Research Methods Tutor
The AutoTutor system, developed at the University of Memphis with Art
Graesser, PI, was the first in a new breed of intelligent tutoring
systems: those that interact as human tutors do -- with natural language
dialog. AutoTutor was developed for the domain of Computer Literacy,
and has recently been "ported" to the domain of conceptual physics.
The behavior of the AutoTutor system is based on studies of human
tutors who were tutoring research methods in Psychology. We think
that this domain would provide a good challenge for extending the
capabilities of computerized dialog-based tutoring systems because of
the richness of the dialog structure and the "case-based" nature of
the domain. The primary goal of Research Methods Tutor is to
implement a dialog-based tutor that can be used in conjunction with a
college-level research methods class. The secondary goals deal with
extending the capabilities of the tutor to allow it to support
more complicated dialogs.
- What is the best way of integrating RMT into the curriculum of a
research methods class?
- Can a dialog-based tutor be provided efficiently via the web?
- Can the natural language understanding mechanism be improved
sufficiently to recognize student inputs more precisely?
- What type of user model is necessary to represent the state of the
learner's knowledge at any moment, and over the course of a term?
- How can the tutor generate more natural-sounding utterances?
An initial prototype of RMT is working with a curriculum script that
covers five topics from research methods courses.
We are currently pilot testing the system, and will evaluate it in the
classroom in the Winter term of 2004.
The initial domain topics for research methods were developed by Lori
Malatesta at the University of Edinburgh. A completely new curriculum
script which corresponds with the DePaul research methods currriculum
has been written by Elizabeth Arnott and David Allbritton in
the Psychology Department at DePaul. Technical development of the
system has been done by Jesse Efron, Oussama BenKhadra, and Peter
Last update: 20031103
Peter Hastings (email@example.com)