ABSTRACT
An important task in tutoring is choosing which task the student should attempt next (Corbett, Koedinger, & Anderson, 1997; Kalchman, & Koedinger, 2005). This decision should be efficient for whomever makes it-the tutor or student, especially a young student. Its outcome should be effective in educational terms. To summarize with a pun, the “choice task” of deciding what to work on should efficiently yield a “choice task” in the sense of helping the student learn. This paper analyzes the efficiency and effectiveness of task choice in the context of a tutor that listens to children read aloud: specifically, the Reading Tutor from Project LISTEN, further described in the chapter by Mostow et al. (this volume). This tutor uses automated speech recognition (ASR) to detect and diagnose the reading problems of children as they read stories aloud from the computer screen. The ASR techniques, and the methods by which we developed and tested them, have been covered elsewhere (e.g., Mostow & Aist, 1999a; Mostow, Roth, Hauptmann, & Kane, 1994). Here, we describe design features added to successive versions of the Reading Tutor to improve efficiency and effectiveness of choosing stories, defining efficiency as the time to pick a story and effectiveness in terms of exposing students to new material. We follow with a quantitative evaluation of the resulting versions of the tutor, making use of data collected through the speech recognizer, and conclude with lessons learned en route.
