The INFACT Gnome Environment: A Platform for Autonomous Agents Generating Automatic Student Assessments and Feedback

by
Nick Benson

The majority of on-line learning environments still provide a very limited set of features for analyzing and assessing student data. The ability to rapidly create and deploy modular applications implementing a variety of assessment techniques is potentially very valuable. With this idea in mind, we have developed a platform for the easy creation and deployment of autonomous assessment agents or "Gnomes," written in Java and operating independently within INFACT, an on-line learning environment developed at the University of Washington.

The Gnome Environment provides a uniform interface for easy access to data such as web-forum posts, student sketches, logged application events, manual instructor markup and student preconception databases within INFACT. Additionally, Gnomes receive real-time notification of events taking place within the system (student and instructor activity, automatic assessment, etc.)

One possible application for Gnomes is the generation of rapid, dynamic responses for students based on their activity within the INFACT environment. Responses can range from feedback on whether or not the student submitted a correct answer to additional questions that may help the system to form a better assessment of the student's understanding of the topic at hand.

In this talk, I will present an overview of the Gnome system, as well as two example Gnomes that analyze student-generated data: one implementing a Bayesian text classifier and the other providing automatic feedback to students based on the content of web-forum posts.

Advised by Steve Tanimoto

EE1 037
Wednesday
April 21, 2004
4:30 - 5:20 pm