Interface Design Issues for Teachable Agent Systems
Karun Viswanath, Ramya Balachandran, Mattie R. Kramer, and Bobby Bodenheimer
Abstract
This paper discusses improvements to the user interface of a system designed to promote
learning through teaching. This system, called Betty’s Brain, is an intelligent agent developed at
Vanderbilt University for studying the learning by teaching paradigm. In the Betty’s Brain system,
students teach a computer agent by creating a concept map using a visual interface. Students
themselves learn through the process of instructing the agent. Students aged 9-11 (U.S. fifth-grade)
in Nashville public schools participated in studies using this system. In this work, we improve and
evaluate the interface components of Betty’s Brain that enable the user to organize and input
problem solving knowledge about a domain for the purpose of instructing the intelligent agent. We
then conducted a comparative user study to evaluate our changes to the interface. Both qualitative
and quantitative improvements in the user’s performance are reported. These results should provide
useful guidance for designers of learning by teaching systems.
Bobby Bodenheimer 