Key points are not available for this paper at this time.
This study aims to identify effective affective states and behaviors of middle-school students' online mathematics learning in predicting their choices to study science, technology, engineering, and mathematics (STEM) in higher education based on a positive-affect-to-success hypothesis. The dataset (591 students and 316,974 actions) was obtained from the ASSISTments project. In the ASSISTments intelligent tutoring system, students completed mathematical problem-solving tasks, and the data was processed to infer their action-level affective states and behaviors, which were averaged to form student-level measures. The students' future STEM choice was predicted by the student- and action-level affective states and behaviors using logistic regression (LR), ordinary least squares regressions with standardized scores (ORz), and random forest with permutation importance and SHAP values (RFPS). The results revealed that student- and action-level gaming behavior consistently predict STEM choice. In addition to gaming, female students are more likely to study STEM if they are less bored and more off-task, and male students if more concentrated and less frustrated. ORz generates theoretically plausible results and identifies sufficiently distinguishable affective states and behaviors. Suggestions for educational practice and research are provided for adaptive teaching.
Mei-Shiu Chiu (Sun,) studied this question.