Using Topic Modeling to Extract Pre-Service Teachers' Understandings of Computational Thinking From Their Coding Reflections

Contribution: This paper employs the automatic scoring of short essays as a novel way to determine pre-service teachers' knowledge of and attitudes toward computational thinking (CT) from their written reflections. Implications about designing CT courses for pre-service teachers are discussed. Background: CT is an essential 21st-century competency that supports the development of problem-solving skills. Inspired by computing science problem-solving practices, CT should transcend disciplines, but few universities or colleges include CT courses or CT content in their core courses. It is also difficult to know what pre-service teachers think about CT and their role in promoting it. Research Questions: Do pre-service teachers' coding reflections reveal any important information about their knowledge of, skills in, and attitudes toward CT? Methodology: Traditional qualitative techniques based on human raters are impractical in analyzing hundreds of essays. Topic modeling, an unsupervised machine learning modeling technique, was employed to extract topical features from participants' reflections. In one section of an undergraduate Introduction to Educational Technology course offered at a large university in Western Canada, n = 139 pre-service teachers wrote a short reflection on their experience following a 20 h Accelerated Intro to Computer Science Code.org course. Topics were identified by analyzing contextual trends in participants' written reflections. Findings: Results showed that pre-service teachers' reflections included CT concepts, practices, and perspectives. Specifically, participants connected the coding activity to prior knowledge and experiences.