Exploring large language models for persona construction: a case study in educational ergonomics

Abstract

This study investigates the potential of large language models (LLMs), specifically ChatGPT 4o, to support the creation of data-driven personas using a real-world dataset on ergonomic awareness among educators in Ho Chi Minh City, Vietnam. The proposed workflow combined automated data cleaning, clustering using k-prototypes algorithm, statistical testing, and narrative persona generation. Three distinct educator profiles were identified, each differing in age, teaching status, teaching experience, and musculoskeletal discomfort. These personas highlight relevant risk factors and opportunities for targeted ergonomic interventions in educational settings. The process demonstrated that LLMs can lower the technical barriers typically associated with quantitative methods and narrative construction, making persona development more accessible in low-resource environments. Although the dataset was limited in size and depth, and expert validation of statistical procedures was not performed, the approach proved feasible and reproducible. The study suggests that combining clustering algorithms with LLM-supported interpretation can bridge analytical rigor and human-centered storytelling. This hybrid method holds promise for design researchers and practitioners seeking scalable, interpretable, and empathetic tools for understanding user groups in small organizations and businesses. It also opens new directions for applying generative AI in occupational health, education and information design contexts where data interpretation and stakeholder communication are essential.