How large language models can help to categorise master’s theses in teacher education: results from a pilot study

In many countries, master’s theses are an integral part of teacher education. While previous research categorised the thesis texts manually, we pilot if artificial intelligence (AI) large language models (LLMs) can be utilised to create a large-scale overview of master’s theses’ characteristics efficiently. Specifically, we developed a coding instrument to categorise master’s theses and (1) investigate if an LLM performs as well as teacher educators at the text coding task, and (2) map general characteristics of 278 theses from a Norwegian university. The LLM (GPT-4 Turbo) showed promising potential to categorise the theses similarly to teacher educators. Most theses were qualitative, small-scale interview and/or classroom observation studies. We discuss implications (1) for the potential of LLMs as research tools in the field of teacher education and (2) for master’s theses as a research topic to foster high-quality teacher education.