IASS Webinar 42: Leveraging AI for Survey Research

Abstract: This presentation scrutinises the transformative potential of Large Language Models (LLMs) in survey research, focusing on three critical areas: questionnaire design, synthetic data creation, and the role of LLMs as qualitative interviewers. In the domain of questionnaire design, the lecture delves into if and how LLMs can construct contextually accurate and highly effective survey items. However, there are valid concerns about the model’s understanding and potential biases, which we will critically evaluate. She also discusses LLMs’ ability to fabricate synthetic data, preserving core statistical properties whilst ensuring privacy. Here too, the ethical implications and the potential for misuse of this capability pose challenges that need to be addressed. Lastly, the lecture explores how LLMs, with their human-like conversational ability, can act as qualitative interviewers, allowing in-depth information gathering at scale. Yet, questions about their ability to fully capture the complexity and subtleties of human interaction and response also remain. The underlying theme of this talk is the question on how research in this space should be structured.