Large language models (LLMs) are increasingly used to generate educational and clinical material, yet their outputs can reproduce demographic stereotypes. A recent international study found substantial racial and gender misrepresentation in AI-generated clinical cases, but considered only the United States.
This project will investigate whether similar biases arise in an Aotearoa New Zealand context. The student will identify medical conditions for which reliable New Zealand epidemiological data are available, generate clinical vignettes using widely used LLMs, and compare the demographic profiles produced by the models with observed New Zealand distributions. Particular attention will be given to representation of Māori and other major ethnic groups in Aotearoa.
The project combines machine learning, statistical analysis and AI ethics. It would suit a student interested in data science, AI, health, or responsible technology and is intended to lead to a research publication.
Supervisors
Primary Supervisor: Elizabeth Stewart
Other Supervisor(s): Katharina Dost
Application and funding
This is a UC Commitments Summer Scholarship research project. Emailed applications will not be considered. Find out more here: University of Canterbury Scholarship Portal - UC Commitments Summer Scholarship
Final date for receiving applications
25 September 2026
