Assessing Physics Students' Scientific Argumentation using Natural Language Processing

Scientific argumentation is an important science and engineering practice and a necessary 21st Century workforce skill. Due to the nature of large enrollment classes, it is difficult to individually assess students and provide feedback on their argumentation. The recent developments in Natural Language Processing (NLP) and Machine Learning (ML) may provide a solution. In this study we investigate methods using NLP and ML to assess and understand students' argumentation. Specifically, we investigate the use of topic modeling to analyze student essays of argumentation after solving a problem in the recitation section of an introductory calculus-based physics course four semesters. We report on the emergent themes present in each semester.