Non-intrusive certitude detection in computer-based instruction

This study examined two different certitude detection methods. Sixty-one physics teachers used a computer-based instructional program to discover rules applied to electric circuit problems. A problem presented on the screen required an attempted response. In one version of the program the subject rated the confidence in that response on a Likert scale. In another version of the program the subject wagered points that the response was correct. Response confidence and the correctness of response combined to define certitude. The time a subject viewed the displayed correct answer was feedback study time. Subsequent trial performance was the number of particular problem attempts before two consecutive high confidence of response correct answers were generated. Certitude generated predictive feedback study time and subsequent trial performance provided a basis for comparing certitude detection methods. Wagering behavior produced more highly correlated predictive functions for feedback study time. Subsequent trials failed to show a significant difference, using predictive correlations, between the Likert and Wager methods. A proposed modification to the Feedback Instructional Model (Kulhavy and Stock, 1989) may account for the observed differences between the certitude detection methods.