Abstract: The research reported in this paper focuses on the hypothesis that
an intelligent tutoring system that provides guidance with respect to students'
meta-cognitive abilities can help them to become better learners. Our strategy
is to extend a Cognitive Tutor (Anderson, Corbett, Koedinger, & Pelletier,
1995) so that it not only helps students acquire domain-specific skills, but
also develop better general help-seeking strategies. In developing the Help
Tutor, we used the same Cognitive Tutor technology at the meta-cognitive level
that has been proven to be very effective at the cognitive level. A key
challenge is to develop a model of how students should use a Cognitive Tutor's
help facilities. We created a preliminary model, implemented by 57 production
rules that capture both effective and ineffective help-seeking behavior. As a
first test of the model's efficacy, we used it off-line to evaluate students'
help-seeking behavior in an existing data set of student-tutor interactions. We
then refined the model based on the results of this analysis. Finally, we
conducted a pilot study with the Help Tutor involving four students. During one
session, we saw a statistically significant reduction in students'
meta-cognitive error rate, as determined by the Help Tutor's model. These
preliminary results inspire confidence as we gear up for a larger-scale
controlled experiment to evaluate whether tutoring on help seeking has a
positive effect on students' learning outcomes.