Presenters:
Ido Roll, Carnegie Mellon University
Vincent Aleven, Carnegie Mellon University
Bruce M. McLaren, Carnegie Mellon University
Eunjeong Ryu, Carnegie Mellon University
Kenneth R. Koedinger, Carnegie Mellon University
Abstract: Intelligent tutoring systems typically offer help to students in the form of context-specific hints. However, students often fail to make optimal use of this information. Instead, they tend to make a wide-spectrum of errors concerning help: sometimes avoiding needed help, or abusing redundant information. Our hypothesis is that students lack the requisite metacognitive knowledge that would enable them to use the information provided by intelligent tutors more effectively. In this poster we describe project aimed at improving this important aspect of students' metacognitive decisions while using Intelligent Tutoring Systems - their help-seeking behavior.
We have constructed a computer-based system called the "Help-Tutor" - a domain-independent tutor-agent that can be added as an adjunct to any Cognitive Tutor. The Help-Tutor teaches better help-seeking skills by tracing students' actions on a (meta)cognitive model and giving immediate tailored feedback to students after they perform a help-seeking related error.
In a classroom evaluation of the Help-Tutor, The Help-Tutor captured help-seeking errors that were associated with poorer learning. The faulty behavior captured by the Help-Tutor also transferred to a paper-and-pencil environment, suggesting that such behavior is not tied to specific environment. While working with the tutor, students improved their help-seeking actions and committed less errors. However, we do not yet have firm evidence that students learned the intended help-seeking skills, or learned the domain knowledge better.
A new version of the tutor that includes a self-reflection component and explicit help-seeking instruction, complementary to the metacognitive feedback, is now being evaluated.