Classifying the quality of robotics-enhanced lesson plans using motivation variables, word count, and sentiment analysis of reflections

While teaching lesson planning, it is critical to uncover signs that additional support is needed. The conceptual framework of this study grounded in two fundamental perspectives of teachers as learners and teachers as designers guided our research inquires to motivational factors and reflection processes. In this study, a discriminant analysis model incorporating motivation variables, word count, and sentiment analysis of reflections was used to classify lesson plan quality. Next, support vector machines identified misclassified lesson plans. Finally, we used ordinal logistic regression to estimate Betas for each predictor. The strongest classifiers were sentiment analysis scores and word counts. The model was fairly accurate, yielding a 12.82%, a 7.69%, and a 23.08% misclassification rate. Ordinal logistic regression indicated that performance goal orientation was a significant predictor of front-end analysis quality.