Title:
The Use of Natural Language Processing in Science Education Research
Abstract:
The barriers to using Natural Language Processing (NLP) in science education research have steadily decreased in recent years, driven by improved performance of everyday computers and the availability of numerous, often free, software tools. This development is beneficial because large amounts of verbal and textual data can now be summarized, visualized, and quantified. It expands the possibilities of content-analytic research, enables the identification of patterns in both new and existing datasets, and supports the re-analysis of previously collected data. It also makes it possible to create language-based interactive teaching and learning materials, for example through automated and dynamic feedback for learners while they are working on tasks.
At the same time, this development is associated with challenges. Despite the availability of very good documentation, the time required to become familiar with programming syntax for NLP applications is still considerably longer than for established methods of analysis. The complex and partly opaque computational procedures are highly sensitive to their initial and boundary conditions. In addition, reasonably accurate predictive models – for example for summative assessment – are inconceivable without human coding of tasks and carefully curated training data.
Drawing on examples from the field of chemistry education research, this presentation argues for exploring these exciting new tools. It concludes with a set of critical guiding questions intended to open a discussion of validity issues in both machine-supported and non-machine-oriented research designs.
About Marvin
Marvin Rost is a postdoctoral researcher in chemistry education at the Technical University of Munich. He previously worked as a postdoctoral researcher at the University of Vienna and completed his doctorate in chemistry education at Humboldt-Universität zu Berlin, where he investigated the intersection between quantitative modeling of competencies and epistemology. His academic background includes chemistry, philosophy of science, and teacher education.
His research focuses on chemistry teaching and learning, especially on modeling, scientific reasoning, and the use of digital and AI-based tools, preferably in longitudinal educational contexts. He works on questions related to how learners engage with chemistry and how these processes can be analyzed in classroom and laboratory settings via state-of-the-art methods such as natural language processing, and causal inference.
Where
Huygens building, HG02.053
