Wizard-of-Oz-Driven Bootstrapping of a Socially Intelligent Tutoring Strategy
Jozef Tvarožek, Mária Bieliková · 2011
Abstract: This research investigates ways to bootstrap a socially intelligent tutoring agent using a Wizard-of-Oz design in which human teachers work as wizards. Students work in an interactive learning environment featuring problem solving and course notes activities, in either individual or group mode. They interact with the agent during off-task dialogs. This paper explores the effort needed on the part of classroom teachers in order to bootstrap an automated agent that can execute near-optimal tutoring actions within the environment. The proposed bootstrapping method can detect peers who will be able to interact well socially, and it can balance individual and social activities for students. Consequently, an automated tutor assisted by classroom teachers, rather than costly expert modeling, can be deployed in interactive learning environments in order to supplement and add to the attention a teacher is able to give students.