SimStudent: Authoring Expert Models by Tutoring
Proceedings of the Second Annual GIFT Users Symposium
2014
Abstract
One aim of the Generalized Intelligent Framework for Tutoring (GIFT) is to reduce the time and cost of authoring Intelligent Tutoring Systems. Recent work with SimStudent offers a promising approach to the efficient authoring of expert models and misconception libraries. SimStudent works by inducing general production rule models from author demonstrations and feedback. Importantly, the demonstration and feedback takes place directly in the tutor interface and requires no programming. Empirical results have shown that models induced by SimStudent fit student data better than models hand-authored by domain experts. Additionally, an analysis with the Goals, Operators, Methods, and Selection rules (GOMS) model showed that authoring with SimStudent is more efficient than authoring with current approaches, namely Example-Tracing. This paper reviews those results and provides an example of constructing a simple algebra tutor with SimStudent. This work with SimStudent presents several concepts that may be useful in the design and development of GIFT: modularization to allow for tutor authoring by non-programmers, generation of likely student misconceptions as a byproduct of expert-model creation, and methods for comparing and evaluating authoring tools.
BibTeX
@inproceedings{maclellan-gift-2014,
title = {SimStudent: Authoring Expert Models by Tutoring},
author = {MacLellan, Christopher J. and Stampfer Wiese, Eliane and Matsuda, Noboru and Koedinger, Kenneth R.},
booktitle = {Proceedings of the Second Annual GIFT Users Symposium},
year = {2014},
}