Dissertation Abstract

The Role of Intelligent Agency in Synthetic Instructor and Human Student Dialogue

Publication Number:  AAT9987579
Author:  Kroetz, Arthur William
School:  University of Southern California
Date:  1999
Pages:  193
Subject:  Educational Software,  Computer Science,  Artificial Intelligence

An enduring problem with computerized tutoring systems is the human-computer interface is often quite unengaging to the user. The ability to make human-computer interactions more life-like was investigated by studying an intelligent tutoring system (ITS) enhanced by the addition of intelligent agency in the form of an animated tutor. This pedagogical tutor named STEVE, inhabits a virtual reality simulation of a compressor control panel used aboard U.S. Navy ships. STEVE, embodied as an upper torso with moveable head, facial features, limited verbalizations, and arms, dynamically demonstrates the procedure to ready this compressor for the shipboard propulsion system. This simulation was evaluated using a single case study method with qualitative measures. Relevant publications, personal interviews, system documentation, the program environment, and the simulation procedure itself were examined. The simulation procedure studied 10 non-randomly chosen subjects interacting with the tutor. Subjects observed the STEVE tutor demonstrate the readiness procedure and then had to replicate the procedure themselves. Three constructs were chosen to elucidate the factors of animated character enhancement; tutor believability, active participation by the subject (constructivism), and what kind of mental models of the simulation were produced. Results of the experiment showed that even though STEVE lacked emotions, subjects deemed him believable to the level of being a sentient humanoid. Subjects also expressed freedom to complete the procedure in a natural, engaging manner but felt constrained to touch other controls not shown in the demonstration. The mental models formed by the subjects showed the lack of ability to remember details of the simulation and what was the overall function of the simulation. The presence of STEVE within the simulation elicits humanlike responses from the subjects, allows for efficient learning of a complex procedure, and provides for situated, active participation in a procedural learning environment.

This dissertation citation and abstract are published with permission of ProQuest Information and Learning. Further reproduction is prohibited without permission.