognitive Tutor Authoring Tools (CTAT) so that the
authors do not have to build a cognitive model by hand, but instead simply
demonstrate solutions for the Simulated Students to automatically generate a
cognitive model. The Simulated-Student technology could then be used to
model human students' performance as well. To evaluate applicability of the
Simulated Students as a tool for modeling real students, we applied the
Simulated Students to a genuine learning log gathered from classroom
instructions using the Algebra I Cognitive Tutor. Such data can be seen as
the human students' "demonstrations" on how to solve problems. The results
from the empirical study show that the Simulated Students can indeed model
human students' performances. After training on 20 problems solved by a set
of human students, a cognitive model generated by Simulated Students
explained 82% of the problem-solving steps performed correctly by another
set of human students.
PDF file (228KB)