LLM-Assisted Sentiment Analysis Enhances Mixed-Methods Education Research

A new study published on arXiv (2605.27403) explores the use of large language models (LLMs) for sentiment analysis in qualitative education research. The researchers analyzed 151 written reflection assignments from a study abroad program, employing LLM-assisted sentiment analysis to enable longitudinal mixed-methods research that combines computational and thematic analyses. Statistical testing was used to quantitatively compare sentiment differences across seven student variables, demonstrating that LLMs can efficiently process qualitative data that is typically time-consuming to analyze manually. The study highlights the potential of LLMs as research assistants for comparing findings across multiple participant groups, overcoming the limitation of previous studies that often compare only one variable at a time.