EVALUATING STUDENT ANXIETY AS A PREDICTOR OF STEM PERFORMANCE USING STORYTELLING AND MACHINE LEARNING

Abstract

– Anxiety can be a formidable barrier in STEM. Incorporating anxiety-informed pedagogy presents a potential approach to overcome this challenge, such as ‘Breadboardia,’ a graphical storybook to learn electronics. In this paper, machine learning (ML) was employed to predict students’ STEM performance based on their self-scored anxiety levels. Throughout, ML concepts are introduced to encourage their adoption in educational research. The instrument was administered to 200 high-school girls prior, immediately thereafter, and two weeks following exposure to ‘Breadboardia.’ A sentiment analysis was also performed to evaluate student perceptions of the storybook. Preliminary results suggest that the participants remained largely neutral towards the storybook, independent of their performance on technical questions. Initial prediction model results are promising, with average errors of 1.5% on training data and 23.1% on testing data. These models elucidate how ML techniques can be leveraged in engineering pedagogy and inform the development of targeted interventions that enhance engagement.