An Evaluation of Self-explanation in a Programming Tutor

摘要

A controlled study was conducted in-natura to evaluate the effectiveness of presenting passive self-explanation questions in a problem-solving tutor on code-tracing. Data was collected from multiple institutions over three semesters using a tutor on selection statements: fall 2012-fall 2013. ANOVA and ANCOVA were used to analyze the collected data. After accounting for the additional time provided to test group students to answer self-explanation questions, test group was found to fare no better than control group on the number of concepts practiced, the pre-post change in score or the number of practice problems solved per practiced concept. It is speculated that this lack of difference might be attributable to self-efficacy issues, and that the features of tutors found to be effective in-vivo might need self-efficacy supports to also be effective in-natura.